⬆️ Update ggerganov/llama.cpp (#2603)

Signed-off-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: mudler <2420543+mudler@users.noreply.github.com>

.dockerignore

@@ -1,6 +1,16 @@
 .idea
+.github
+.vscode
 models
 examples/chatbot-ui/models
 examples/rwkv/models
 examples/**/models
-Dockerfile
+Dockerfile*
+__pycache__
+
+# SonarQube
+.scannerwork
+
+# backend virtual environments
+**/venv
+backend/python/**/source

.editorconfig

@@ -0,0 +1,31 @@
+
+root = true
+
+[*]
+indent_style = space
+indent_size = 2
+end_of_line = lf
+charset = utf-8
+trim_trailing_whitespace = true
+insert_final_newline = true
+
+[*.go]
+indent_style = tab
+
+[Makefile]
+indent_style = tab
+
+[*.proto]
+indent_size = 2
+
+[*.py]
+indent_size = 4
+
+[*.js]
+indent_size = 2
+
+[*.yaml]
+indent_size = 2
+
+[*.md]
+trim_trailing_whitespace = false

.env

@@ -1,33 +1,33 @@
 ## Set number of threads.
 ## Note: prefer the number of physical cores. Overbooking the CPU degrades performance notably.
-# THREADS=14
+# LOCALAI_THREADS=14
 
 ## Specify a different bind address (defaults to ":8080")
-# ADDRESS=127.0.0.1:8080
+# LOCALAI_ADDRESS=127.0.0.1:8080
 
 ## Default models context size
-# CONTEXT_SIZE=512
+# LOCALAI_CONTEXT_SIZE=512
 #
 ## Define galleries.
 ## models will to install will be visible in `/models/available`
-# GALLERIES=[{"name":"model-gallery", "url":"github:go-skynet/model-gallery/index.yaml"}]
+# LOCALAI_GALLERIES=[{"name":"localai", "url":"github:mudler/LocalAI/gallery/index.yaml@master"}]
 
 ## CORS settings
-# CORS=true
-# CORS_ALLOW_ORIGINS=*
+# LOCALAI_CORS=true
+# LOCALAI_CORS_ALLOW_ORIGINS=*
 
 ## Default path for models
 #
-# MODELS_PATH=/models
+# LOCALAI_MODELS_PATH=/models
 
 ## Enable debug mode
-# DEBUG=true
+# LOCALAI_LOG_LEVEL=debug
 
 ## Disables COMPEL (Diffusers)
 # COMPEL=0
 
 ## Enable/Disable single backend (useful if only one GPU is available)
-# SINGLE_ACTIVE_BACKEND=true
+# LOCALAI_SINGLE_ACTIVE_BACKEND=true
 
 ## Specify a build type. Available: cublas, openblas, clblas.
 ## cuBLAS: This is a GPU-accelerated version of the complete standard BLAS (Basic Linear Algebra Subprograms) library. It's provided by Nvidia and is part of their CUDA toolkit.
@@ -46,13 +46,13 @@
 # GO_TAGS=stablediffusion
 
 ## Path where to store generated images
-# IMAGE_PATH=/tmp
+# LOCALAI_IMAGE_PATH=/tmp/generated/images
 
 ## Specify a default upload limit in MB (whisper)
-# UPLOAD_LIMIT
+# LOCALAI_UPLOAD_LIMIT=15
 
 ## List of external GRPC backends (note on the container image this variable is already set to use extra backends available in extra/)
-# EXTERNAL_GRPC_BACKENDS=my-backend:127.0.0.1:9000,my-backend2:/usr/bin/backend.py
+# LOCALAI_EXTERNAL_GRPC_BACKENDS=my-backend:127.0.0.1:9000,my-backend2:/usr/bin/backend.py
 
 ### Advanced settings ###
 ### Those are not really used by LocalAI, but from components in the stack ###
@@ -71,19 +71,24 @@
 ### Define the number of parallel LLAMA.cpp workers (Defaults to 1)
 # LLAMACPP_PARALLEL=1
 
+### Define a list of GRPC Servers for llama-cpp workers to distribute the load
+# https://github.com/ggerganov/llama.cpp/pull/6829
+# https://github.com/ggerganov/llama.cpp/blob/master/examples/rpc/README.md
+# LLAMACPP_GRPC_SERVERS=""
+
 ### Enable to run parallel requests
-# PARALLEL_REQUESTS=true
+# LOCALAI_PARALLEL_REQUESTS=true
 
 ### Watchdog settings
 ###
 # Enables watchdog to kill backends that are inactive for too much time
-# WATCHDOG_IDLE=true
-#
-# Enables watchdog to kill backends that are busy for too much time
-# WATCHDOG_BUSY=true
+# LOCALAI_WATCHDOG_IDLE=true
 #
 # Time in duration format (e.g. 1h30m) after which a backend is considered idle
-# WATCHDOG_IDLE_TIMEOUT=5m
+# LOCALAI_WATCHDOG_IDLE_TIMEOUT=5m
+#
+# Enables watchdog to kill backends that are busy for too much time
+# LOCALAI_WATCHDOG_BUSY=true
 #
 # Time in duration format (e.g. 1h30m) after which a backend is considered busy
-# WATCHDOG_BUSY_TIMEOUT=5m
+# LOCALAI_WATCHDOG_BUSY_TIMEOUT=5m

.github/bump_docs.sh

@@ -2,6 +2,6 @@
 set -xe
 REPO=$1
 
-LATEST_TAG=$(curl -s "https://api.github.com/repos/$REPO/releases/latest" | jq -r '.name')
+LATEST_TAG=$(curl -s "https://api.github.com/repos/$REPO/releases/latest" | jq -r '.tag_name')
 
 cat <<< $(jq ".version = \"$LATEST_TAG\"" docs/data/version.json) > docs/data/version.json

.github/checksum_checker.sh

@@ -0,0 +1,126 @@
+#!/bin/bash
+# This scripts needs yq and huggingface_hub to be installed
+# to install hugingface_hub run pip install huggingface_hub
+
+# Path to the input YAML file
+input_yaml=$1
+
+# Function to download file and check checksum using Python
+function check_and_update_checksum() {
+    model_name="$1"
+    file_name="$2"
+    uri="$3"
+    old_checksum="$4"
+    idx="$5"
+
+    # Download the file and calculate new checksum using Python
+    new_checksum=$(python3 -c "
+import hashlib
+from huggingface_hub import hf_hub_download, get_paths_info
+import requests
+import sys
+import os
+
+uri = '$uri'
+file_name = uri.split('/')[-1]
+
+# Function to parse the URI and determine download method
+# Function to parse the URI and determine download method
+def parse_uri(uri):
+    if uri.startswith('huggingface://'):
+        repo_id = uri.split('://')[1]
+        return 'huggingface', repo_id.rsplit('/', 1)[0]
+    elif 'huggingface.co' in uri:
+        parts = uri.split('/resolve/')
+        if len(parts) > 1:
+            repo_path = parts[0].split('https://huggingface.co/')[-1]
+            return 'huggingface', repo_path
+    return 'direct', uri
+
+def calculate_sha256(file_path):
+    sha256_hash = hashlib.sha256()
+    with open(file_path, 'rb') as f:
+        for byte_block in iter(lambda: f.read(4096), b''):
+            sha256_hash.update(byte_block)
+    return sha256_hash.hexdigest()
+
+download_type, repo_id_or_url = parse_uri(uri)
+
+new_checksum =  None
+
+# Decide download method based on URI type
+if download_type == 'huggingface':
+    # Use HF API to pull sha
+    for file in get_paths_info(repo_id_or_url, [file_name], repo_type='model'):
+        try:
+            new_checksum = file.lfs.sha256
+            break
+        except Exception as e:
+            print(f'Error from Hugging Face Hub: {str(e)}', file=sys.stderr)
+            sys.exit(2)
+    if new_checksum is None:
+        try:
+            file_path = hf_hub_download(repo_id=repo_id_or_url, filename=file_name)
+        except Exception as e:
+            print(f'Error from Hugging Face Hub: {str(e)}', file=sys.stderr)
+            sys.exit(2)
+else:
+    response = requests.get(repo_id_or_url)
+    if response.status_code == 200:
+        with open(file_name, 'wb') as f:
+            f.write(response.content)
+        file_path = file_name
+    elif response.status_code == 404:
+        print(f'File not found: {response.status_code}', file=sys.stderr)
+        sys.exit(2)
+    else:
+        print(f'Error downloading file: {response.status_code}', file=sys.stderr)
+        sys.exit(1)
+
+if new_checksum is None:
+    new_checksum = calculate_sha256(file_path)
+    print(new_checksum)
+    os.remove(file_path)
+else:
+    print(new_checksum)
+
+")
+
+    if [[ "$new_checksum" == "" ]]; then
+        echo "Error calculating checksum for $file_name. Skipping..."
+        return
+    fi
+
+    echo "Checksum for $file_name: $new_checksum"
+
+    # Compare and update the YAML file if checksums do not match
+    result=$?
+    if [[ $result -eq 2 ]]; then
+        echo "File not found, deleting entry for $file_name..."
+        # yq eval -i "del(.[$idx].files[] | select(.filename == \"$file_name\"))" "$input_yaml"
+    elif [[ "$old_checksum" != "$new_checksum" ]]; then
+        echo "Checksum mismatch for $file_name. Updating..."
+        yq eval -i "del(.[$idx].files[] | select(.filename == \"$file_name\").sha256)" "$input_yaml"
+        yq eval -i "(.[$idx].files[] | select(.filename == \"$file_name\")).sha256 = \"$new_checksum\"" "$input_yaml"
+    elif [[ $result -ne 0 ]]; then
+        echo "Error downloading file $file_name. Skipping..."
+    else
+        echo "Checksum match for $file_name. No update needed."
+    fi
+}
+
+# Read the YAML and process each file
+len=$(yq eval '. | length' "$input_yaml")
+for ((i=0; i<$len; i++))
+do
+    name=$(yq eval ".[$i].name" "$input_yaml")
+    files_len=$(yq eval ".[$i].files | length" "$input_yaml")
+    for ((j=0; j<$files_len; j++))
+    do
+        filename=$(yq eval ".[$i].files[$j].filename" "$input_yaml")
+        uri=$(yq eval ".[$i].files[$j].uri" "$input_yaml")
+        checksum=$(yq eval ".[$i].files[$j].sha256" "$input_yaml")
+        echo "Checking model $name, file $filename. URI = $uri, Checksum = $checksum"
+        check_and_update_checksum "$name" "$filename" "$uri" "$checksum" "$i"
+    done
+done

.github/ci/modelslist.go

@@ -0,0 +1,297 @@
+package main
+
+import (
+	"fmt"
+	"html/template"
+	"io/ioutil"
+	"os"
+
+	"gopkg.in/yaml.v3"
+)
+
+var modelPageTemplate string = `
+<!DOCTYPE html>
+<html>
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>LocalAI models</title>
+    <link href="https://cdnjs.cloudflare.com/ajax/libs/flowbite/2.3.0/flowbite.min.css" rel="stylesheet" />
+    <script src="https://cdn.jsdelivr.net/npm/vanilla-lazyload@19.1.3/dist/lazyload.min.js"></script>
+
+    <link
+    rel="stylesheet"
+    href="https://cdn.jsdelivr.net/gh/highlightjs/cdn-release@11.8.0/build/styles/default.min.css"
+  />
+    <script
+    defer
+    src="https://cdn.jsdelivr.net/gh/highlightjs/cdn-release@11.8.0/build/highlight.min.js"
+  ></script>
+    <script
+    defer
+    src="https://cdn.jsdelivr.net/npm/alpinejs@3.x.x/dist/cdn.min.js"
+  ></script>
+  <script
+    defer
+    src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"
+  ></script>
+  <script
+    defer
+    src="https://cdn.jsdelivr.net/npm/dompurify@3.0.6/dist/purify.min.js"
+  ></script>
+
+  <link href="/static/general.css" rel="stylesheet" />
+    <link href="https://fonts.googleapis.com/css2?family=Inter:wght@400;600;700&family=Roboto:wght@400;500&display=swap" rel="stylesheet">
+    <link
+    href="https://fonts.googleapis.com/css?family=Roboto:300,400,500,700,900&display=swap"
+    rel="stylesheet" />
+  <link
+    rel="stylesheet"
+    href="https://cdn.jsdelivr.net/npm/tw-elements/css/tw-elements.min.css" />
+  <script src="https://cdn.tailwindcss.com/3.3.0"></script>
+  <script>
+    tailwind.config = {
+      darkMode: "class",
+      theme: {
+        fontFamily: {
+          sans: ["Roboto", "sans-serif"],
+          body: ["Roboto", "sans-serif"],
+          mono: ["ui-monospace", "monospace"],
+        },
+      },
+      corePlugins: {
+        preflight: false,
+      },
+    };
+  </script>
+    <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.1.1/css/all.min.css">
+    <script src="https://unpkg.com/htmx.org@1.9.12" integrity="sha384-ujb1lZYygJmzgSwoxRggbCHcjc0rB2XoQrxeTUQyRjrOnlCoYta87iKBWq3EsdM2" crossorigin="anonymous"></script>
+</head>
+
+<body class="bg-gray-900 text-gray-200">
+<div class="flex flex-col min-h-screen">
+
+<nav class="bg-gray-800 shadow-lg">
+    <div class="container mx-auto px-4 py-4">
+        <div class="flex items-center justify-between">
+            <div class="flex items-center">
+                <a href="/" class="text-white text-xl font-bold"><img src="https://github.com/go-skynet/LocalAI/assets/2420543/0966aa2a-166e-4f99-a3e5-6c915fc997dd" alt="LocalAI Logo" class="h-10 mr-3 border-2 border-gray-300 shadow rounded"></a>
+                <a href="/" class="text-white text-xl font-bold">LocalAI</a>
+            </div>
+            <!-- Menu button for small screens -->
+            <div class="lg:hidden">
+                <button id="menu-toggle" class="text-gray-400 hover:text-white focus:outline-none">
+                    <i class="fas fa-bars fa-lg"></i>
+                </button>
+            </div>
+            <!-- Navigation links -->
+            <div class="hidden lg:flex lg:items-center lg:justify-end lg:flex-1 lg:w-0">
+                <a href="https://localai.io" class="text-gray-400 hover:text-white px-3 py-2 rounded" target="_blank" ><i class="fas fa-book-reader pr-2"></i> Documentation</a>
+            </div>
+        </div>
+        <!-- Collapsible menu for small screens -->
+        <div class="hidden lg:hidden" id="mobile-menu">
+            <div class="pt-4 pb-3 border-t border-gray-700">
+                
+                <a href="https://localai.io" class="block text-gray-400 hover:text-white px-3 py-2 rounded mt-1" target="_blank" ><i class="fas fa-book-reader pr-2"></i> Documentation</a>
+               
+            </div>
+        </div>
+    </div>
+</nav>
+
+<style>
+  .is-hidden {
+	display: none;
+	  }
+</style>
+
+<div class="container mx-auto px-4 flex-grow">
+
+<div class="models mt-12">
+	<h2 class="text-center text-3xl font-semibold text-gray-100">
+	LocalAI model gallery list </h2><br>
+
+	<h2 class="text-center text-3xl font-semibold text-gray-100">
+
+	 🖼️ Available {{.AvailableModels}} models</i> <a href="https://localai.io/models/" target="_blank" >
+			<i class="fas fa-circle-info pr-2"></i>
+		</a></h2> 
+
+	<h3>	  
+	Refer to the Model gallery <a href="https://localai.io/models/" target="_blank" ><i class="fas fa-circle-info pr-2"></i></a> for more information on how to use the models with LocalAI.<br>
+
+	You can install models with the CLI command <code>local-ai models install <model-name></code>. or by using the WebUI.
+	</h3>
+  
+	<input class="form-control appearance-none block w-full mt-5 px-3 py-2 text-base font-normal text-gray-300 pb-2 mb-5 bg-gray-800 bg-clip-padding border border-solid border-gray-600 rounded transition ease-in-out m-0 focus:text-gray-300 focus:bg-gray-900 focus:border-blue-500 focus:outline-none" type="search" 
+	id="searchbox" placeholder="Live search keyword..">
+	  <div class="dark grid grid-cols-1 grid-rows-1 md:grid-cols-3 block rounded-lg shadow-secondary-1 dark:bg-surface-dark">
+		{{ range $_, $model := .Models }}
+		<div class="box me-4 mb-2 block rounded-lg bg-white shadow-secondary-1  dark:bg-gray-800 dark:bg-surface-dark dark:text-white text-surface pb-2">
+		<div>
+		    {{ $icon := "https://upload.wikimedia.org/wikipedia/commons/6/65/No-Image-Placeholder.svg" }}
+			{{ if $model.Icon }}
+	  		{{ $icon = $model.Icon }}
+	  		{{ end }}
+			<div class="flex justify-center items-center">
+				<img data-src="{{ $icon }}" alt="{{$model.Name}}" class="rounded-t-lg max-h-48 max-w-96 object-cover mt-3 lazy">
+			</div>
+	  		<div class="p-6 text-surface dark:text-white">
+				<h5 class="mb-2 text-xl font-medium leading-tight">{{$model.Name}}</h5>
+				
+				   
+				<p class="mb-4 text-base truncate">{{ $model.Description }}</p>
+		
+			</div>
+			<div class="px-6 pt-4 pb-2">
+
+      <!-- Modal toggle -->
+      <button data-modal-target="{{ $model.Name}}-modal" data-modal-toggle="{{ $model.Name }}-modal" class="block text-white bg-blue-700 hover:bg-blue-800 focus:ring-4 focus:outline-none focus:ring-blue-300 font-medium rounded-lg text-sm px-5 py-2.5 text-center dark:bg-blue-600 dark:hover:bg-blue-700 dark:focus:ring-blue-800" type="button">
+        More info
+      </button>
+
+    <!-- Main modal -->
+    <div id="{{ $model.Name}}-modal" tabindex="-1" aria-hidden="true" class="hidden overflow-y-auto overflow-x-hidden fixed top-0 right-0 left-0 z-50 justify-center items-center w-full md:inset-0 h-[calc(100%-1rem)] max-h-full">
+        <div class="relative p-4 w-full max-w-2xl max-h-full">
+            <!-- Modal content -->
+            <div class="relative bg-white rounded-lg shadow dark:bg-gray-700">
+                <!-- Modal header -->
+                <div class="flex items-center justify-between p-4 md:p-5 border-b rounded-t dark:border-gray-600">
+                    <h3 class="text-xl font-semibold text-gray-900 dark:text-white">
+                        {{ $model.Name}}
+                    </h3>
+                    <button type="button" class="text-gray-400 bg-transparent hover:bg-gray-200 hover:text-gray-900 rounded-lg text-sm w-8 h-8 ms-auto inline-flex justify-center items-center dark:hover:bg-gray-600 dark:hover:text-white" data-modal-hide="{{$model.Name}}-modal">
+                        <svg class="w-3 h-3" aria-hidden="true" xmlns="http://www.w3.org/2000/svg" fill="none" viewBox="0 0 14 14">
+                            <path stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="m1 1 6 6m0 0 6 6M7 7l6-6M7 7l-6 6"/>
+                        </svg>
+                        <span class="sr-only">Close modal</span>
+                    </button>
+                </div>
+                <!-- Modal body -->
+                <div class="p-4 md:p-5 space-y-4">
+                    <div class="flex justify-center items-center">
+                    <img data-src="{{ $icon }}" alt="{{$model.Name}}" class="lazy rounded-t-lg max-h-48 max-w-96 object-cover mt-3">
+                  </div>
+
+                    <p class="text-base leading-relaxed text-gray-500 dark:text-gray-400">
+                    {{ $model.Description }}
+
+                    </p>
+                    
+                    <p class="text-base leading-relaxed text-gray-500 dark:text-gray-400">
+                    To install the model with the CLI, run: <br>
+                    <code> local-ai models install {{$model.Name}} </code> <br>
+
+                    <hr>
+                    See also <a href="https://localai.io/models/" target="_blank" >
+                    Installation <i class="fas fa-circle-info pr-2"></i>
+                    </a> to see how to install models with the REST API.
+                    </p>
+
+                    <p class="text-base leading-relaxed text-gray-500 dark:text-gray-400">
+                    <ul>
+                    {{ range $_, $u := $model.URLs }}
+                    <li><a href="{{ $u }}" target=_blank><i class="fa-solid fa-link"></i> {{ $u }}</a></li>
+                    {{ end }}  
+                    </ul>
+                    </p>
+                </div>
+                <!-- Modal footer -->
+                <div class="flex items-center p-4 md:p-5 border-t border-gray-200 rounded-b dark:border-gray-600">
+                    <button data-modal-hide="{{ $model.Name}}-modal" type="button" class="py-2.5 px-5 ms-3 text-sm font-medium text-gray-900 focus:outline-none bg-white rounded-lg border border-gray-200 hover:bg-gray-100 hover:text-blue-700 focus:z-10 focus:ring-4 focus:ring-gray-100 dark:focus:ring-gray-700 dark:bg-gray-800 dark:text-gray-400 dark:border-gray-600 dark:hover:text-white dark:hover:bg-gray-700">Close</button>
+                </div>
+            </div>
+        </div>
+    </div>
+
+
+			</div>
+		</div>
+		</div>
+		{{ end }}      
+
+		</div>
+  </div>
+</div>
+
+<script>
+var lazyLoadInstance = new LazyLoad({
+  // Your custom settings go here
+});
+
+let cards = document.querySelectorAll('.box')
+    
+function liveSearch() {
+    let search_query = document.getElementById("searchbox").value;
+    
+    //Use innerText if all contents are visible
+    //Use textContent for including hidden elements
+    for (var i = 0; i < cards.length; i++) {
+        if(cards[i].textContent.toLowerCase()
+                .includes(search_query.toLowerCase())) {
+            cards[i].classList.remove("is-hidden");
+        } else {
+            cards[i].classList.add("is-hidden");
+        }
+    }
+}
+
+//A little delay
+let typingTimer;               
+let typeInterval = 500;  
+let searchInput = document.getElementById('searchbox');
+
+searchInput.addEventListener('keyup', () => {
+    clearTimeout(typingTimer);
+    typingTimer = setTimeout(liveSearch, typeInterval);
+});
+</script>
+
+</div>
+
+<script src="https://cdnjs.cloudflare.com/ajax/libs/flowbite/2.3.0/flowbite.min.js"></script>
+</body>
+</html>
+`
+
+type GalleryModel struct {
+	Name        string   `json:"name" yaml:"name"`
+	URLs        []string `json:"urls" yaml:"urls"`
+	Icon        string   `json:"icon" yaml:"icon"`
+	Description string   `json:"description" yaml:"description"`
+}
+
+func main() {
+	// read the YAML file which contains the models
+
+	f, err := ioutil.ReadFile(os.Args[1])
+	if err != nil {
+		fmt.Println("Error reading file:", err)
+		return
+	}
+
+	models := []*GalleryModel{}
+	err = yaml.Unmarshal(f, &models)
+	if err != nil {
+		// write to stderr
+		os.Stderr.WriteString("Error unmarshaling YAML: " + err.Error() + "\n")
+		return
+	}
+
+	// render the template
+	data := struct {
+		Models          []*GalleryModel
+		AvailableModels int
+	}{
+		Models:          models,
+		AvailableModels: len(models),
+	}
+	tmpl := template.Must(template.New("modelPage").Parse(modelPageTemplate))
+
+	err = tmpl.Execute(os.Stdout, data)
+	if err != nil {
+		fmt.Println("Error executing template:", err)
+		return
+	}
+}

.github/dependabot.yml

@@ -0,0 +1,25 @@
+# https://docs.github.com/en/code-security/dependabot/dependabot-version-updates/configuration-options-for-the-dependabot.yml-file
+version: 2
+updates:
+  - package-ecosystem: "gomod"
+    directory: "/"
+    schedule:
+      interval: "weekly"
+  - package-ecosystem: "github-actions"
+    # Workflow files stored in the default location of `.github/workflows`. (You don't need to specify `/.github/workflows` for `directory`. You can use `directory: "/"`.)
+    directory: "/"
+    schedule:
+      # Check for updates to GitHub Actions every weekday
+      interval: "weekly"
+  - package-ecosystem: "pip"
+    # Workflow files stored in the default location of `.github/workflows`. (You don't need to specify `/.github/workflows` for `directory`. You can use `directory: "/"`.)
+    directory: "/"
+    schedule:
+      # Check for updates to GitHub Actions every weekday
+      interval: "weekly"
+  - package-ecosystem: "docker"
+    # Workflow files stored in the default location of `.github/workflows`. (You don't need to specify `/.github/workflows` for `directory`. You can use `directory: "/"`.)
+    directory: "/"
+    schedule:
+      # Check for updates to GitHub Actions every weekday
+      interval: "weekly"

.github/labeler.yml

@@ -0,0 +1,24 @@
+enhancements:
+ - head-branch: ['^feature', 'feature']
+
+kind/documentation:
+- any:
+  - changed-files:
+    - any-glob-to-any-file: 'docs/*'
+  - changed-files:
+    - any-glob-to-any-file: '*.md'
+
+area/ai-model:
+- any:
+  - changed-files:
+    - any-glob-to-any-file: 'gallery/*'
+
+examples:
+- any:
+  - changed-files:
+    - any-glob-to-any-file: 'examples/*'
+
+ci:
+- any:
+  - changed-files:
+    - any-glob-to-any-file: '.github/*'

.github/release.yml

@@ -12,13 +12,23 @@ changelog:
     - title: "Bug fixes :bug:"
       labels:
         - bug
+        - regression
     - title: Exciting New Features 🎉
       labels:
         - Semver-Minor
         - enhancement
+        - ux
+        - roadmap
+    - title: 🧠 Models
+      labels:
+        - area/ai-model
+    - title: 📖 Documentation and examples
+      labels:
+        - kind/documentation
+        - examples
     - title: 👒 Dependencies
       labels:
         - dependencies
     - title: Other Changes
       labels:
-        - "*"
+        - "*"

.github/workflows/bump_deps.yaml

@@ -49,7 +49,7 @@ jobs:
         run: |
           bash .github/bump_deps.sh ${{ matrix.repository }} ${{ matrix.branch }} ${{ matrix.variable }}
       - name: Create Pull Request
-        uses: peter-evans/create-pull-request@v5
+        uses: peter-evans/create-pull-request@v6
         with:
           token: ${{ secrets.UPDATE_BOT_TOKEN }}
           push-to-fork: ci-forks/LocalAI

.github/workflows/bump_docs.yaml

@@ -17,7 +17,7 @@ jobs:
         run: |
           bash .github/bump_docs.sh ${{ matrix.repository }}
       - name: Create Pull Request
-        uses: peter-evans/create-pull-request@v5
+        uses: peter-evans/create-pull-request@v6
         with:
           token: ${{ secrets.UPDATE_BOT_TOKEN }}
           push-to-fork: ci-forks/LocalAI

.github/workflows/checksum_checker.yaml

@@ -0,0 +1,47 @@
+name: Check if checksums are up-to-date
+on:
+  schedule:
+    - cron: 0 20 * * *
+  workflow_dispatch:
+jobs:
+  checksum_check:
+    runs-on: arc-runner-set
+    steps:
+      - name: Force Install GIT latest
+        run: |
+          sudo apt-get update \
+          && sudo apt-get install -y software-properties-common \
+          && sudo apt-get update \
+          && sudo add-apt-repository -y ppa:git-core/ppa \
+          && sudo apt-get update \
+          && sudo apt-get install -y git
+      - uses: actions/checkout@v4
+      - name: Install dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y pip wget
+          sudo pip install --upgrade pip 
+          pip install huggingface_hub
+      - name: 'Setup yq'
+        uses: dcarbone/install-yq-action@v1.1.1
+        with:
+          version: 'v4.43.1'
+          download-compressed: true
+          force: true
+
+      - name: Checksum checker 🔧
+        run: |
+          export HF_HOME=/hf_cache
+          sudo mkdir /hf_cache
+          sudo chmod 777 /hf_cache
+          bash .github/checksum_checker.sh gallery/index.yaml
+      - name: Create Pull Request
+        uses: peter-evans/create-pull-request@v6
+        with:
+          token: ${{ secrets.UPDATE_BOT_TOKEN }}
+          push-to-fork: ci-forks/LocalAI
+          commit-message: ':arrow_up: Checksum updates in gallery/index.yaml'
+          title: 'models(gallery): :arrow_up: update checksum'
+          branch: "update/checksum"
+          body: Updating checksums in gallery/index.yaml
+          signoff: true

.github/workflows/dependabot_auto.yml

@@ -0,0 +1,43 @@
+name: Dependabot auto-merge
+on:
+- pull_request_target
+
+permissions:
+  contents: write
+  pull-requests: write
+  packages: read
+
+jobs:
+  dependabot:
+    runs-on: ubuntu-latest
+    if: ${{ github.actor == 'dependabot[bot]' }}
+    steps:
+      - name: Dependabot metadata
+        id: metadata
+        uses: dependabot/fetch-metadata@v2.1.0
+        with:
+          github-token: "${{ secrets.GITHUB_TOKEN }}"
+          skip-commit-verification: true
+
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Approve a PR if not already approved
+        run: |
+          gh pr checkout "$PR_URL"
+            if [ "$(gh pr status --json reviewDecision -q .currentBranch.reviewDecision)" != "APPROVED" ];
+          then
+            gh pr review --approve "$PR_URL"
+          else
+            echo "PR already approved.";
+          fi
+        env:
+          PR_URL: ${{github.event.pull_request.html_url}}
+          GITHUB_TOKEN: ${{secrets.GITHUB_TOKEN}}
+
+      - name: Enable auto-merge for Dependabot PRs
+        if: ${{ contains(github.event.pull_request.title, 'bump')}}
+        run: gh pr merge --auto --squash "$PR_URL"
+        env:
+          PR_URL: ${{github.event.pull_request.html_url}}
+          GITHUB_TOKEN: ${{secrets.GITHUB_TOKEN}}

.github/workflows/generate_grpc_cache.yaml

@@ -0,0 +1,94 @@
+name: 'generate and publish GRPC docker caches'
+
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - master
+
+concurrency:
+  group: grpc-cache-${{ github.head_ref || github.ref }}-${{ github.repository }}
+  cancel-in-progress: true
+
+jobs:
+  generate_caches:
+    strategy:
+      matrix:
+        include:
+          - grpc-base-image: ubuntu:22.04
+            runs-on: 'ubuntu-latest'
+            platforms: 'linux/amd64,linux/arm64'
+    runs-on: ${{matrix.runs-on}}
+    steps:
+      - name: Release space from worker
+        if: matrix.runs-on == 'ubuntu-latest'
+        run: |
+          echo "Listing top largest packages"
+          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+          head -n 30 <<< "${pkgs}"
+          echo
+          df -h
+          echo
+          sudo apt-get remove -y '^llvm-.*|^libllvm.*' || true
+          sudo apt-get remove --auto-remove android-sdk-platform-tools || true
+          sudo apt-get purge --auto-remove android-sdk-platform-tools || true
+          sudo rm -rf /usr/local/lib/android
+          sudo apt-get remove -y '^dotnet-.*|^aspnetcore-.*' || true
+          sudo rm -rf /usr/share/dotnet
+          sudo apt-get remove -y '^mono-.*' || true
+          sudo apt-get remove -y '^ghc-.*' || true
+          sudo apt-get remove -y '.*jdk.*|.*jre.*' || true
+          sudo apt-get remove -y 'php.*' || true
+          sudo apt-get remove -y hhvm powershell firefox monodoc-manual msbuild || true
+          sudo apt-get remove -y '^google-.*' || true
+          sudo apt-get remove -y azure-cli || true
+          sudo apt-get remove -y '^mongo.*-.*|^postgresql-.*|^mysql-.*|^mssql-.*' || true
+          sudo apt-get remove -y '^gfortran-.*' || true
+          sudo apt-get remove -y microsoft-edge-stable || true
+          sudo apt-get remove -y firefox || true
+          sudo apt-get remove -y powershell || true
+          sudo apt-get remove -y r-base-core || true
+          sudo apt-get autoremove -y
+          sudo apt-get clean
+          echo
+          echo "Listing top largest packages"
+          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+          head -n 30 <<< "${pkgs}"
+          echo
+          sudo rm -rfv build || true
+          sudo rm -rf /usr/share/dotnet || true
+          sudo rm -rf /opt/ghc || true
+          sudo rm -rf "/usr/local/share/boost" || true
+          sudo rm -rf "$AGENT_TOOLSDIRECTORY" || true
+          df -h
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@master
+        with:
+          platforms: all
+
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@master
+
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Cache GRPC
+        uses: docker/build-push-action@v5
+        with:
+          builder: ${{ steps.buildx.outputs.name }}
+          # The build-args MUST be an EXACT match between the image cache and other workflow steps that want to use that cache.
+          # This means that even the MAKEFLAGS have to be an EXACT match.
+          # If the build-args are not an EXACT match, it will result in a cache miss, which will require GRPC to be built from scratch.
+          build-args: |
+            GRPC_BASE_IMAGE=${{ matrix.grpc-base-image }}
+            GRPC_MAKEFLAGS=--jobs=4 --output-sync=target
+            GRPC_VERSION=v1.64.0
+          context: .
+          file: ./Dockerfile
+          cache-to: type=gha,ignore-error=true
+          cache-from: type=gha
+          target: grpc
+          platforms: ${{ matrix.platforms }}
+          push: false

.github/workflows/generate_intel_image.yaml

@@ -0,0 +1,59 @@
+name: 'generate and publish intel docker caches'
+
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - master
+
+concurrency:
+  group: intel-cache-${{ github.head_ref || github.ref }}-${{ github.repository }}
+  cancel-in-progress: true
+
+jobs:
+  generate_caches:
+    strategy:
+      matrix:
+        include:
+          - base-image: intel/oneapi-basekit:2024.1.0-devel-ubuntu22.04
+            runs-on: 'ubuntu-latest'
+            platforms: 'linux/amd64'
+    runs-on: ${{matrix.runs-on}}
+    steps:
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@master
+        with:
+          platforms: all
+      - name: Login to DockerHub
+        if: github.event_name != 'pull_request'
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+
+      - name: Login to quay
+        if: github.event_name != 'pull_request'
+        uses: docker/login-action@v3
+        with:
+          registry: quay.io
+          username: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+          password: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+      - name: Set up Docker Buildx
+        id: buildx
+        uses: docker/setup-buildx-action@master
+
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Cache Intel images
+        uses: docker/build-push-action@v5
+        with:
+          builder: ${{ steps.buildx.outputs.name }}
+          build-args: |
+            BASE_IMAGE=${{ matrix.base-image }}
+          context: .
+          file: ./Dockerfile
+          tags: quay.io/go-skynet/intel-oneapi-base:latest
+          push: true
+          target: intel
+          platforms: ${{ matrix.platforms }}

.github/workflows/image-pr.yml

@@ -22,6 +22,8 @@ jobs:
       platforms: ${{ matrix.platforms }}
       runs-on: ${{ matrix.runs-on }}
       base-image: ${{ matrix.base-image }}
+      grpc-base-image: ${{ matrix.grpc-base-image }}
+      makeflags: ${{ matrix.makeflags }}
     secrets:
       dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
       dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
@@ -41,6 +43,7 @@ jobs:
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
             cuda-minor-version: "1"
@@ -51,22 +54,27 @@ jobs:
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'hipblas'
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-hipblas'
             ffmpeg: 'false'
             image-type: 'extras'
-            base-image: "rocm/dev-ubuntu-22.04:6.0-complete"
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'sycl_f16'
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            base-image: "intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04"
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
             tag-suffix: 'sycl-f16-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
   core-image-build:
     uses: ./.github/workflows/image_build.yml
     with:
@@ -80,6 +88,8 @@ jobs:
       platforms: ${{ matrix.platforms }}
       runs-on: ${{ matrix.runs-on }}
       base-image: ${{ matrix.base-image }}
+      grpc-base-image: ${{ matrix.grpc-base-image }}
+      makeflags: ${{ matrix.makeflags }}
     secrets:
       dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
       dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
@@ -96,14 +106,17 @@ jobs:
             image-type: 'core'
             runs-on: 'ubuntu-latest'
             base-image: "ubuntu:22.04"
+            makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'sycl_f16'
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            base-image: "intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04"
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
             tag-suffix: 'sycl-f16-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
             cuda-minor-version: "1"
@@ -113,4 +126,5 @@ jobs:
             ffmpeg: 'true'
             image-type: 'core'
             runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
+            base-image: "ubuntu:22.04"
+            makeflags: "--jobs=4 --output-sync=target"

.github/workflows/image.yml

@@ -26,6 +26,11 @@ jobs:
       platforms: ${{ matrix.platforms }}
       runs-on: ${{ matrix.runs-on }}
       base-image: ${{ matrix.base-image }}
+      grpc-base-image: ${{ matrix.grpc-base-image }}
+      aio: ${{ matrix.aio }}
+      makeflags: ${{ matrix.makeflags }}
+      latest-image: ${{ matrix.latest-image }}
+      latest-image-aio: ${{ matrix.latest-image-aio }}
     secrets:
       dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
       dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
@@ -47,14 +52,16 @@ jobs:
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: ''
             platforms: 'linux/amd64'
-            tag-latest: 'false'
+            tag-latest: 'auto'
             tag-suffix: '-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "11"
             cuda-minor-version: "7"
@@ -65,6 +72,7 @@ jobs:
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
             cuda-minor-version: "1"
@@ -75,26 +83,35 @@ jobs:
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "11"
             cuda-minor-version: "7"
             platforms: 'linux/amd64'
-            tag-latest: 'false'
+            tag-latest: 'auto'
             tag-suffix: '-cublas-cuda11-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
+            aio: "-aio-gpu-nvidia-cuda-11"
+            latest-image: 'latest-gpu-nvidia-cuda-11'
+            latest-image-aio: 'latest-aio-gpu-nvidia-cuda-11'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
             cuda-minor-version: "1"
             platforms: 'linux/amd64'
-            tag-latest: 'false'
+            tag-latest: 'auto'
             tag-suffix: '-cublas-cuda12-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             runs-on: 'arc-runner-set'
             base-image: "ubuntu:22.04"
+            aio: "-aio-gpu-nvidia-cuda-12"
+            latest-image: 'latest-gpu-nvidia-cuda-12'
+            latest-image-aio: 'latest-aio-gpu-nvidia-cuda-12'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: ''
             #platforms: 'linux/amd64,linux/arm64'
             platforms: 'linux/amd64'
@@ -104,87 +121,117 @@ jobs:
             image-type: 'extras'
             base-image: "ubuntu:22.04"
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'hipblas'
             platforms: 'linux/amd64'
-            tag-latest: 'false'
+            tag-latest: 'auto'
             tag-suffix: '-hipblas-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
-            base-image: "rocm/dev-ubuntu-22.04:6.0-complete"
+            aio: "-aio-gpu-hipblas"
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
+            latest-image: 'latest-gpu-hipblas'
+            latest-image-aio: 'latest-aio-gpu-hipblas'
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'hipblas'
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-hipblas'
             ffmpeg: 'false'
             image-type: 'extras'
-            base-image: "rocm/dev-ubuntu-22.04:6.0-complete"
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'sycl_f16'
             platforms: 'linux/amd64'
-            tag-latest: 'false'
-            base-image: "intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04"
+            tag-latest: 'auto'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
             tag-suffix: '-sycl-f16-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             runs-on: 'arc-runner-set'
+            aio: "-aio-gpu-intel-f16"
+            latest-image: 'latest-gpu-intel-f16'
+            latest-image-aio: 'latest-aio-gpu-intel-f16'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'sycl_f32'
             platforms: 'linux/amd64'
-            tag-latest: 'false'
-            base-image: "intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04"
+            tag-latest: 'auto'
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
             tag-suffix: '-sycl-f32-ffmpeg'
             ffmpeg: 'true'
             image-type: 'extras'
             runs-on: 'arc-runner-set'
+            aio: "-aio-gpu-intel-f32"
+            latest-image: 'latest-gpu-intel-f32'
+            latest-image-aio: 'latest-aio-gpu-intel-f32'
+            makeflags: "--jobs=3 --output-sync=target"
           # Core images
           - build-type: 'sycl_f16'
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            base-image: "intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04"
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
             tag-suffix: '-sycl-f16-core'
             ffmpeg: 'false'
             image-type: 'core'
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'sycl_f32'
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            base-image: "intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04"
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
             tag-suffix: '-sycl-f32-core'
             ffmpeg: 'false'
             image-type: 'core'
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'sycl_f16'
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            base-image: "intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04"
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
             tag-suffix: '-sycl-f16-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'sycl_f32'
             platforms: 'linux/amd64'
             tag-latest: 'false'
-            base-image: "intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04"
+            base-image: "quay.io/go-skynet/intel-oneapi-base:latest"
+            grpc-base-image: "ubuntu:22.04"
             tag-suffix: '-sycl-f32-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'hipblas'
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-hipblas-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
-            base-image: "rocm/dev-ubuntu-22.04:6.0-complete"
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
           - build-type: 'hipblas'
             platforms: 'linux/amd64'
             tag-latest: 'false'
             tag-suffix: '-hipblas-core'
             ffmpeg: 'false'
             image-type: 'core'
-            base-image: "rocm/dev-ubuntu-22.04:6.0-complete"
+            base-image: "rocm/dev-ubuntu-22.04:6.1"
+            grpc-base-image: "ubuntu:22.04"
             runs-on: 'arc-runner-set'
+            makeflags: "--jobs=3 --output-sync=target"
   
   core-image-build:
     uses: ./.github/workflows/image_build.yml
@@ -198,7 +245,12 @@ jobs:
       cuda-minor-version: ${{ matrix.cuda-minor-version }}
       platforms: ${{ matrix.platforms }}
       runs-on: ${{ matrix.runs-on }}
+      aio: ${{ matrix.aio }}
       base-image: ${{ matrix.base-image }}
+      grpc-base-image: ${{ matrix.grpc-base-image }}
+      makeflags: ${{ matrix.makeflags }}
+      latest-image: ${{ matrix.latest-image }}
+      latest-image-aio: ${{ matrix.latest-image-aio }}
     secrets:
       dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
       dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
@@ -208,13 +260,17 @@ jobs:
       matrix:
         include:
           - build-type: ''
-            platforms: 'linux/amd64'
-            tag-latest: 'false'
+            platforms: 'linux/amd64,linux/arm64'
+            tag-latest: 'auto'
             tag-suffix: '-ffmpeg-core'
             ffmpeg: 'true'
             image-type: 'core'
             base-image: "ubuntu:22.04"
             runs-on: 'ubuntu-latest'
+            aio: "-aio-cpu"
+            latest-image: 'latest-cpu'
+            latest-image-aio: 'latest-aio-cpu'
+            makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "11"
             cuda-minor-version: "7"
@@ -225,6 +281,7 @@ jobs:
             image-type: 'core'
             base-image: "ubuntu:22.04"
             runs-on: 'ubuntu-latest'
+            makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
             cuda-minor-version: "1"
@@ -235,6 +292,7 @@ jobs:
             image-type: 'core'
             base-image: "ubuntu:22.04"
             runs-on: 'ubuntu-latest'
+            makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "11"
             cuda-minor-version: "7"
@@ -245,6 +303,7 @@ jobs:
             image-type: 'core'
             runs-on: 'ubuntu-latest'
             base-image: "ubuntu:22.04"
+            makeflags: "--jobs=4 --output-sync=target"
           - build-type: 'cublas'
             cuda-major-version: "12"
             cuda-minor-version: "1"
@@ -255,3 +314,4 @@ jobs:
             image-type: 'core'
             runs-on: 'ubuntu-latest'
             base-image: "ubuntu:22.04"
+            makeflags: "--jobs=4 --output-sync=target"

.github/workflows/image_build.yml

@@ -6,6 +6,10 @@ on:
     inputs:
       base-image:
         description: 'Base image'
+        required: true
+        type: string
+      grpc-base-image:
+        description: 'GRPC Base image, must be a compatible image with base-image'
         required: false
         default: ''
         type: string
@@ -29,6 +33,14 @@ on:
         description: 'Tag latest'
         default: ''
         type: string
+      latest-image:
+          description: 'Tag latest'
+          default: ''
+          type: string
+      latest-image-aio:
+          description: 'Tag latest'
+          default: ''
+          type: string
       tag-suffix:
         description: 'Tag suffix'
         default: ''
@@ -46,6 +58,16 @@ on:
         required: true
         default: ''
         type: string
+      makeflags:
+        description: 'Make Flags'
+        required: false
+        default: '--jobs=4 --output-sync=target'
+        type: string
+      aio:
+        description: 'AIO Image Name'
+        required: false
+        default: ''
+        type: string
     secrets:
       dockerUsername:
         required: true
@@ -69,6 +91,7 @@ jobs:
           && sudo apt-get install -y git
       - name: Checkout
         uses: actions/checkout@v4
+
       - name: Release space from worker
         if: inputs.runs-on == 'ubuntu-latest'
         run: |
@@ -110,8 +133,10 @@ jobs:
           sudo rm -rf "/usr/local/share/boost" || true
           sudo rm -rf "$AGENT_TOOLSDIRECTORY" || true
           df -h
+
       - name: Docker meta
         id: meta
+        if: github.event_name != 'pull_request'
         uses: docker/metadata-action@v5
         with:
           images: |
@@ -124,6 +149,46 @@ jobs:
           flavor: |
             latest=${{ inputs.tag-latest }}
             suffix=${{ inputs.tag-suffix }}
+      - name: Docker meta for PR
+        id: meta_pull_request
+        if: github.event_name == 'pull_request'
+        uses: docker/metadata-action@v5
+        with:
+          images: |
+            ttl.sh/localai-ci-pr-${{ github.event.number }}
+          tags: |
+            type=ref,event=branch
+            type=semver,pattern={{raw}}
+            type=sha
+          flavor: |
+            latest=${{ inputs.tag-latest }}
+            suffix=${{ inputs.tag-suffix }}
+      - name: Docker meta AIO (quay.io)
+        if: inputs.aio != ''
+        id: meta_aio
+        uses: docker/metadata-action@v5
+        with:
+          images: |
+            quay.io/go-skynet/local-ai
+          tags: |
+            type=ref,event=branch
+            type=semver,pattern={{raw}}
+          flavor: |
+            latest=${{ inputs.tag-latest }}
+            suffix=${{ inputs.aio }}
+
+      - name: Docker meta AIO (dockerhub)
+        if: inputs.aio != ''
+        id: meta_aio_dockerhub
+        uses: docker/metadata-action@v5
+        with:
+          images: |
+            localai/localai
+          tags: |
+            type=ref,event=branch
+            type=semver,pattern={{raw}}
+          flavor: |
+            suffix=${{ inputs.aio }}
 
       - name: Set up QEMU
         uses: docker/setup-qemu-action@master
@@ -151,8 +216,13 @@ jobs:
 
       - name: Build and push
         uses: docker/build-push-action@v5
+        if: github.event_name != 'pull_request'
         with:
           builder: ${{ steps.buildx.outputs.name }}
+          # The build-args MUST be an EXACT match between the image cache and other workflow steps that want to use that cache.
+          # This means that even the MAKEFLAGS have to be an EXACT match.
+          # If the build-args are not an EXACT match, it will result in a cache miss, which will require GRPC to be built from scratch.
+          # This is why some build args like GRPC_VERSION and MAKEFLAGS are hardcoded
           build-args: |
             BUILD_TYPE=${{ inputs.build-type }}
             CUDA_MAJOR_VERSION=${{ inputs.cuda-major-version }}
@@ -160,12 +230,106 @@ jobs:
             FFMPEG=${{ inputs.ffmpeg }}
             IMAGE_TYPE=${{ inputs.image-type }}
             BASE_IMAGE=${{ inputs.base-image }}
+            GRPC_BASE_IMAGE=${{ inputs.grpc-base-image || inputs.base-image }}
+            GRPC_MAKEFLAGS=--jobs=4 --output-sync=target
+            GRPC_VERSION=v1.64.0
+            MAKEFLAGS=${{ inputs.makeflags }}
           context: .
           file: ./Dockerfile
+          cache-from: type=gha
           platforms: ${{ inputs.platforms }}
           push: ${{ github.event_name != 'pull_request' }}
           tags: ${{ steps.meta.outputs.tags }}
           labels: ${{ steps.meta.outputs.labels }}
+### Start testing image
+      - name: Build and push
+        uses: docker/build-push-action@v5
+        if: github.event_name == 'pull_request'
+        with:
+          builder: ${{ steps.buildx.outputs.name }}
+          # The build-args MUST be an EXACT match between the image cache and other workflow steps that want to use that cache.
+          # This means that even the MAKEFLAGS have to be an EXACT match.
+          # If the build-args are not an EXACT match, it will result in a cache miss, which will require GRPC to be built from scratch.
+          # This is why some build args like GRPC_VERSION and MAKEFLAGS are hardcoded
+          build-args: |
+            BUILD_TYPE=${{ inputs.build-type }}
+            CUDA_MAJOR_VERSION=${{ inputs.cuda-major-version }}
+            CUDA_MINOR_VERSION=${{ inputs.cuda-minor-version }}
+            FFMPEG=${{ inputs.ffmpeg }}
+            IMAGE_TYPE=${{ inputs.image-type }}
+            BASE_IMAGE=${{ inputs.base-image }}
+            GRPC_BASE_IMAGE=${{ inputs.grpc-base-image || inputs.base-image }}
+            GRPC_MAKEFLAGS=--jobs=4 --output-sync=target
+            GRPC_VERSION=v1.64.0
+            MAKEFLAGS=${{ inputs.makeflags }}
+          context: .
+          file: ./Dockerfile
+          cache-from: type=gha
+          platforms: ${{ inputs.platforms }}
+          push: true
+          tags: ${{ steps.meta_pull_request.outputs.tags }}
+          labels: ${{ steps.meta_pull_request.outputs.labels }}
+      - name: Testing image
+        if: github.event_name == 'pull_request'
+        run: |
+          echo "Image is available at ttl.sh/localai-ci-pr-${{ github.event.number }}:${{ steps.meta_pull_request.outputs.version }}" >> $GITHUB_STEP_SUMMARY
+## End testing image
+      - name: Build and push AIO image
+        if: inputs.aio != ''
+        uses: docker/build-push-action@v5
+        with:
+          builder: ${{ steps.buildx.outputs.name }}
+          build-args: |
+            BASE_IMAGE=quay.io/go-skynet/local-ai:${{ steps.meta.outputs.version }}
+            MAKEFLAGS=${{ inputs.makeflags }}
+          context: .
+          file: ./Dockerfile.aio
+          platforms: ${{ inputs.platforms }}
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.meta_aio.outputs.tags }}
+          labels: ${{ steps.meta_aio.outputs.labels }}
+
+      - name: Build and push AIO image (dockerhub)
+        if: inputs.aio != ''
+        uses: docker/build-push-action@v5
+        with:
+          builder: ${{ steps.buildx.outputs.name }}
+          build-args: |
+            BASE_IMAGE=localai/localai:${{ steps.meta.outputs.version }}
+            MAKEFLAGS=${{ inputs.makeflags }}
+          context: .
+          file: ./Dockerfile.aio
+          platforms: ${{ inputs.platforms }}
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.meta_aio_dockerhub.outputs.tags }}
+          labels: ${{ steps.meta_aio_dockerhub.outputs.labels }}
+
+      - name: Latest tag
+        # run this on branches, when it is a tag and there is a latest-image defined
+        if: github.event_name != 'pull_request' && inputs.latest-image != ''  && github.ref_type == 'tag'
+        run: |
+          docker pull localai/localai:${{ steps.meta.outputs.version }}
+          docker tag localai/localai:${{ steps.meta.outputs.version }} localai/localai:${{ inputs.latest-image }}
+          docker push localai/localai:${{ inputs.latest-image }}
+          docker pull quay.io/go-skynet/local-ai:${{ steps.meta.outputs.version }}
+          docker tag quay.io/go-skynet/local-ai:${{ steps.meta.outputs.version }} quay.io/go-skynet/local-ai:${{ inputs.latest-image }}
+          docker push quay.io/go-skynet/local-ai:${{ inputs.latest-image }}
+      - name: Latest AIO tag
+        # run this on branches, when it is a tag and there is a latest-image defined
+        if: github.event_name != 'pull_request' && inputs.latest-image-aio != ''  && github.ref_type == 'tag'
+        run: |
+          docker pull localai/localai:${{ steps.meta_aio_dockerhub.outputs.version }}
+          docker tag localai/localai:${{ steps.meta_aio_dockerhub.outputs.version }} localai/localai:${{ inputs.latest-image-aio }}
+          docker push localai/localai:${{ inputs.latest-image-aio }}
+          docker pull quay.io/go-skynet/local-ai:${{ steps.meta_aio.outputs.version }}
+          docker tag quay.io/go-skynet/local-ai:${{ steps.meta_aio.outputs.version }} quay.io/go-skynet/local-ai:${{ inputs.latest-image-aio }}
+          docker push quay.io/go-skynet/local-ai:${{ inputs.latest-image-aio }}
+  
       - name: job summary
         run: |
           echo "Built image: ${{ steps.meta.outputs.labels }}" >> $GITHUB_STEP_SUMMARY
+
+      - name: job summary(AIO)
+        if: inputs.aio != ''
+        run: |
+          echo "Built image: ${{ steps.meta_aio.outputs.labels }}" >> $GITHUB_STEP_SUMMARY

.github/workflows/labeler.yml

@@ -0,0 +1,12 @@
+name: "Pull Request Labeler"
+on:
+- pull_request_target
+
+jobs:
+  labeler:
+    permissions:
+      contents: read
+      pull-requests: write
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/labeler@v5

.github/workflows/localaibot_automerge.yml

@@ -0,0 +1,35 @@
+name: LocalAI-bot auto-merge
+on:
+- pull_request_target
+
+permissions:
+  contents: write
+  pull-requests: write
+  packages: read
+
+jobs:
+  dependabot:
+    runs-on: ubuntu-latest
+    if: ${{ github.actor == 'localai-bot' }}
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Approve a PR if not already approved
+        run: |
+          gh pr checkout "$PR_URL"
+            if [ "$(gh pr status --json reviewDecision -q .currentBranch.reviewDecision)" != "APPROVED" ];
+          then
+            gh pr review --approve "$PR_URL"
+          else
+            echo "PR already approved.";
+          fi
+        env:
+          PR_URL: ${{github.event.pull_request.html_url}}
+          GITHUB_TOKEN: ${{secrets.GITHUB_TOKEN}}
+
+      - name: Enable auto-merge for LocalAIBot PRs
+        run: gh pr merge --auto --squash "$PR_URL"
+        env:
+          PR_URL: ${{github.event.pull_request.html_url}}
+          GITHUB_TOKEN: ${{secrets.GITHUB_TOKEN}}

.github/workflows/release.yaml

@@ -1,6 +1,11 @@
 name: Build and Release
 
-on: push
+on:
+- push
+- pull_request
+
+env:
+  GRPC_VERSION: v1.64.0
 
 permissions:
   contents: write
@@ -10,85 +15,225 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  build-linux:
-    strategy:
-      matrix:
-        include:
-          - build: 'avx2'
-            defines: ''
-          - build: 'avx'
-            defines: '-DLLAMA_AVX2=OFF'
-          - build: 'avx512'
-            defines: '-DLLAMA_AVX512=ON'
-          - build: 'cuda12'
-            defines: ''
-          - build: 'cuda11'
-            defines: ''
+
+  build-linux-arm:
     runs-on: ubuntu-latest
     steps:
       - name: Clone
         uses: actions/checkout@v4
         with:
           submodules: true
-      - uses: actions/setup-go@v4
+      - uses: actions/setup-go@v5
         with:
-          go-version: '>=1.21.0'
+          go-version: '1.21.x'
+          cache: false
+
       - name: Dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential ffmpeg
+          sudo apt-get install build-essential ffmpeg protobuf-compiler ccache
+          sudo apt-get install -qy binutils-aarch64-linux-gnu gcc-aarch64-linux-gnu g++-aarch64-linux-gnu
       - name: Install CUDA Dependencies
-        if: ${{ matrix.build == 'cuda12' || matrix.build == 'cuda11' }}
         run: |
-          if [ "${{ matrix.build }}" == "cuda12" ]; then
-            export CUDA_VERSION=12-3
-          else
-            export CUDA_VERSION=11-7
-          fi
-          curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
+          curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/cross-linux-aarch64/cuda-keyring_1.1-1_all.deb
           sudo dpkg -i cuda-keyring_1.1-1_all.deb
           sudo apt-get update
-          sudo apt-get install -y cuda-nvcc-${CUDA_VERSION} libcublas-dev-${CUDA_VERSION}
+          sudo apt-get install -y cuda-cross-aarch64 cuda-nvcc-cross-aarch64-${CUDA_VERSION} libcublas-cross-aarch64-${CUDA_VERSION}
+        env:
+          CUDA_VERSION: 12-4
       - name: Cache grpc
         id: cache-grpc
-        uses: actions/cache@v3
+        uses: actions/cache@v4
         with:
           path: grpc
-          key: ${{ runner.os }}-grpc
+          key: ${{ runner.os }}-arm-grpc-${{ env.GRPC_VERSION }}
       - name: Build grpc
         if: steps.cache-grpc.outputs.cache-hit != 'true'
         run: |
-          git clone --recurse-submodules -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+
+          git clone --recurse-submodules -b ${{ env.GRPC_VERSION }} --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
           cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
             -DgRPC_BUILD_TESTS=OFF \
-            ../.. && sudo make -j12
+            ../.. && sudo make --jobs 5 --output-sync=target
       - name: Install gRPC
         run: |
-          cd grpc && cd cmake/build && sudo make -j12 install
+          GNU_HOST=aarch64-linux-gnu
+          C_COMPILER_ARM_LINUX=$GNU_HOST-gcc
+          CXX_COMPILER_ARM_LINUX=$GNU_HOST-g++
+
+          CROSS_TOOLCHAIN=/usr/$GNU_HOST
+          CROSS_STAGING_PREFIX=$CROSS_TOOLCHAIN/stage
+          CMAKE_CROSS_TOOLCHAIN=/tmp/arm.toolchain.cmake
+
+          # https://cmake.org/cmake/help/v3.13/manual/cmake-toolchains.7.html#cross-compiling-for-linux
+          echo "set(CMAKE_SYSTEM_NAME Linux)" >> $CMAKE_CROSS_TOOLCHAIN && \
+            echo "set(CMAKE_SYSTEM_PROCESSOR arm)" >> $CMAKE_CROSS_TOOLCHAIN && \
+            echo "set(CMAKE_STAGING_PREFIX $CROSS_STAGING_PREFIX)" >> $CMAKE_CROSS_TOOLCHAIN && \
+            echo "set(CMAKE_SYSROOT ${CROSS_TOOLCHAIN}/sysroot)" >> $CMAKE_CROSS_TOOLCHAIN && \
+            echo "set(CMAKE_C_COMPILER /usr/bin/$C_COMPILER_ARM_LINUX)" >> $CMAKE_CROSS_TOOLCHAIN && \
+            echo "set(CMAKE_CXX_COMPILER /usr/bin/$CXX_COMPILER_ARM_LINUX)" >> $CMAKE_CROSS_TOOLCHAIN && \
+            echo "set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)" >> $CMAKE_CROSS_TOOLCHAIN && \
+            echo "set(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)" >> $CMAKE_CROSS_TOOLCHAIN && \
+            echo "set(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)" >> $CMAKE_CROSS_TOOLCHAIN && \
+            echo "set(CMAKE_FIND_ROOT_PATH_MODE_PACKAGE ONLY)" >> $CMAKE_CROSS_TOOLCHAIN
+          GRPC_DIR=$PWD/grpc
+          cd grpc && cd cmake/build && sudo make --jobs 5 --output-sync=target install && \
+          GRPC_CROSS_BUILD_DIR=$GRPC_DIR/cmake/cross_build && \
+          mkdir -p $GRPC_CROSS_BUILD_DIR && \
+          cd $GRPC_CROSS_BUILD_DIR && \
+          cmake -DCMAKE_TOOLCHAIN_FILE=$CMAKE_CROSS_TOOLCHAIN \
+            -DCMAKE_BUILD_TYPE=Release \
+            -DCMAKE_INSTALL_PREFIX=$CROSS_TOOLCHAIN/grpc_install \
+            ../.. && \
+          sudo make -j`nproc` install
       - name: Build
         id: build
-        env:
-          CMAKE_ARGS: "${{ matrix.defines }}"
-          BUILD_ID: "${{ matrix.build }}"
         run: |
-          if [ "${{ matrix.build }}" == "cuda12" ] || [ "${{ matrix.build }}" == "cuda11" ]; then
-            export BUILD_TYPE=cublas
-            export PATH=/usr/local/cuda/bin:$PATH
-            make dist
-          else
-            STATIC=true make dist
-          fi
-      - uses: actions/upload-artifact@v3
+          GNU_HOST=aarch64-linux-gnu
+          C_COMPILER_ARM_LINUX=$GNU_HOST-gcc
+          CXX_COMPILER_ARM_LINUX=$GNU_HOST-g++
+
+          CROSS_TOOLCHAIN=/usr/$GNU_HOST
+          CROSS_STAGING_PREFIX=$CROSS_TOOLCHAIN/stage
+          CMAKE_CROSS_TOOLCHAIN=/tmp/arm.toolchain.cmake
+          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@8ba23be9613c672d40ae261d2a1335d639bdd59b
+          go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.0
+          export PATH=$PATH:$GOPATH/bin
+          export PATH=/usr/local/cuda/bin:$PATH
+          sudo rm -rf /usr/aarch64-linux-gnu/lib/libstdc++.so.6
+          sudo cp -rf /usr/aarch64-linux-gnu/lib/libstdc++.so* /usr/aarch64-linux-gnu/lib/libstdc++.so.6
+          sudo cp /usr/aarch64-linux-gnu/lib/ld-linux-aarch64.so.1 ld.so
+          GO_TAGS=p2p \
+          BACKEND_LIBS="./grpc/cmake/cross_build/third_party/re2/libre2.a ./grpc/cmake/cross_build/libgrpc.a ./grpc/cmake/cross_build/libgrpc++.a ./grpc/cmake/cross_build/third_party/protobuf/libprotobuf.a /usr/aarch64-linux-gnu/lib/libc.so.6 /usr/aarch64-linux-gnu/lib/libstdc++.so.6 /usr/aarch64-linux-gnu/lib/libgomp.so.1 /usr/aarch64-linux-gnu/lib/libm.so.6 /usr/aarch64-linux-gnu/lib/libgcc_s.so.1 /usr/aarch64-linux-gnu/lib/libdl.so.2 /usr/aarch64-linux-gnu/lib/libpthread.so.0 ./ld.so" \
+          GOOS=linux \
+          GOARCH=arm64 \
+          CMAKE_ARGS="-DProtobuf_INCLUDE_DIRS=$CROSS_STAGING_PREFIX/include -DProtobuf_DIR=$CROSS_STAGING_PREFIX/lib/cmake/protobuf -DgRPC_DIR=$CROSS_STAGING_PREFIX/lib/cmake/grpc -DCMAKE_TOOLCHAIN_FILE=$CMAKE_CROSS_TOOLCHAIN -DCMAKE_C_COMPILER=aarch64-linux-gnu-gcc -DCMAKE_CXX_COMPILER=aarch64-linux-gnu-g++" make dist-cross-linux-arm64
+      - uses: actions/upload-artifact@v4
         with:
-          name: ${{ matrix.build }}
+          name: LocalAI-linux-arm64
           path: release/
       - name: Release
-        uses: softprops/action-gh-release@v1
+        uses: softprops/action-gh-release@v2
         if: startsWith(github.ref, 'refs/tags/')
         with:
           files: |
             release/*
+      - name: Setup tmate session if tests fail
+        if: ${{ failure() }}
+        uses: mxschmitt/action-tmate@v3.18
+        with:
+          detached: true
+          connect-timeout-seconds: 180
+          limit-access-to-actor: true
+  build-linux:
+    runs-on: arc-runner-set
+    steps:
+      - name: Force Install GIT latest
+        run: |
+          sudo apt-get update \
+          && sudo apt-get install -y software-properties-common \
+          && sudo apt-get update \
+          && sudo add-apt-repository -y ppa:git-core/ppa \
+          && sudo apt-get update \
+          && sudo apt-get install -y git
+      - name: Clone
+        uses: actions/checkout@v4
+        with:
+          submodules: true
+      - uses: actions/setup-go@v5
+        with:
+          go-version: '1.21.x'
+          cache: false
+      - name: Dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y wget curl build-essential ffmpeg protobuf-compiler ccache cmake
+      - name: Intel Dependencies
+        run: |
+          wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB | gpg --dearmor | sudo tee /usr/share/keyrings/oneapi-archive-keyring.gpg > /dev/null
+          echo "deb [signed-by=/usr/share/keyrings/oneapi-archive-keyring.gpg] https://apt.repos.intel.com/oneapi all main" | sudo tee /etc/apt/sources.list.d/oneAPI.list
+          sudo apt update
+          sudo apt install -y intel-basekit
+      - name: Install CUDA Dependencies
+        run: |
+          curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
+          sudo dpkg -i cuda-keyring_1.1-1_all.deb
+          sudo apt-get update
+          sudo apt-get install -y cuda-nvcc-${CUDA_VERSION} libcublas-dev-${CUDA_VERSION}
+        env:
+          CUDA_VERSION: 12-3
+      - name: "Install Hipblas"
+        env:
+          ROCM_VERSION: "6.1"
+          AMDGPU_VERSION: "6.1"
+        run: |
+            set -ex 
 
+            sudo apt-get update
+            sudo DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends ca-certificates curl libnuma-dev gnupg 
+            
+            curl -sL https://repo.radeon.com/rocm/rocm.gpg.key | sudo apt-key add - 
+              
+            printf "deb [arch=amd64] https://repo.radeon.com/rocm/apt/$ROCM_VERSION/ jammy main" | sudo tee /etc/apt/sources.list.d/rocm.list
+            
+            printf "deb [arch=amd64] https://repo.radeon.com/amdgpu/$AMDGPU_VERSION/ubuntu jammy main" | sudo tee /etc/apt/sources.list.d/amdgpu.list
+            printf 'Package: *\nPin: release o=repo.radeon.com\nPin-Priority: 600' | sudo tee /etc/apt/preferences.d/rocm-pin-600
+            sudo apt-get update
+
+            sudo DEBIAN_FRONTEND=noninteractive apt-get install -y \
+                hipblas-dev rocm-dev \
+                rocblas-dev
+          
+            sudo apt-get clean
+            sudo rm -rf /var/lib/apt/lists/*
+            sudo ldconfig 
+      - name: Cache grpc
+        id: cache-grpc
+        uses: actions/cache@v4
+        with:
+          path: grpc
+          key: ${{ runner.os }}-grpc-${{ env.GRPC_VERSION }}
+      - name: Build grpc
+        if: steps.cache-grpc.outputs.cache-hit != 'true'
+        run: |
+          git clone --recurse-submodules -b ${{ env.GRPC_VERSION }} --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+          cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
+            -DgRPC_BUILD_TESTS=OFF \
+            ../.. && sudo make --jobs 5 --output-sync=target
+      - name: Install gRPC
+        run: |
+          cd grpc && cd cmake/build && sudo make --jobs 5 --output-sync=target install
+      # BACKEND_LIBS needed for gpu-workload: /opt/intel/oneapi/*/lib/libiomp5.so /opt/intel/oneapi/*/lib/libmkl_core.so /opt/intel/oneapi/*/lib/libmkl_core.so.2 /opt/intel/oneapi/*/lib/libmkl_intel_ilp64.so /opt/intel/oneapi/*/lib/libmkl_intel_ilp64.so.2 /opt/intel/oneapi/*/lib/libmkl_sycl_blas.so /opt/intel/oneapi/*/lib/libmkl_sycl_blas.so.4 /opt/intel/oneapi/*/lib/libmkl_tbb_thread.so /opt/intel/oneapi/*/lib/libmkl_tbb_thread.so.2 /opt/intel/oneapi/*/lib/libsycl.so /opt/intel/oneapi/*/lib/libsycl.so.7 /opt/intel/oneapi/*/lib/libsycl.so.7.1.0 /opt/rocm-*/lib/libamdhip64.so /opt/rocm-*/lib/libamdhip64.so.5 /opt/rocm-*/lib/libamdhip64.so.6 /opt/rocm-*/lib/libamdhip64.so.6.1.60100 /opt/rocm-*/lib/libhipblas.so /opt/rocm-*/lib/libhipblas.so.2 /opt/rocm-*/lib/libhipblas.so.2.1.60100 /opt/rocm-*/lib/librocblas.so /opt/rocm-*/lib/librocblas.so.4 /opt/rocm-*/lib/librocblas.so.4.1.60100 /usr/lib/x86_64-linux-gnu/libstdc++.so.6 /usr/lib/x86_64-linux-gnu/libOpenCL.so.1 /usr/lib/x86_64-linux-gnu/libOpenCL.so.1.0.0 /usr/lib/x86_64-linux-gnu/libm.so.6 /usr/lib/x86_64-linux-gnu/libgcc_s.so.1 /usr/lib/x86_64-linux-gnu/libc.so.6 /usr/lib/x86_64-linux-gnu/librt.so.1 /usr/local/cuda-*/targets/x86_64-linux/lib/libcublas.so /usr/local/cuda-*/targets/x86_64-linux/lib/libcublasLt.so /usr/local/cuda-*/targets/x86_64-linux/lib/libcudart.so /usr/local/cuda-*/targets/x86_64-linux/lib/stubs/libcuda.so
+      - name: Build
+        id: build
+        run: |
+          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@8ba23be9613c672d40ae261d2a1335d639bdd59b
+          go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.0
+          export PATH=$PATH:$GOPATH/bin
+          export PATH=/usr/local/cuda/bin:$PATH
+          export PATH=/opt/rocm/bin:$PATH
+          source /opt/intel/oneapi/setvars.sh
+          sudo cp /lib64/ld-linux-x86-64.so.2 ld.so
+          GO_TAGS=p2p \
+          BACKEND_LIBS="./ld.so /usr/lib/x86_64-linux-gnu/libstdc++.so.6 /usr/lib/x86_64-linux-gnu/libm.so.6 /usr/lib/x86_64-linux-gnu/libgcc_s.so.1 /usr/lib/x86_64-linux-gnu/libc.so.6 /usr/lib/x86_64-linux-gnu/libgomp.so.1" \
+          make -j4 dist
+      - uses: actions/upload-artifact@v4
+        with:
+          name: LocalAI-linux
+          path: release/
+      - name: Release
+        uses: softprops/action-gh-release@v2
+        if: startsWith(github.ref, 'refs/tags/')
+        with:
+          files: |
+            release/*
+      - name: Setup tmate session if tests fail
+        if: ${{ failure() }}
+        uses: mxschmitt/action-tmate@v3.18
+        with:
+          detached: true
+          connect-timeout-seconds: 180
+          limit-access-to-actor: true
   build-stablediffusion:
     runs-on: ubuntu-latest
     steps:
@@ -96,66 +241,72 @@ jobs:
         uses: actions/checkout@v4
         with:
           submodules: true
-      - uses: actions/setup-go@v4
+      - uses: actions/setup-go@v5
         with:
-          go-version: '>=1.21.0'
+          go-version: '1.21.x'
+          cache: false
       - name: Dependencies
         run: |
-          sudo apt-get install -y --no-install-recommends libopencv-dev
-          sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends libopencv-dev protobuf-compiler ccache
+          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@8ba23be9613c672d40ae261d2a1335d639bdd59b
+          go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.0
       - name: Build stablediffusion
         run: |
+          export PATH=$PATH:$GOPATH/bin
           make backend-assets/grpc/stablediffusion
           mkdir -p release && cp backend-assets/grpc/stablediffusion release
-      - uses: actions/upload-artifact@v3
+        env:
+          GO_TAGS: stablediffusion
+      - uses: actions/upload-artifact@v4
         with:
           name: stablediffusion
           path: release/
       - name: Release
-        uses: softprops/action-gh-release@v1
+        uses: softprops/action-gh-release@v2
         if: startsWith(github.ref, 'refs/tags/')
         with:
           files: |
             release/*
 
-  build-macOS:
-    strategy:
-      matrix:
-        include:
-          - build: 'avx2'
-            defines: ''
-          - build: 'avx'
-            defines: '-DLLAMA_AVX2=OFF'
-          - build: 'avx512'
-            defines: '-DLLAMA_AVX512=ON'
-    runs-on: macOS-latest
+  build-macOS-arm64:
+    runs-on: macos-14
     steps:
       - name: Clone
         uses: actions/checkout@v4
         with:
           submodules: true
-      - uses: actions/setup-go@v4
+      - uses: actions/setup-go@v5
         with:
-          go-version: '>=1.21.0'
+          go-version: '1.21.x'
+          cache: false
       - name: Dependencies
         run: |
           brew install protobuf grpc
+          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@8ba23be9613c672d40ae261d2a1335d639bdd59b
+          go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.0
       - name: Build
         id: build
-        env:
-          CMAKE_ARGS: "${{ matrix.defines }}"
-          BUILD_ID: "${{ matrix.build }}"
         run: |
           export C_INCLUDE_PATH=/usr/local/include
           export CPLUS_INCLUDE_PATH=/usr/local/include
-          make dist
-      - uses: actions/upload-artifact@v3
+          export PATH=$PATH:$GOPATH/bin
+          
+          BACKEND_LIBS="$(ls /opt/homebrew/opt/grpc/lib/*.dylib /opt/homebrew/opt/re2/lib/*.dylib /opt/homebrew/opt/openssl@3/lib/*.dylib /opt/homebrew/opt/protobuf/lib/*.dylib /opt/homebrew/opt/abseil/lib/*.dylib | xargs)" GO_TAGS=p2p make dist
+      - uses: actions/upload-artifact@v4
         with:
-          name: ${{ matrix.build }}
+          name: LocalAI-MacOS-arm64
           path: release/
       - name: Release
-        uses: softprops/action-gh-release@v1
+        uses: softprops/action-gh-release@v2
         if: startsWith(github.ref, 'refs/tags/')
         with:
           files: |
             release/*
+      - name: Setup tmate session if tests fail
+        if: ${{ failure() }}
+        uses: mxschmitt/action-tmate@v3.18
+        with:
+          detached: true
+          connect-timeout-seconds: 180
+          limit-access-to-actor: true

.github/workflows/secscan.yaml

@@ -0,0 +1,30 @@
+name: "Security Scan"
+
+# Run workflow each time code is pushed to your repository and on a schedule.
+# The scheduled workflow runs every at 00:00 on Sunday UTC time.
+on:
+  push:
+  schedule:
+  - cron: '0 0 * * 0'
+
+jobs:
+  tests:
+    runs-on: ubuntu-latest
+    env:
+      GO111MODULE: on
+    steps:
+      - name: Checkout Source
+        uses: actions/checkout@v4
+        if: ${{ github.actor != 'dependabot[bot]' }}
+      - name: Run Gosec Security Scanner
+        if: ${{ github.actor != 'dependabot[bot]' }}
+        uses: securego/gosec@master
+        with:
+          # we let the report trigger content trigger a failure using the GitHub Security features.
+          args: '-no-fail -fmt sarif -out results.sarif ./...'
+      - name: Upload SARIF file
+        if: ${{ github.actor != 'dependabot[bot]' }}
+        uses: github/codeql-action/upload-sarif@v3
+        with:
+          # Path to SARIF file relative to the root of the repository
+          sarif_file: results.sarif

.github/workflows/test-extra.yml

@@ -25,23 +25,16 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install build-essential ffmpeg
-          curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-             sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
-             sudo apt-get update && \
-             sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
-          sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools==1.64.0
           
-          sudo rm -rfv /usr/bin/conda || true
-
       - name: Test transformers
         run: |
-           export PATH=$PATH:/opt/conda/bin
-           make -C backend/python/transformers
-           make -C backend/python/transformers test
+           make --jobs=5 --output-sync=target -C backend/python/transformers
+           make --jobs=5 --output-sync=target -C backend/python/transformers test
 
   tests-sentencetransformers:
     runs-on: ubuntu-latest
@@ -54,23 +47,39 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install build-essential ffmpeg
-          curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-             sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
-             sudo apt-get update && \
-             sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
-          sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools==1.64.0
           
-          sudo rm -rfv /usr/bin/conda || true
-
       - name: Test sentencetransformers
         run: |
-           export PATH=$PATH:/opt/conda/bin
-           make -C backend/python/sentencetransformers
-           make -C backend/python/sentencetransformers test
+           make --jobs=5 --output-sync=target -C backend/python/sentencetransformers
+           make --jobs=5 --output-sync=target -C backend/python/sentencetransformers test
+
+
+  tests-rerankers:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Clone
+        uses: actions/checkout@v4
+        with: 
+          submodules: true
+      - name: Dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install build-essential ffmpeg
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools==1.64.0
+
+      - name: Test rerankers
+        run: |
+           make --jobs=5 --output-sync=target -C backend/python/rerankers
+           make --jobs=5 --output-sync=target -C backend/python/rerankers test
 
   tests-diffusers:
     runs-on: ubuntu-latest
@@ -82,25 +91,60 @@ jobs:
       - name: Dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential ffmpeg
-          curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-             sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
-             sudo apt-get update && \
-             sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
-          sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
-          
-          sudo rm -rfv /usr/bin/conda || true
-
+          sudo apt-get install -y build-essential ffmpeg
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          pip install --user grpcio-tools==1.64.0
       - name: Test diffusers
         run: |
-           export PATH=$PATH:/opt/conda/bin
-           make -C backend/python/diffusers
-           make -C backend/python/diffusers test
+          make --jobs=5 --output-sync=target -C backend/python/diffusers
+          make --jobs=5 --output-sync=target -C backend/python/diffusers test
 
+  tests-parler-tts:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Clone
+        uses: actions/checkout@v4
+        with: 
+          submodules: true
+      - name: Dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install build-essential ffmpeg
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools==1.64.0
+
+      - name: Test parler-tts
+        run: |
+           make --jobs=5 --output-sync=target -C backend/python/parler-tts
+           make --jobs=5 --output-sync=target -C backend/python/parler-tts test
+  
+  tests-openvoice:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Clone
+        uses: actions/checkout@v4
+        with: 
+          submodules: true
+      - name: Dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install build-essential ffmpeg
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools==1.64.0
+
+      - name: Test openvoice
+        run: |
+           make --jobs=5 --output-sync=target -C backend/python/openvoice
+           make --jobs=5 --output-sync=target -C backend/python/openvoice test
 
   tests-transformers-musicgen:
     runs-on: ubuntu-latest
@@ -113,54 +157,40 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install build-essential ffmpeg
-          curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-             sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
-             sudo apt-get update && \
-             sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
-          sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
-          
-          sudo rm -rfv /usr/bin/conda || true
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools==1.64.0
 
       - name: Test transformers-musicgen
         run: |
-           export PATH=$PATH:/opt/conda/bin
-           make -C backend/python/transformers-musicgen
-           make -C backend/python/transformers-musicgen test
+           make --jobs=5 --output-sync=target -C backend/python/transformers-musicgen
+           make --jobs=5 --output-sync=target -C backend/python/transformers-musicgen test
 
 
 
-  tests-petals:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Clone
-        uses: actions/checkout@v4
-        with: 
-          submodules: true
-      - name: Dependencies
-        run: |
-          sudo apt-get update
-          sudo apt-get install build-essential ffmpeg
-          curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-             sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
-             sudo apt-get update && \
-             sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
-          sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
-          
-          sudo rm -rfv /usr/bin/conda || true
+  # tests-petals:
+  #   runs-on: ubuntu-latest
+  #   steps:
+  #     - name: Clone
+  #       uses: actions/checkout@v4
+  #       with: 
+  #         submodules: true
+  #     - name: Dependencies
+  #       run: |
+  #         sudo apt-get update
+  #         sudo apt-get install build-essential ffmpeg
+  #         # Install UV
+  #         curl -LsSf https://astral.sh/uv/install.sh | sh
+  #         sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+  #         sudo apt-get install -y libopencv-dev
+  #         pip install --user grpcio-tools==1.64.0
 
-      - name: Test petals
-        run: |
-           export PATH=$PATH:/opt/conda/bin
-           make -C backend/python/petals
-           make -C backend/python/petals test
+  #     - name: Test petals
+  #       run: |
+  #          make --jobs=5 --output-sync=target -C backend/python/petals
+  #          make --jobs=5 --output-sync=target -C backend/python/petals test
 
            
 
@@ -215,23 +245,16 @@ jobs:
   #       run: |
   #         sudo apt-get update
   #         sudo apt-get install build-essential ffmpeg
-  #         curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-  #            sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-  #             gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-  #            sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
-  #            sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
-  #            sudo apt-get update && \
-  #            sudo apt-get install -y conda
-  #         sudo apt-get install -y ca-certificates cmake curl patch
-  #         sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
-          
-  #         sudo rm -rfv /usr/bin/conda || true
+  #         # Install UV
+  #         curl -LsSf https://astral.sh/uv/install.sh | sh
+  #         sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+  #         sudo apt-get install -y libopencv-dev
+  #         pip install --user grpcio-tools==1.64.0
 
   #     - name: Test bark
   #       run: |
-  #          export PATH=$PATH:/opt/conda/bin
-  #          make -C backend/python/bark
-  #          make -C backend/python/bark test
+  #          make --jobs=5 --output-sync=target -C backend/python/bark
+  #          make --jobs=5 --output-sync=target -C backend/python/bark test
 
            
   # Below tests needs GPU. Commented out for now
@@ -247,21 +270,15 @@ jobs:
   #       run: |
   #         sudo apt-get update
   #         sudo apt-get install build-essential ffmpeg
-  #         curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-  #            sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-  #             gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-  #            sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
-  #            sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
-  #            sudo apt-get update && \
-  #            sudo apt-get install -y conda
-  #         sudo apt-get install -y ca-certificates cmake curl patch
-  #         sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
-  #         sudo rm -rfv /usr/bin/conda || true
+  #         # Install UV
+  #         curl -LsSf https://astral.sh/uv/install.sh | sh
+  #         sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+  #         sudo apt-get install -y libopencv-dev
+  #         pip install --user grpcio-tools==1.64.0
   #     - name: Test vllm
   #       run: |
-  #          export PATH=$PATH:/opt/conda/bin
-  #          make -C backend/python/vllm
-  #          make -C backend/python/vllm test
+  #          make --jobs=5 --output-sync=target -C backend/python/vllm
+  #          make --jobs=5 --output-sync=target -C backend/python/vllm test
   tests-vallex:
     runs-on: ubuntu-latest
     steps:
@@ -273,21 +290,15 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install build-essential ffmpeg
-          curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-             sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
-             sudo apt-get update && \
-             sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
-          sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2    
-          sudo rm -rfv /usr/bin/conda || true
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          sudo apt-get install -y ca-certificates cmake curl patch python3-pip
+          sudo apt-get install -y libopencv-dev
+          pip install --user grpcio-tools==1.64.0
       - name: Test vall-e-x
         run: |
-           export PATH=$PATH:/opt/conda/bin
-           make -C backend/python/vall-e-x
-           make -C backend/python/vall-e-x test
+           make --jobs=5 --output-sync=target -C backend/python/vall-e-x
+           make --jobs=5 --output-sync=target -C backend/python/vall-e-x test
 
   tests-coqui:
     runs-on: ubuntu-latest
@@ -300,18 +311,11 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install build-essential ffmpeg
-          curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-             sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
-             sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
-             sudo apt-get update && \
-             sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch espeak espeak-ng          
-          sudo rm -rfv /usr/bin/conda || true
-
+          sudo apt-get install -y ca-certificates cmake curl patch espeak espeak-ng python3-pip
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          pip install --user grpcio-tools==1.64.0
       - name: Test coqui
         run: |
-           export PATH=$PATH:/opt/conda/bin
-           make -C backend/python/coqui
-           make -C backend/python/coqui test
+          make --jobs=5 --output-sync=target -C backend/python/coqui
+          make --jobs=5 --output-sync=target -C backend/python/coqui test

.github/workflows/test.yml

@@ -9,6 +9,9 @@ on:
     tags:
       - '*'
 
+env:
+  GRPC_VERSION: v1.64.0
+
 concurrency:
   group: ci-tests-${{ github.head_ref || github.ref }}-${{ github.repository }}
   cancel-in-progress: true
@@ -54,29 +57,48 @@ jobs:
           df -h
       - name: Clone
         uses: actions/checkout@v4
-        with: 
+        with:
           submodules: true
       - name: Setup Go ${{ matrix.go-version }}
-        uses: actions/setup-go@v4
+        uses: actions/setup-go@v5
         with:
           go-version: ${{ matrix.go-version }}
+          cache: false
       # You can test your matrix by printing the current Go version
       - name: Display Go version
         run: go version
       - name: Dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential ffmpeg
+          sudo apt-get install build-essential curl ffmpeg
           curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
              sudo install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-              gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
+             gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list' && \
              sudo /bin/bash -c 'echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list' && \
              sudo apt-get update && \
              sudo apt-get install -y conda
-          sudo apt-get install -y ca-certificates cmake curl patch
-          sudo apt-get install -y libopencv-dev && sudo ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
-          
+          # Install UV
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          sudo apt-get install -y ca-certificates cmake patch python3-pip unzip
+          sudo apt-get install -y libopencv-dev
+
+          curl -L -s https://github.com/protocolbuffers/protobuf/releases/download/v26.1/protoc-26.1-linux-x86_64.zip -o protoc.zip && \
+          unzip -j -d /usr/local/bin protoc.zip bin/protoc && \
+          rm protoc.zip
+
+          curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
+          sudo dpkg -i cuda-keyring_1.1-1_all.deb
+          sudo apt-get update
+          sudo apt-get install -y cuda-nvcc-${CUDA_VERSION} libcublas-dev-${CUDA_VERSION}
+          export CUDACXX=/usr/local/cuda/bin/nvcc
+
+          go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.0
+          go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@8ba23be9613c672d40ae261d2a1335d639bdd59b
+
+          # The python3-grpc-tools package in 22.04 is too old
+          pip install --user grpcio-tools
+
           sudo rm -rfv /usr/bin/conda || true
           PATH=$PATH:/opt/conda/bin make -C backend/python/sentencetransformers
 
@@ -85,49 +107,124 @@ jobs:
           GO_TAGS="tts" make -C sources/go-piper piper.o && \
           sudo cp -rfv sources/go-piper/piper-phonemize/pi/lib/. /usr/lib/ && \
           # Pre-build stable diffusion before we install a newer version of abseil (not compatible with stablediffusion-ncn)
-          GO_TAGS="stablediffusion tts" GRPC_BACKENDS=backend-assets/grpc/stablediffusion make build
+          PATH="$PATH:/root/go/bin" GO_TAGS="stablediffusion tts" GRPC_BACKENDS=backend-assets/grpc/stablediffusion make build
+        env:
+          CUDA_VERSION: 12-3
       - name: Cache grpc
         id: cache-grpc
-        uses: actions/cache@v3
+        uses: actions/cache@v4
         with:
           path: grpc
-          key: ${{ runner.os }}-grpc
+          key: ${{ runner.os }}-grpc-${{ env.GRPC_VERSION }}
       - name: Build grpc
         if: steps.cache-grpc.outputs.cache-hit != 'true'
         run: |
-          git clone --recurse-submodules -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+          git clone --recurse-submodules -b ${{ env.GRPC_VERSION }} --depth 1 --jobs 5 --shallow-submodules https://github.com/grpc/grpc && \
           cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
             -DgRPC_BUILD_TESTS=OFF \
-            ../.. && sudo make -j12
+            ../.. && sudo make --jobs 5
       - name: Install gRPC
         run: |
-          cd grpc && cd cmake/build && sudo make -j12 install
+          cd grpc && cd cmake/build && sudo make --jobs 5 install
       - name: Test
         run: |
-          GO_TAGS="stablediffusion tts" make test
+          PATH="$PATH:/root/go/bin" GO_TAGS="stablediffusion tts" make --jobs 5 --output-sync=target test
+      - name: Setup tmate session if tests fail
+        if: ${{ failure() }}
+        uses: mxschmitt/action-tmate@v3.18
+        with:
+          detached: true
+          connect-timeout-seconds: 180
+          limit-access-to-actor: true
+
+  tests-aio-container:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Release space from worker
+        run: |
+          echo "Listing top largest packages"
+          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+          head -n 30 <<< "${pkgs}"
+          echo
+          df -h
+          echo
+          sudo apt-get remove -y '^llvm-.*|^libllvm.*' || true
+          sudo apt-get remove --auto-remove android-sdk-platform-tools || true
+          sudo apt-get purge --auto-remove android-sdk-platform-tools || true
+          sudo rm -rf /usr/local/lib/android
+          sudo apt-get remove -y '^dotnet-.*|^aspnetcore-.*' || true
+          sudo rm -rf /usr/share/dotnet
+          sudo apt-get remove -y '^mono-.*' || true
+          sudo apt-get remove -y '^ghc-.*' || true
+          sudo apt-get remove -y '.*jdk.*|.*jre.*' || true
+          sudo apt-get remove -y 'php.*' || true
+          sudo apt-get remove -y hhvm powershell firefox monodoc-manual msbuild || true
+          sudo apt-get remove -y '^google-.*' || true
+          sudo apt-get remove -y azure-cli || true
+          sudo apt-get remove -y '^mongo.*-.*|^postgresql-.*|^mysql-.*|^mssql-.*' || true
+          sudo apt-get remove -y '^gfortran-.*' || true
+          sudo apt-get autoremove -y
+          sudo apt-get clean
+          echo
+          echo "Listing top largest packages"
+          pkgs=$(dpkg-query -Wf '${Installed-Size}\t${Package}\t${Status}\n' | awk '$NF == "installed"{print $1 "\t" $2}' | sort -nr)
+          head -n 30 <<< "${pkgs}"
+          echo
+          sudo rm -rfv build || true
+          df -h
+      - name: Clone
+        uses: actions/checkout@v4
+        with:
+          submodules: true
+      - name: Build images
+        run: |
+          docker build --build-arg FFMPEG=true --build-arg IMAGE_TYPE=extras --build-arg EXTRA_BACKENDS=rerankers --build-arg MAKEFLAGS="--jobs=5 --output-sync=target" -t local-ai:tests -f Dockerfile .
+          BASE_IMAGE=local-ai:tests DOCKER_AIO_IMAGE=local-ai-aio:test make docker-aio
+      - name: Test
+        run: |
+          LOCALAI_MODELS_DIR=$PWD/models LOCALAI_IMAGE_TAG=test LOCALAI_IMAGE=local-ai-aio \
+            make run-e2e-aio
+      - name: Setup tmate session if tests fail
+        if: ${{ failure() }}
+        uses: mxschmitt/action-tmate@v3.18
+        with:
+          detached: true
+          connect-timeout-seconds: 180
+          limit-access-to-actor: true
 
   tests-apple:
-    runs-on: macOS-latest
+    runs-on: macOS-14
     strategy:
       matrix:
         go-version: ['1.21.x']
     steps:
       - name: Clone
         uses: actions/checkout@v4
-        with: 
+        with:
           submodules: true
       - name: Setup Go ${{ matrix.go-version }}
-        uses: actions/setup-go@v4
+        uses: actions/setup-go@v5
         with:
           go-version: ${{ matrix.go-version }}
+          cache: false
       # You can test your matrix by printing the current Go version
       - name: Display Go version
         run: go version
       - name: Dependencies
         run: |
-          brew install protobuf grpc
+          brew install protobuf grpc make protoc-gen-go protoc-gen-go-grpc
+          pip install --user grpcio-tools==1.64.0
       - name: Test
         run: |
           export C_INCLUDE_PATH=/usr/local/include
           export CPLUS_INCLUDE_PATH=/usr/local/include
-          CMAKE_ARGS="-DLLAMA_F16C=OFF -DLLAMA_AVX512=OFF -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF" make test
+          # Used to run the newer GNUMake version from brew that supports --output-sync
+          export PATH="/opt/homebrew/opt/make/libexec/gnubin:$PATH"
+          BUILD_TYPE="GITHUB_CI_HAS_BROKEN_METAL" CMAKE_ARGS="-DLLAMA_F16C=OFF -DLLAMA_AVX512=OFF -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF" make --jobs 4 --output-sync=target test
+      - name: Setup tmate session if tests fail
+        if: ${{ failure() }}
+        uses: mxschmitt/action-tmate@v3.18
+        with:
+          detached: true
+          connect-timeout-seconds: 180
+          limit-access-to-actor: true

.github/workflows/update_swagger.yaml

@@ -0,0 +1,31 @@
+name: Update swagger
+on:
+  schedule:
+    - cron: 0 20 * * *
+  workflow_dispatch:
+jobs:
+  swagger:
+    strategy:
+      fail-fast: false
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-go@v5
+        with:
+          go-version: 'stable'
+      - run: |
+          go install github.com/swaggo/swag/cmd/swag@latest
+      - name: Bump swagger 🔧
+        run: |
+          make swagger
+      - name: Create Pull Request
+        uses: peter-evans/create-pull-request@v6
+        with:
+          token: ${{ secrets.UPDATE_BOT_TOKEN }}
+          push-to-fork: ci-forks/LocalAI
+          commit-message: 'feat(swagger): update swagger'
+          title: 'feat(swagger): update swagger'
+          branch: "update/swagger"
+          body:  Update swagger
+          signoff: true
+

.github/workflows/yaml-check.yml

@@ -0,0 +1,18 @@
+name: 'Yamllint GitHub Actions'
+on:
+  - pull_request
+jobs:
+  yamllint:
+    name: 'Yamllint'
+    runs-on: ubuntu-latest
+    steps:
+      - name: 'Checkout'
+        uses: actions/checkout@master
+      - name: 'Yamllint'
+        uses: karancode/yamllint-github-action@master
+        with:
+          yamllint_file_or_dir: 'gallery'
+          yamllint_strict: false
+          yamllint_comment: true
+        env:
+          GITHUB_ACCESS_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.gitignore

@@ -6,6 +6,9 @@ get-sources
 prepare-sources
 /backend/cpp/llama/grpc-server
 /backend/cpp/llama/llama.cpp
+/backend/cpp/llama-*
+
+*.log
 
 go-ggml-transformers
 go-gpt2
@@ -39,3 +42,15 @@ backend-assets/*
 !backend-assets/.keep
 prepare
 /ggml-metal.metal
+docs/static/gallery.html
+
+# Protobuf generated files
+*.pb.go
+*pb2.py
+*pb2_grpc.py
+
+# SonarQube
+.scannerwork
+
+# backend virtual environments
+**/venv

.vscode/extensions.json

@@ -0,0 +1,5 @@
+{
+    "recommendations": [
+        "golang.go"
+    ]
+}

.yamllint

@@ -0,0 +1,4 @@
+extends: default
+
+rules:
+    line-length: disable

CONTRIBUTING.md

@@ -1,4 +1,4 @@
-# Contributing to localAI
+# Contributing to LocalAI
 
 Thank you for your interest in contributing to LocalAI! We appreciate your time and effort in helping to improve our project. Before you get started, please take a moment to review these guidelines.
 
@@ -29,8 +29,9 @@ Thank you for your interest in contributing to LocalAI! We appreciate your time
 
 1. Clone the repository: `git clone https://github.com/go-skynet/LocalAI.git`
 2. Navigate to the project directory: `cd LocalAI`
-3. Install the required dependencies: `make prepare`
-4. Run LocalAI: `make run`
+3. Install the required dependencies ( see https://localai.io/basics/build/#build-localai-locally )
+4. Build LocalAI: `make build`
+5. Run LocalAI: `./local-ai`
 
 ## Contributing
 
@@ -59,14 +60,29 @@ If you find a bug, have a feature request, or encounter any issues, please check
 
 `make test` cannot handle all the model now. Please be sure to add a test case for the new features or the part was changed.
 
+### Running AIO tests
+
+All-In-One images has a set of tests that automatically verifies that most of the endpoints works correctly, a flow can be :
+
+```bash
+# Build the LocalAI docker image
+make DOCKER_IMAGE=local-ai docker
+
+# Build the corresponding AIO image
+BASE_IMAGE=local-ai DOCKER_AIO_IMAGE=local-ai-aio:test make docker-aio
+
+# Run the AIO e2e tests
+LOCALAI_IMAGE_TAG=test LOCALAI_IMAGE=local-ai-aio make run-e2e-aio
+```
+
 ## Documentation
 
-- We are welcome the contribution of the documents, please open new PR in the official document repo [localai-website](https://github.com/go-skynet/localai-website)
-
+We are welcome the contribution of the documents, please open new PR or create a new issue. The documentation is available under `docs/` https://github.com/mudler/LocalAI/tree/master/docs
+ 
 ## Community and Communication
 
 - You can reach out via the Github issue tracker.
 - Open a new discussion at [Discussion](https://github.com/go-skynet/LocalAI/discussions)
 - Join the Discord channel [Discord](https://discord.gg/uJAeKSAGDy)
 
----
+---

Dockerfile

@@ -1,30 +1,40 @@
 ARG IMAGE_TYPE=extras
 ARG BASE_IMAGE=ubuntu:22.04
+ARG GRPC_BASE_IMAGE=${BASE_IMAGE}
+ARG INTEL_BASE_IMAGE=${BASE_IMAGE}
 
-# extras or core
-FROM ${BASE_IMAGE} as requirements-core
+# The requirements-core target is common to all images.  It should not be placed in requirements-core unless every single build will use it.
+FROM ${BASE_IMAGE} AS requirements-core
 
 USER root
 
-ARG GO_VERSION=1.21.7
-ARG BUILD_TYPE
-ARG CUDA_MAJOR_VERSION=11
-ARG CUDA_MINOR_VERSION=7
+ARG GO_VERSION=1.22.4
 ARG TARGETARCH
 ARG TARGETVARIANT
 
-ENV BUILD_TYPE=${BUILD_TYPE}
 ENV DEBIAN_FRONTEND=noninteractive
-ENV EXTERNAL_GRPC_BACKENDS="coqui:/build/backend/python/coqui/run.sh,huggingface-embeddings:/build/backend/python/sentencetransformers/run.sh,petals:/build/backend/python/petals/run.sh,transformers:/build/backend/python/transformers/run.sh,sentencetransformers:/build/backend/python/sentencetransformers/run.sh,autogptq:/build/backend/python/autogptq/run.sh,bark:/build/backend/python/bark/run.sh,diffusers:/build/backend/python/diffusers/run.sh,exllama:/build/backend/python/exllama/run.sh,vall-e-x:/build/backend/python/vall-e-x/run.sh,vllm:/build/backend/python/vllm/run.sh,mamba:/build/backend/python/mamba/run.sh,exllama2:/build/backend/python/exllama2/run.sh,transformers-musicgen:/build/backend/python/transformers-musicgen/run.sh"
+ENV EXTERNAL_GRPC_BACKENDS="coqui:/build/backend/python/coqui/run.sh,huggingface-embeddings:/build/backend/python/sentencetransformers/run.sh,petals:/build/backend/python/petals/run.sh,transformers:/build/backend/python/transformers/run.sh,sentencetransformers:/build/backend/python/sentencetransformers/run.sh,rerankers:/build/backend/python/rerankers/run.sh,autogptq:/build/backend/python/autogptq/run.sh,bark:/build/backend/python/bark/run.sh,diffusers:/build/backend/python/diffusers/run.sh,exllama:/build/backend/python/exllama/run.sh,openvoice:/build/backend/python/openvoice/run.sh,vall-e-x:/build/backend/python/vall-e-x/run.sh,vllm:/build/backend/python/vllm/run.sh,mamba:/build/backend/python/mamba/run.sh,exllama2:/build/backend/python/exllama2/run.sh,transformers-musicgen:/build/backend/python/transformers-musicgen/run.sh,parler-tts:/build/backend/python/parler-tts/run.sh"
 
-ARG GO_TAGS="stablediffusion tinydream tts"
 
 RUN apt-get update && \
-    apt-get install -y ca-certificates curl patch pip cmake git && apt-get clean
+    apt-get install -y --no-install-recommends \
+        build-essential \
+        ccache \
+        ca-certificates \
+        cmake \
+        curl \
+        git \
+        unzip && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
 
 # Install Go
-RUN curl -L -s https://go.dev/dl/go$GO_VERSION.linux-$TARGETARCH.tar.gz | tar -C /usr/local -xz
-ENV PATH $PATH:/usr/local/go/bin
+RUN curl -L -s https://go.dev/dl/go${GO_VERSION}.linux-${TARGETARCH}.tar.gz | tar -C /usr/local -xz
+ENV PATH $PATH:/root/go/bin:/usr/local/go/bin
+
+# Install grpc compilers
+RUN go install google.golang.org/protobuf/cmd/protoc-gen-go@v1.34.1 && \
+    go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@1958fcbe2ca8bd93af633f11e97d44e567e945af
 
 COPY --chmod=644 custom-ca-certs/* /usr/local/share/ca-certificates/
 RUN update-ca-certificates
@@ -33,16 +43,6 @@ RUN update-ca-certificates
 RUN echo "Target Architecture: $TARGETARCH"
 RUN echo "Target Variant: $TARGETVARIANT"
 
-# CuBLAS requirements
-RUN if [ "${BUILD_TYPE}" = "cublas" ]; then \
-    apt-get install -y software-properties-common && \
-    curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb && \
-    dpkg -i cuda-keyring_1.1-1_all.deb && \
-    rm -f cuda-keyring_1.1-1_all.deb && \
-    apt-get update && \
-    apt-get install -y cuda-nvcc-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcurand-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcublas-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcusparse-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} libcusolver-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION}  && apt-get clean \
-    ; fi
-
 # Cuda
 ENV PATH /usr/local/cuda/bin:${PATH}
 
@@ -50,10 +50,12 @@ ENV PATH /usr/local/cuda/bin:${PATH}
 ENV PATH /opt/rocm/bin:${PATH}
 
 # OpenBLAS requirements and stable diffusion
-RUN apt-get install -y \
-    libopenblas-dev \
-    libopencv-dev \ 
-    && apt-get clean
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        libopenblas-dev \
+        libopencv-dev && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
 
 # Set up OpenCV
 RUN ln -s /usr/include/opencv4/opencv2 /usr/include/opencv2
@@ -63,38 +65,171 @@ WORKDIR /build
 RUN test -n "$TARGETARCH" \
     || (echo 'warn: missing $TARGETARCH, either set this `ARG` manually, or run using `docker buildkit`')
 
-# Extras requirements
-FROM requirements-core as requirements-extras
+###################################
+###################################
 
-RUN curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
-    install -o root -g root -m 644 conda.gpg /usr/share/keyrings/conda-archive-keyring.gpg && \
-    gpg --keyring /usr/share/keyrings/conda-archive-keyring.gpg --no-default-keyring --fingerprint 34161F5BF5EB1D4BFBBB8F0A8AEB4F8B29D82806 && \
-    echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" > /etc/apt/sources.list.d/conda.list && \
-    echo "deb [arch=amd64 signed-by=/usr/share/keyrings/conda-archive-keyring.gpg] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | tee -a /etc/apt/sources.list.d/conda.list && \
-    apt-get update && \
-    apt-get install -y conda && apt-get clean
+# The requirements-extras target is for any builds with IMAGE_TYPE=extras. It should not be placed in this target unless every IMAGE_TYPE=extras build will use it
+FROM requirements-core AS requirements-extras
 
+RUN curl -LsSf https://astral.sh/uv/install.sh | sh
 ENV PATH="/root/.cargo/bin:${PATH}"
-RUN apt-get install -y python3-pip && apt-get clean
-RUN pip install --upgrade pip
 
 RUN curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y
-RUN apt-get install -y espeak-ng espeak && apt-get clean
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        espeak-ng \
+        espeak \
+        python3-pip \
+        python-is-python3 \
+        python3-dev \
+        python3-venv && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/* && \
+    pip install --upgrade pip
 
-RUN if [ ! -e /usr/bin/python ]; then \
-	  ln -s /usr/bin/python3 /usr/bin/python \
+# Install grpcio-tools (the version in 22.04 is too old)
+RUN pip install --user grpcio-tools
+
+###################################
+###################################
+
+# The requirements-drivers target is for BUILD_TYPE specific items.  If you need to install something specific to CUDA, or specific to ROCM, it goes here.
+# This target will be built on top of requirements-core or requirements-extras as retermined by the IMAGE_TYPE build-arg
+FROM requirements-${IMAGE_TYPE} AS requirements-drivers
+
+ARG BUILD_TYPE
+ARG CUDA_MAJOR_VERSION=11
+ARG CUDA_MINOR_VERSION=8
+
+ENV BUILD_TYPE=${BUILD_TYPE}
+
+# CuBLAS requirements
+RUN <<EOT bash
+    if [ "${BUILD_TYPE}" = "cublas" ]; then
+        apt-get update && \
+        apt-get install -y  --no-install-recommends \
+                        software-properties-common pciutils
+        if [ "amd64" = "$TARGETARCH" ]; then
+            curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
+            fi
+        if [ "arm64" = "$TARGETARCH" ]; then
+            curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/arm64/cuda-keyring_1.1-1_all.deb
+        fi
+        dpkg -i cuda-keyring_1.1-1_all.deb && \
+            rm -f cuda-keyring_1.1-1_all.deb && \
+            apt-get update && \
+            apt-get install -y --no-install-recommends \
+                cuda-nvcc-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+                libcufft-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+                libcurand-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+                libcublas-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+                libcusparse-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+                libcusolver-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} && \
+            apt-get clean && \
+        rm -rf /var/lib/apt/lists/*
+    fi
+EOT
+
+RUN if [ "${BUILD_TYPE}" = "cublas" ]; then \
+        apt-get update && \
+        apt-get install -y  --no-install-recommends \
+            software-properties-common pciutils && \
+        curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb && \
+        dpkg -i cuda-keyring_1.1-1_all.deb && \
+        rm -f cuda-keyring_1.1-1_all.deb && \
+        apt-get update && \
+        apt-get install -y --no-install-recommends \
+            cuda-nvcc-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcufft-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcurand-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcublas-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcusparse-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} \
+            libcusolver-dev-${CUDA_MAJOR_VERSION}-${CUDA_MINOR_VERSION} && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/* \
+    ; fi
+
+# If we are building with clblas support, we need the libraries for the builds
+RUN if [ "${BUILD_TYPE}" = "clblas" ]; then \
+        apt-get update && \
+        apt-get install -y --no-install-recommends \
+            libclblast-dev && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/* \
+    ; fi
+
+RUN if [ "${BUILD_TYPE}" = "hipblas" ]; then \
+        apt-get update && \
+        apt-get install -y --no-install-recommends \
+            hipblas-dev \
+            rocblas-dev && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/* && \
+        # I have no idea why, but the ROCM lib packages don't trigger ldconfig after they install, which results in local-ai and others not being able
+        # to locate the libraries. We run ldconfig ourselves to work around this packaging deficiency
+        ldconfig \
     ; fi
 
 ###################################
 ###################################
 
-FROM requirements-${IMAGE_TYPE} as builder
+# Temporary workaround for Intel's repository to work correctly
+# https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/APT-Repository-not-working-signatures-invalid/m-p/1599436/highlight/true#M36143
+# This is a temporary workaround until Intel fixes their repository
+FROM ${INTEL_BASE_IMAGE} AS intel
+RUN wget -qO - https://repositories.intel.com/gpu/intel-graphics.key | \
+gpg --yes --dearmor --output /usr/share/keyrings/intel-graphics.gpg
+RUN echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/gpu/ubuntu jammy/lts/2350 unified" > /etc/apt/sources.list.d/intel-graphics.list
 
-ARG GO_TAGS="stablediffusion tts"
+###################################
+###################################
+
+# The grpc target does one thing, it builds and installs GRPC.  This is in it's own layer so that it can be effectively cached by CI.
+# You probably don't need to change anything here, and if you do, make sure that CI is adjusted so that the cache continues to work.
+FROM ${GRPC_BASE_IMAGE} AS grpc
+
+# This is a bit of a hack, but it's required in order to be able to effectively cache this layer in CI
+ARG GRPC_MAKEFLAGS="-j4 -Otarget"
+ARG GRPC_VERSION=v1.64.2
+
+ENV MAKEFLAGS=${GRPC_MAKEFLAGS}
+
+WORKDIR /build
+
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+        ca-certificates \
+        build-essential \
+        cmake \
+        git && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+# We install GRPC to a different prefix here so that we can copy in only the build artifacts later
+# saves several hundred MB on the final docker image size vs copying in the entire GRPC source tree
+# and running make install in the target container
+RUN git clone --recurse-submodules --jobs 4 -b ${GRPC_VERSION} --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+    mkdir -p /build/grpc/cmake/build && \
+    cd /build/grpc/cmake/build && \
+    cmake -DgRPC_INSTALL=ON -DgRPC_BUILD_TESTS=OFF -DCMAKE_INSTALL_PREFIX:PATH=/opt/grpc ../.. && \
+    make && \
+    make install && \
+    rm -rf /build
+
+###################################
+###################################
+
+# The builder target compiles LocalAI. This target is not the target that will be uploaded to the registry.
+# Adjustments to the build process should likely be made here.
+FROM requirements-drivers AS builder
+
+ARG GO_TAGS="stablediffusion tts p2p"
 ARG GRPC_BACKENDS
-ARG BUILD_GRPC=true
+ARG MAKEFLAGS
+
 ENV GRPC_BACKENDS=${GRPC_BACKENDS}
 ENV GO_TAGS=${GO_TAGS}
+ENV MAKEFLAGS=${MAKEFLAGS}
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
 ENV NVIDIA_REQUIRE_CUDA="cuda>=${CUDA_MAJOR_VERSION}.0"
 ENV NVIDIA_VISIBLE_DEVICES=all
@@ -103,63 +238,72 @@ WORKDIR /build
 
 COPY . .
 COPY .git .
+RUN echo "GO_TAGS: $GO_TAGS"
+
 RUN make prepare
 
-# If we are building with clblas support, we need the libraries for the builds
-RUN if [ "${BUILD_TYPE}" = "clblas" ]; then \
-    apt-get update && \
-    apt-get install -y libclblast-dev && \
-    apt-get clean \
-    ; fi
+# We need protoc installed, and the version in 22.04 is too old.  We will create one as part installing the GRPC build below
+# but that will also being in a newer version of absl which stablediffusion cannot compile with.  This version of protoc is only
+# here so that we can generate the grpc code for the stablediffusion build
+RUN <<EOT bash
+    if [ "amd64" = "$TARGETARCH" ]; then
+        curl -L -s https://github.com/protocolbuffers/protobuf/releases/download/v27.1/protoc-27.1-linux-x86_64.zip -o protoc.zip && \
+        unzip -j -d /usr/local/bin protoc.zip bin/protoc && \
+        rm protoc.zip
+    fi
+    if [ "arm64" = "$TARGETARCH" ]; then
+        curl -L -s https://github.com/protocolbuffers/protobuf/releases/download/v27.1/protoc-27.1-linux-aarch_64.zip -o protoc.zip && \
+        unzip -j -d /usr/local/bin protoc.zip bin/protoc && \
+        rm protoc.zip
+    fi
+EOT
 
 # stablediffusion does not tolerate a newer version of abseil, build it first
 RUN GRPC_BACKENDS=backend-assets/grpc/stablediffusion make build
 
-RUN if [ "${BUILD_GRPC}" = "true" ]; then \
-    git clone --recurse-submodules -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
-    cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
-      -DgRPC_BUILD_TESTS=OFF \
-       ../.. && make -j12 install \
-    ; fi
+# Install the pre-built GRPC
+COPY --from=grpc /opt/grpc /usr/local
 
 # Rebuild with defaults backends
+WORKDIR /build
 RUN make build
 
 RUN if [ ! -d "/build/sources/go-piper/piper-phonemize/pi/lib/" ]; then \
-    mkdir -p /build/sources/go-piper/piper-phonemize/pi/lib/ \
-    touch /build/sources/go-piper/piper-phonemize/pi/lib/keep \
+        mkdir -p /build/sources/go-piper/piper-phonemize/pi/lib/ \
+        touch /build/sources/go-piper/piper-phonemize/pi/lib/keep \
     ; fi
 
 ###################################
 ###################################
 
-FROM requirements-${IMAGE_TYPE}
+# This is the final target. The result of this target will be the image uploaded to the registry.
+# If you cannot find a more suitable place for an addition, this layer is a suitable place for it.
+FROM requirements-drivers
 
 ARG FFMPEG
 ARG BUILD_TYPE
 ARG TARGETARCH
 ARG IMAGE_TYPE=extras
+ARG EXTRA_BACKENDS
+ARG MAKEFLAGS
 
 ENV BUILD_TYPE=${BUILD_TYPE}
 ENV REBUILD=false
 ENV HEALTHCHECK_ENDPOINT=http://localhost:8080/readyz
+ENV MAKEFLAGS=${MAKEFLAGS}
 
 ARG CUDA_MAJOR_VERSION=11
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
 ENV NVIDIA_REQUIRE_CUDA="cuda>=${CUDA_MAJOR_VERSION}.0"
 ENV NVIDIA_VISIBLE_DEVICES=all
-ENV PIP_CACHE_PURGE=true
 
 # Add FFmpeg
 RUN if [ "${FFMPEG}" = "true" ]; then \
-    apt-get install -y ffmpeg && apt-get clean \
-    ; fi
-
-# Add OpenCL
-RUN if [ "${BUILD_TYPE}" = "clblas" ]; then \
-    apt-get update && \
-    apt-get install -y libclblast1 && \
-    apt-get clean \
+        apt-get update && \
+        apt-get install -y --no-install-recommends \
+            ffmpeg && \
+        apt-get clean && \
+        rm -rf /var/lib/apt/lists/* \
     ; fi
 
 WORKDIR /build
@@ -171,9 +315,9 @@ WORKDIR /build
 COPY . .
 
 COPY --from=builder /build/sources ./sources/
-COPY --from=builder /build/grpc ./grpc/
+COPY --from=grpc /opt/grpc /usr/local
 
-RUN make prepare-sources && cd /build/grpc/cmake/build && make install && rm -rf grpc
+RUN make prepare-sources
 
 # Copy the binary
 COPY --from=builder /build/local-ai ./
@@ -184,45 +328,61 @@ COPY --from=builder /build/sources/go-piper/piper-phonemize/pi/lib/* /usr/lib/
 # do not let stablediffusion rebuild (requires an older version of absl)
 COPY --from=builder /build/backend-assets/grpc/stablediffusion ./backend-assets/grpc/stablediffusion
 
-## Duplicated from Makefile to avoid having a big layer that's hard to push
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/autogptq \
+# Change the shell to bash so we can use [[ tests below
+SHELL ["/bin/bash", "-c"]
+# We try to strike a balance between individual layer size (as that affects total push time) and total image size
+# Splitting the backends into more groups with fewer items results in a larger image, but a smaller size for the largest layer
+# Splitting the backends into fewer groups with more items results in a smaller image, but a larger size for the largest layer
+
+RUN if [[ ( "${EXTRA_BACKENDS}" =~ "coqui" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/coqui \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "parler-tts" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/parler-tts \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "diffusers" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/diffusers \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "transformers-musicgen" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/transformers-musicgen \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "exllama1" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/exllama \
     ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/bark \
+
+RUN if [[ ( "${EXTRA_BACKENDS}" =~ "vall-e-x" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/vall-e-x \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "openvoice" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/openvoice \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "petals" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/petals \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "sentencetransformers" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/sentencetransformers \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "exllama2" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/exllama2 \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "transformers" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/transformers \
     ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/diffusers \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/vllm \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/mamba \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/sentencetransformers \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/transformers \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/vall-e-x \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/exllama \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-     make -C backend/python/exllama2 \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/petals \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/transformers-musicgen \
-    ; fi
-RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/coqui \
+
+RUN if [[ ( "${EXTRA_BACKENDS}" =~ "vllm" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/vllm \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "autogptq" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/autogptq \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "bark" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/bark \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "rerankers" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/rerankers \
+    ; fi && \
+    if [[ ( "${EXTRA_BACKENDS}" =~ "mamba" || -z "${EXTRA_BACKENDS}" ) && "$IMAGE_TYPE" == "extras" ]]; then \
+        make -C backend/python/mamba \
     ; fi
 
 # Make sure the models directory exists
@@ -230,7 +390,8 @@ RUN mkdir -p /build/models
 
 # Define the health check command
 HEALTHCHECK --interval=1m --timeout=10m --retries=10 \
-  CMD curl -f $HEALTHCHECK_ENDPOINT || exit 1
+  CMD curl -f ${HEALTHCHECK_ENDPOINT} || exit 1
 
+VOLUME /build/models
 EXPOSE 8080
 ENTRYPOINT [ "/build/entrypoint.sh" ]

Dockerfile.aio

@@ -0,0 +1,8 @@
+ARG BASE_IMAGE=ubuntu:22.04
+
+FROM ${BASE_IMAGE} 
+
+RUN apt-get update && apt-get install -y pciutils && apt-get clean
+
+COPY aio/ /aio
+ENTRYPOINT [ "/aio/entrypoint.sh" ]

Makefile

@@ -4,11 +4,8 @@ GOVET=$(GOCMD) vet
 BINARY_NAME=local-ai
 
 # llama.cpp versions
-GOLLAMA_VERSION?=aeba71ee842819da681ea537e78846dc75949ac0
-
-GOLLAMA_STABLE_VERSION?=50cee7712066d9e38306eccadcfbb44ea87df4b7
-
-CPPLLAMA_VERSION?=19885d205e768579ab090d1e99281cae58c21b54
+GOLLAMA_STABLE_VERSION?=2b57a8ae43e4699d3dc5d1496a1ccd42922993be
+CPPLLAMA_VERSION?=37bef8943312d91183ff06d8f1214082a17344a5
 
 # gpt4all version
 GPT4ALL_REPO?=https://github.com/nomic-ai/gpt4all
@@ -19,28 +16,29 @@ RWKV_REPO?=https://github.com/donomii/go-rwkv.cpp
 RWKV_VERSION?=661e7ae26d442f5cfebd2a0881b44e8c55949ec6
 
 # whisper.cpp version
-WHISPER_CPP_VERSION?=37a709f6558c6d9783199e2b8cbb136e1c41d346
+WHISPER_CPP_VERSION?=b29b3b29240aac8b71ce8e5a4360c1f1562ad66f
 
 # bert.cpp version
-BERT_VERSION?=6abe312cded14042f6b7c3cd8edf082713334a4d
+BERT_VERSION?=710044b124545415f555e4260d16b146c725a6e4
 
 # go-piper version
-PIPER_VERSION?=d6b6275ba037dabdba4a8b65dfdf6b2a73a67f07
+PIPER_VERSION?=9d0100873a7dbb0824dfea40e8cec70a1b110759
 
 # stablediffusion version
-STABLEDIFFUSION_VERSION?=362df9da29f882dbf09ade61972d16a1f53c3485
+STABLEDIFFUSION_VERSION?=4a3cd6aeae6f66ee57eae9a0075f8c58c3a6a38f
 
 # tinydream version
-TINYDREAM_VERSION?=772a9c0d9aaf768290e63cca3c904fe69faf677a
+TINYDREAM_VERSION?=c04fa463ace9d9a6464313aa5f9cd0f953b6c057
 
 export BUILD_TYPE?=
 export STABLE_BUILD_TYPE?=$(BUILD_TYPE)
 export CMAKE_ARGS?=
 
 CGO_LDFLAGS?=
+CGO_LDFLAGS_WHISPER?=
 CUDA_LIBPATH?=/usr/local/cuda/lib64/
 GO_TAGS?=
-BUILD_ID?=git
+BUILD_ID?=
 
 TEST_DIR=/tmp/test
 
@@ -72,7 +70,7 @@ UNAME_S := $(shell uname -s)
 endif
 
 ifeq ($(OS),Darwin)
-	CGO_LDFLAGS += -lcblas -framework Accelerate
+
 	ifeq ($(OSX_SIGNING_IDENTITY),)
 		OSX_SIGNING_IDENTITY := $(shell security find-identity -v -p codesigning | grep '"' | head -n 1 | sed -E 's/.*"(.*)"/\1/')
 	endif
@@ -83,6 +81,12 @@ ifeq ($(OS),Darwin)
 	# disable metal if on Darwin and any other value is explicitly passed.
 	else ifneq ($(BUILD_TYPE),metal)
 		CMAKE_ARGS+=-DLLAMA_METAL=OFF
+		export LLAMA_NO_ACCELERATE=1
+	endif
+
+	ifeq ($(BUILD_TYPE),metal)
+#			-lcblas 	removed: it seems to always be listed as a duplicate flag.
+		CGO_LDFLAGS += -framework Accelerate
 	endif
 endif
 
@@ -91,10 +95,12 @@ ifeq ($(BUILD_TYPE),openblas)
 	export WHISPER_OPENBLAS=1
 endif
 
+
 ifeq ($(BUILD_TYPE),cublas)
 	CGO_LDFLAGS+=-lcublas -lcudart -L$(CUDA_LIBPATH)
 	export LLAMA_CUBLAS=1
-	export WHISPER_CUBLAS=1
+	export WHISPER_CUDA=1
+	CGO_LDFLAGS_WHISPER+=-L$(CUDA_LIBPATH)/stubs/ -lcuda -lcufft
 endif
 
 ifeq ($(BUILD_TYPE),hipblas)
@@ -106,7 +112,7 @@ ifeq ($(BUILD_TYPE),hipblas)
 	# llama-ggml has no hipblas support, so override it here.
 	export STABLE_BUILD_TYPE=
 	export WHISPER_HIPBLAS=1
-	GPU_TARGETS ?= gfx900,gfx90a,gfx1030,gfx1031,gfx1100
+	GPU_TARGETS ?= gfx900,gfx906,gfx908,gfx940,gfx941,gfx942,gfx90a,gfx1030,gfx1031,gfx1100,gfx1101
 	AMDGPU_TARGETS ?= "$(GPU_TARGETS)"
 	CMAKE_ARGS+=-DLLAMA_HIPBLAS=ON -DAMDGPU_TARGETS="$(AMDGPU_TARGETS)" -DGPU_TARGETS="$(GPU_TARGETS)"
 	CGO_LDFLAGS += -O3 --rtlib=compiler-rt -unwindlib=libgcc -lhipblas -lrocblas --hip-link -L${ROCM_HOME}/lib/llvm/lib
@@ -146,14 +152,18 @@ ifeq ($(findstring tts,$(GO_TAGS)),tts)
 	OPTIONAL_GRPC+=backend-assets/grpc/piper
 endif
 
-ALL_GRPC_BACKENDS=backend-assets/grpc/langchain-huggingface
+ALL_GRPC_BACKENDS=backend-assets/grpc/huggingface
 ALL_GRPC_BACKENDS+=backend-assets/grpc/bert-embeddings
-ALL_GRPC_BACKENDS+=backend-assets/grpc/llama
-ALL_GRPC_BACKENDS+=backend-assets/grpc/llama-cpp
+ALL_GRPC_BACKENDS+=backend-assets/grpc/llama-cpp-avx
+ALL_GRPC_BACKENDS+=backend-assets/grpc/llama-cpp-avx2
+ALL_GRPC_BACKENDS+=backend-assets/grpc/llama-cpp-fallback
 ALL_GRPC_BACKENDS+=backend-assets/grpc/llama-ggml
+ALL_GRPC_BACKENDS+=backend-assets/grpc/llama-cpp-grpc
+ALL_GRPC_BACKENDS+=backend-assets/util/llama-cpp-rpc-server
 ALL_GRPC_BACKENDS+=backend-assets/grpc/gpt4all
 ALL_GRPC_BACKENDS+=backend-assets/grpc/rwkv
 ALL_GRPC_BACKENDS+=backend-assets/grpc/whisper
+ALL_GRPC_BACKENDS+=backend-assets/grpc/local-store
 ALL_GRPC_BACKENDS+=$(OPTIONAL_GRPC)
 
 GRPC_BACKENDS?=$(ALL_GRPC_BACKENDS) $(OPTIONAL_GRPC)
@@ -168,126 +178,115 @@ ifeq ($(BUILD_API_ONLY),true)
 	GRPC_BACKENDS=
 endif
 
-.PHONY: all test build vendor
+.PHONY: all test build vendor get-sources prepare-sources prepare
 
 all: help
 
-## GPT4ALL
-sources/gpt4all:
-	git clone --recurse-submodules $(GPT4ALL_REPO) sources/gpt4all
-	cd sources/gpt4all && git checkout -b build $(GPT4ALL_VERSION) && git submodule update --init --recursive --depth 1
+## BERT embeddings
+sources/go-bert.cpp:
+	git clone --recurse-submodules https://github.com/go-skynet/go-bert.cpp sources/go-bert.cpp
+	cd sources/go-bert.cpp && git checkout -b build $(BERT_VERSION) && git submodule update --init --recursive --depth 1
+
+sources/go-bert.cpp/libgobert.a: sources/go-bert.cpp
+	$(MAKE) -C sources/go-bert.cpp libgobert.a
+
+## go-llama.cpp
+sources/go-llama.cpp:
+	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp sources/go-llama.cpp
+	cd sources/go-llama.cpp && git checkout -b build $(GOLLAMA_STABLE_VERSION) && git submodule update --init --recursive --depth 1
+
+sources/go-llama.cpp/libbinding.a: sources/go-llama.cpp
+	$(MAKE) -C sources/go-llama.cpp BUILD_TYPE=$(STABLE_BUILD_TYPE) libbinding.a
 
 ## go-piper
 sources/go-piper:
 	git clone --recurse-submodules https://github.com/mudler/go-piper sources/go-piper
 	cd sources/go-piper && git checkout -b build $(PIPER_VERSION) && git submodule update --init --recursive --depth 1
 
-## BERT embeddings
-sources/go-bert:
-	git clone --recurse-submodules https://github.com/go-skynet/go-bert.cpp sources/go-bert
-	cd sources/go-bert && git checkout -b build $(BERT_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-piper/libpiper_binding.a: sources/go-piper
+	$(MAKE) -C sources/go-piper libpiper_binding.a example/main piper.o
+
+## GPT4ALL
+sources/gpt4all:
+	git clone --recurse-submodules $(GPT4ALL_REPO) sources/gpt4all
+	cd sources/gpt4all && git checkout -b build $(GPT4ALL_VERSION) && git submodule update --init --recursive --depth 1
+
+sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a: sources/gpt4all
+	$(MAKE) -C sources/gpt4all/gpt4all-bindings/golang/ libgpt4all.a
+
+## RWKV
+sources/go-rwkv.cpp:
+	git clone --recurse-submodules $(RWKV_REPO) sources/go-rwkv.cpp
+	cd sources/go-rwkv.cpp && git checkout -b build $(RWKV_VERSION) && git submodule update --init --recursive --depth 1
+
+sources/go-rwkv.cpp/librwkv.a: sources/go-rwkv.cpp
+	cd sources/go-rwkv.cpp && cd rwkv.cpp &&	cmake . -DRWKV_BUILD_SHARED_LIBRARY=OFF &&	cmake --build . && 	cp librwkv.a ..
 
 ## stable diffusion
 sources/go-stable-diffusion:
 	git clone --recurse-submodules https://github.com/mudler/go-stable-diffusion sources/go-stable-diffusion
 	cd sources/go-stable-diffusion && git checkout -b build $(STABLEDIFFUSION_VERSION) && git submodule update --init --recursive --depth 1
 
-sources/go-stable-diffusion/libstablediffusion.a:
-	$(MAKE) -C sources/go-stable-diffusion libstablediffusion.a
+sources/go-stable-diffusion/libstablediffusion.a: sources/go-stable-diffusion
+	CPATH="$(CPATH):/usr/include/opencv4" $(MAKE) -C sources/go-stable-diffusion libstablediffusion.a
 
 ## tiny-dream
 sources/go-tiny-dream:
 	git clone --recurse-submodules https://github.com/M0Rf30/go-tiny-dream sources/go-tiny-dream
 	cd sources/go-tiny-dream && git checkout -b build $(TINYDREAM_VERSION) && git submodule update --init --recursive --depth 1
 
-sources/go-tiny-dream/libtinydream.a:
+sources/go-tiny-dream/libtinydream.a: sources/go-tiny-dream
 	$(MAKE) -C sources/go-tiny-dream libtinydream.a
 
-## RWKV
-sources/go-rwkv:
-	git clone --recurse-submodules $(RWKV_REPO) sources/go-rwkv
-	cd sources/go-rwkv && git checkout -b build $(RWKV_VERSION) && git submodule update --init --recursive --depth 1
-
-sources/go-rwkv/librwkv.a: sources/go-rwkv
-	cd sources/go-rwkv && cd rwkv.cpp &&	cmake . -DRWKV_BUILD_SHARED_LIBRARY=OFF &&	cmake --build . && 	cp librwkv.a ..
-
-sources/go-bert/libgobert.a: sources/go-bert
-	$(MAKE) -C sources/go-bert libgobert.a
-
-backend-assets/gpt4all: sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a
-	mkdir -p backend-assets/gpt4all
-	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.so backend-assets/gpt4all/ || true
-	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dylib backend-assets/gpt4all/ || true
-	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dll backend-assets/gpt4all/ || true
-
-backend-assets/espeak-ng-data: sources/go-piper
-	mkdir -p backend-assets/espeak-ng-data
-	$(MAKE) -C sources/go-piper piper.o
-	@cp -rf sources/go-piper/piper-phonemize/pi/share/espeak-ng-data/. backend-assets/espeak-ng-data
-
-sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a: sources/gpt4all
-	$(MAKE) -C sources/gpt4all/gpt4all-bindings/golang/ libgpt4all.a
-
+## whisper
 sources/whisper.cpp:
-	git clone https://github.com/ggerganov/whisper.cpp.git sources/whisper.cpp
+	git clone https://github.com/ggerganov/whisper.cpp sources/whisper.cpp
 	cd sources/whisper.cpp && git checkout -b build $(WHISPER_CPP_VERSION) && git submodule update --init --recursive --depth 1
 
 sources/whisper.cpp/libwhisper.a: sources/whisper.cpp
-	cd sources/whisper.cpp && make libwhisper.a
+	cd sources/whisper.cpp && $(MAKE) libwhisper.a
 
-sources/go-llama:
-	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp sources/go-llama
-	cd sources/go-llama && git checkout -b build $(GOLLAMA_VERSION) && git submodule update --init --recursive --depth 1
-
-sources/go-llama-ggml:
-	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp sources/go-llama-ggml
-	cd sources/go-llama-ggml && git checkout -b build $(GOLLAMA_STABLE_VERSION) && git submodule update --init --recursive --depth 1
-
-sources/go-llama/libbinding.a: sources/go-llama
-	$(MAKE) -C sources/go-llama BUILD_TYPE=$(BUILD_TYPE) libbinding.a
-
-sources/go-llama-ggml/libbinding.a: sources/go-llama-ggml
-	$(MAKE) -C sources/go-llama-ggml BUILD_TYPE=$(STABLE_BUILD_TYPE) libbinding.a
-
-sources/go-piper/libpiper_binding.a: sources/go-piper
-	$(MAKE) -C sources/go-piper libpiper_binding.a example/main
-
-backend/cpp/llama/llama.cpp:
-	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama llama.cpp
-
-get-sources: backend/cpp/llama/llama.cpp sources/go-llama sources/go-llama-ggml sources/gpt4all sources/go-piper sources/go-rwkv sources/whisper.cpp sources/go-bert sources/go-stable-diffusion sources/go-tiny-dream
-	touch $@
+get-sources: sources/go-llama.cpp sources/gpt4all sources/go-piper sources/go-rwkv.cpp sources/whisper.cpp sources/go-bert.cpp sources/go-stable-diffusion sources/go-tiny-dream
 
 replace:
-	$(GOCMD) mod edit -replace github.com/nomic-ai/gpt4all/gpt4all-bindings/golang=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang
-	$(GOCMD) mod edit -replace github.com/donomii/go-rwkv.cpp=$(CURDIR)/sources/go-rwkv
+	$(GOCMD) mod edit -replace github.com/donomii/go-rwkv.cpp=$(CURDIR)/sources/go-rwkv.cpp
 	$(GOCMD) mod edit -replace github.com/ggerganov/whisper.cpp=$(CURDIR)/sources/whisper.cpp
 	$(GOCMD) mod edit -replace github.com/ggerganov/whisper.cpp/bindings/go=$(CURDIR)/sources/whisper.cpp/bindings/go
-	$(GOCMD) mod edit -replace github.com/go-skynet/go-bert.cpp=$(CURDIR)/sources/go-bert
-	$(GOCMD) mod edit -replace github.com/mudler/go-stable-diffusion=$(CURDIR)/sources/go-stable-diffusion
+	$(GOCMD) mod edit -replace github.com/go-skynet/go-bert.cpp=$(CURDIR)/sources/go-bert.cpp
 	$(GOCMD) mod edit -replace github.com/M0Rf30/go-tiny-dream=$(CURDIR)/sources/go-tiny-dream
 	$(GOCMD) mod edit -replace github.com/mudler/go-piper=$(CURDIR)/sources/go-piper
+	$(GOCMD) mod edit -replace github.com/mudler/go-stable-diffusion=$(CURDIR)/sources/go-stable-diffusion
+	$(GOCMD) mod edit -replace github.com/nomic-ai/gpt4all/gpt4all-bindings/golang=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang
+	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(CURDIR)/sources/go-llama.cpp
+
+dropreplace:
+	$(GOCMD) mod edit -dropreplace github.com/donomii/go-rwkv.cpp
+	$(GOCMD) mod edit -dropreplace github.com/ggerganov/whisper.cpp
+	$(GOCMD) mod edit -dropreplace github.com/ggerganov/whisper.cpp/bindings/go
+	$(GOCMD) mod edit -dropreplace github.com/go-skynet/go-bert.cpp
+	$(GOCMD) mod edit -dropreplace github.com/M0Rf30/go-tiny-dream
+	$(GOCMD) mod edit -dropreplace github.com/mudler/go-piper
+	$(GOCMD) mod edit -dropreplace github.com/mudler/go-stable-diffusion
+	$(GOCMD) mod edit -dropreplace github.com/nomic-ai/gpt4all/gpt4all-bindings/golang
+	$(GOCMD) mod edit -dropreplace github.com/go-skynet/go-llama.cpp
 
 prepare-sources: get-sources replace
 	$(GOCMD) mod download
-	touch $@
 
 ## GENERIC
 rebuild: ## Rebuilds the project
 	$(GOCMD) clean -cache
-	$(MAKE) -C sources/go-llama clean
-	$(MAKE) -C sources/go-llama-ggml clean
+	$(MAKE) -C sources/go-llama.cpp clean
 	$(MAKE) -C sources/gpt4all/gpt4all-bindings/golang/ clean
-	$(MAKE) -C sources/go-rwkv clean
+	$(MAKE) -C sources/go-rwkv.cpp clean
 	$(MAKE) -C sources/whisper.cpp clean
 	$(MAKE) -C sources/go-stable-diffusion clean
-	$(MAKE) -C sources/go-bert clean
+	$(MAKE) -C sources/go-bert.cpp clean
 	$(MAKE) -C sources/go-piper clean
 	$(MAKE) -C sources/go-tiny-dream clean
 	$(MAKE) build
 
 prepare: prepare-sources $(OPTIONAL_TARGETS)
-	touch $@
 
 clean: ## Remove build related file
 	$(GOCMD) clean -cache
@@ -295,22 +294,73 @@ clean: ## Remove build related file
 	rm -rf ./sources
 	rm -rf $(BINARY_NAME)
 	rm -rf release/
-	rm -rf backend-assets
+	rm -rf backend-assets/*
 	$(MAKE) -C backend/cpp/grpc clean
 	$(MAKE) -C backend/cpp/llama clean
+	rm -rf backend/cpp/llama-* || true
+	$(MAKE) dropreplace
+	$(MAKE) protogen-clean
+	rmdir pkg/grpc/proto || true
+
+clean-tests:
+	rm -rf test-models
+	rm -rf test-dir
+	rm -rf core/http/backend-assets
 
 ## Build:
-
-build: backend-assets grpcs prepare ## Build the project
+build: prepare backend-assets grpcs ## Build the project
 	$(info ${GREEN}I local-ai build info:${RESET})
 	$(info ${GREEN}I BUILD_TYPE: ${YELLOW}$(BUILD_TYPE)${RESET})
 	$(info ${GREEN}I GO_TAGS: ${YELLOW}$(GO_TAGS)${RESET})
 	$(info ${GREEN}I LD_FLAGS: ${YELLOW}$(LD_FLAGS)${RESET})
+ifneq ($(BACKEND_LIBS),)
+	$(MAKE) backend-assets/lib
+	cp $(BACKEND_LIBS) backend-assets/lib/
+endif
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o $(BINARY_NAME) ./
 
-dist: build
+build-minimal:
+	BUILD_GRPC_FOR_BACKEND_LLAMA=true GRPC_BACKENDS="backend-assets/grpc/llama-cpp-avx2" GO_TAGS=none $(MAKE) build
+
+build-api:
+	BUILD_GRPC_FOR_BACKEND_LLAMA=true BUILD_API_ONLY=true GO_TAGS=none $(MAKE) build
+
+backend-assets/lib:
+	mkdir -p backend-assets/lib
+
+dist:
+	$(MAKE) backend-assets/grpc/llama-cpp-avx2
+ifeq ($(OS),Darwin)
+	$(info ${GREEN}I Skip CUDA/hipblas build on MacOS${RESET})
+else
+	$(MAKE) backend-assets/grpc/llama-cpp-cuda
+	$(MAKE) backend-assets/grpc/llama-cpp-hipblas
+	$(MAKE) backend-assets/grpc/llama-cpp-sycl_f16
+	$(MAKE) backend-assets/grpc/llama-cpp-sycl_f32
+endif
+	STATIC=true $(MAKE) build
 	mkdir -p release
+# if BUILD_ID is empty, then we don't append it to the binary name
+ifeq ($(BUILD_ID),)
+	cp $(BINARY_NAME) release/$(BINARY_NAME)-$(OS)-$(ARCH)
+	shasum -a 256 release/$(BINARY_NAME)-$(OS)-$(ARCH) > release/$(BINARY_NAME)-$(OS)-$(ARCH).sha256
+else
 	cp $(BINARY_NAME) release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-$(ARCH)
+	shasum -a 256 release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-$(ARCH) > release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-$(ARCH).sha256
+endif
+
+dist-cross-linux-arm64: 
+	CMAKE_ARGS="$(CMAKE_ARGS) -DLLAMA_NATIVE=off" GRPC_BACKENDS="backend-assets/grpc/llama-cpp-fallback backend-assets/grpc/llama-cpp-grpc backend-assets/util/llama-cpp-rpc-server" \
+	STATIC=true $(MAKE) build
+	mkdir -p release
+# if BUILD_ID is empty, then we don't append it to the binary name
+ifeq ($(BUILD_ID),)
+	cp $(BINARY_NAME) release/$(BINARY_NAME)-$(OS)-arm64
+	shasum -a 256 release/$(BINARY_NAME)-$(OS)-arm64 > release/$(BINARY_NAME)-$(OS)-arm64.sha256
+else
+	cp $(BINARY_NAME) release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-arm64
+	shasum -a 256 release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-arm64 > release/$(BINARY_NAME)-$(BUILD_ID)-$(OS)-arm64.sha256
+endif
 
 osx-signed: build
 	codesign --deep --force --sign "$(OSX_SIGNING_IDENTITY)" --entitlements "./Entitlements.plist" "./$(BINARY_NAME)"
@@ -319,10 +369,10 @@ osx-signed: build
 run: prepare ## run local-ai
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GOCMD) run ./
 
-test-models/testmodel:
+test-models/testmodel.ggml:
 	mkdir test-models
 	mkdir test-dir
-	wget -q https://huggingface.co/TheBloke/orca_mini_3B-GGML/resolve/main/orca-mini-3b.ggmlv3.q4_0.bin -O test-models/testmodel
+	wget -q https://huggingface.co/TheBloke/orca_mini_3B-GGML/resolve/main/orca-mini-3b.ggmlv3.q4_0.bin -O test-models/testmodel.ggml
 	wget -q https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.en.bin -O test-models/whisper-en
 	wget -q https://huggingface.co/mudler/all-MiniLM-L6-v2/resolve/main/ggml-model-q4_0.bin -O test-models/bert
 	wget -q https://cdn.openai.com/whisper/draft-20220913a/micro-machines.wav -O test-dir/audio.wav
@@ -334,9 +384,9 @@ prepare-test: grpcs
 	cp -rf backend-assets core/http
 	cp tests/models_fixtures/* test-models
 
-test: prepare test-models/testmodel grpcs
+test: prepare test-models/testmodel.ggml grpcs
 	@echo 'Running tests'
-	export GO_TAGS="tts stablediffusion"
+	export GO_TAGS="tts stablediffusion debug"
 	$(MAKE) prepare-test
 	HUGGINGFACE_GRPC=$(abspath ./)/backend/python/sentencetransformers/run.sh TEST_DIR=$(abspath ./)/test-dir/ FIXTURES=$(abspath ./)/tests/fixtures CONFIG_FILE=$(abspath ./)/test-models/config.yaml MODELS_PATH=$(abspath ./)/test-models \
 	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --label-filter="!gpt4all && !llama && !llama-gguf"  --flake-attempts $(TEST_FLAKES) --fail-fast -v -r $(TEST_PATHS)
@@ -350,12 +400,16 @@ prepare-e2e:
 	mkdir -p $(TEST_DIR)
 	cp -rfv $(abspath ./tests/e2e-fixtures)/gpu.yaml $(TEST_DIR)/gpu.yaml
 	test -e $(TEST_DIR)/ggllm-test-model.bin || wget -q https://huggingface.co/TheBloke/CodeLlama-7B-Instruct-GGUF/resolve/main/codellama-7b-instruct.Q2_K.gguf -O $(TEST_DIR)/ggllm-test-model.bin
-	docker build --build-arg BUILD_GRPC=true --build-arg GRPC_BACKENDS="$(GRPC_BACKENDS)" --build-arg IMAGE_TYPE=core --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg CUDA_MAJOR_VERSION=11 --build-arg CUDA_MINOR_VERSION=7 --build-arg FFMPEG=true -t localai-tests .
+	docker build --build-arg GRPC_BACKENDS="$(GRPC_BACKENDS)" --build-arg IMAGE_TYPE=core --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg CUDA_MAJOR_VERSION=11 --build-arg CUDA_MINOR_VERSION=7 --build-arg FFMPEG=true -t localai-tests .
 
 run-e2e-image:
 	ls -liah $(abspath ./tests/e2e-fixtures)
 	docker run -p 5390:8080 -e MODELS_PATH=/models -e THREADS=1 -e DEBUG=true -d --rm -v $(TEST_DIR):/models --gpus all --name e2e-tests-$(RANDOM) localai-tests
 
+run-e2e-aio:
+	@echo 'Running e2e AIO tests'
+	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --flake-attempts 5 -v -r ./tests/e2e-aio
+
 test-e2e:
 	@echo 'Running e2e tests'
 	BUILD_TYPE=$(BUILD_TYPE) \
@@ -386,6 +440,11 @@ test-stablediffusion: prepare-test
 	TEST_DIR=$(abspath ./)/test-dir/ FIXTURES=$(abspath ./)/tests/fixtures CONFIG_FILE=$(abspath ./)/test-models/config.yaml MODELS_PATH=$(abspath ./)/test-models \
 	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --label-filter="stablediffusion" --flake-attempts 1 -v -r $(TEST_PATHS)
 
+test-stores: backend-assets/grpc/local-store
+	mkdir -p tests/integration/backend-assets/grpc
+	cp -f backend-assets/grpc/local-store tests/integration/backend-assets/grpc/
+	$(GOCMD) run github.com/onsi/ginkgo/v2/ginkgo --label-filter="stores" --flake-attempts 1 -v -r tests/integration
+
 test-container:
 	docker build --target requirements -t local-ai-test-container .
 	docker run -ti --rm --entrypoint /bin/bash -ti -v $(abspath ./):/build local-ai-test-container
@@ -402,30 +461,160 @@ help: ## Show this help.
 		else if (/^## .*$$/) {printf "  ${CYAN}%s${RESET}\n", substr($$1,4)} \
 		}' $(MAKEFILE_LIST)
 
+.PHONY: protogen
 protogen: protogen-go protogen-python
 
+.PHONY: protogen-clean
+protogen-clean: protogen-go-clean protogen-python-clean
+
+.PHONY: protogen-go
 protogen-go:
-	protoc -Ibackend/ --go_out=pkg/grpc/proto/ --go_opt=paths=source_relative --go-grpc_out=pkg/grpc/proto/ --go-grpc_opt=paths=source_relative \
+	mkdir -p pkg/grpc/proto
+	protoc --experimental_allow_proto3_optional -Ibackend/ --go_out=pkg/grpc/proto/ --go_opt=paths=source_relative --go-grpc_out=pkg/grpc/proto/ --go-grpc_opt=paths=source_relative \
     backend/backend.proto
 
-protogen-python:
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/sentencetransformers/ --grpc_python_out=backend/python/sentencetransformers/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/transformers/ --grpc_python_out=backend/python/transformers/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/transformers-musicgen/ --grpc_python_out=backend/python/transformers-musicgen/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/autogptq/ --grpc_python_out=backend/python/autogptq/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/exllama/ --grpc_python_out=backend/python/exllama/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/bark/ --grpc_python_out=backend/python/bark/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/diffusers/ --grpc_python_out=backend/python/diffusers/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/coqui/ --grpc_python_out=backend/python/coqui/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/vall-e-x/ --grpc_python_out=backend/python/vall-e-x/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/vllm/ --grpc_python_out=backend/python/vllm/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/petals/ --grpc_python_out=backend/python/petals/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/mamba/ --grpc_python_out=backend/python/mamba/ backend/backend.proto
-	python3 -m grpc_tools.protoc -Ibackend/ --python_out=backend/python/exllama2/ --grpc_python_out=backend/python/exllama2/ backend/backend.proto
+.PHONY: protogen-go-clean
+protogen-go-clean:
+	$(RM) pkg/grpc/proto/backend.pb.go pkg/grpc/proto/backend_grpc.pb.go
+	$(RM) bin/*
+
+.PHONY: protogen-python
+protogen-python: autogptq-protogen bark-protogen coqui-protogen diffusers-protogen exllama-protogen exllama2-protogen mamba-protogen petals-protogen rerankers-protogen sentencetransformers-protogen transformers-protogen parler-tts-protogen transformers-musicgen-protogen vall-e-x-protogen vllm-protogen openvoice-protogen
+
+.PHONY: protogen-python-clean
+protogen-python-clean: autogptq-protogen-clean bark-protogen-clean coqui-protogen-clean diffusers-protogen-clean exllama-protogen-clean exllama2-protogen-clean mamba-protogen-clean petals-protogen-clean sentencetransformers-protogen-clean rerankers-protogen-clean transformers-protogen-clean transformers-musicgen-protogen-clean parler-tts-protogen-clean vall-e-x-protogen-clean vllm-protogen-clean openvoice-protogen-clean
+
+.PHONY: autogptq-protogen
+autogptq-protogen:
+	$(MAKE) -C backend/python/autogptq protogen
+
+.PHONY: autogptq-protogen-clean
+autogptq-protogen-clean:
+	$(MAKE) -C backend/python/autogptq protogen-clean
+
+.PHONY: bark-protogen
+bark-protogen:
+	$(MAKE) -C backend/python/bark protogen
+
+.PHONY: bark-protogen-clean
+bark-protogen-clean:
+	$(MAKE) -C backend/python/bark protogen-clean
+
+.PHONY: coqui-protogen
+coqui-protogen:
+	$(MAKE) -C backend/python/coqui protogen
+
+.PHONY: coqui-protogen-clean
+coqui-protogen-clean:
+	$(MAKE) -C backend/python/coqui protogen-clean
+
+.PHONY: diffusers-protogen
+diffusers-protogen:
+	$(MAKE) -C backend/python/diffusers protogen
+
+.PHONY: diffusers-protogen-clean
+diffusers-protogen-clean:
+	$(MAKE) -C backend/python/diffusers protogen-clean
+
+.PHONY: exllama-protogen
+exllama-protogen:
+	$(MAKE) -C backend/python/exllama protogen
+
+.PHONY: exllama-protogen-clean
+exllama-protogen-clean:
+	$(MAKE) -C backend/python/exllama protogen-clean
+
+.PHONY: exllama2-protogen
+exllama2-protogen:
+	$(MAKE) -C backend/python/exllama2 protogen
+
+.PHONY: exllama2-protogen-clean
+exllama2-protogen-clean:
+	$(MAKE) -C backend/python/exllama2 protogen-clean
+
+.PHONY: mamba-protogen
+mamba-protogen:
+	$(MAKE) -C backend/python/mamba protogen
+
+.PHONY: mamba-protogen-clean
+mamba-protogen-clean:
+	$(MAKE) -C backend/python/mamba protogen-clean
+
+.PHONY: petals-protogen
+petals-protogen:
+	$(MAKE) -C backend/python/petals protogen
+
+.PHONY: petals-protogen-clean
+petals-protogen-clean:
+	$(MAKE) -C backend/python/petals protogen-clean
+
+.PHONY: rerankers-protogen
+rerankers-protogen:
+	$(MAKE) -C backend/python/rerankers protogen
+
+.PHONY: rerankers-protogen-clean
+rerankers-protogen-clean:
+	$(MAKE) -C backend/python/rerankers protogen-clean
+
+.PHONY: sentencetransformers-protogen
+sentencetransformers-protogen:
+	$(MAKE) -C backend/python/sentencetransformers protogen
+
+.PHONY: sentencetransformers-protogen-clean
+sentencetransformers-protogen-clean:
+	$(MAKE) -C backend/python/sentencetransformers protogen-clean
+
+.PHONY: transformers-protogen
+transformers-protogen:
+	$(MAKE) -C backend/python/transformers protogen
+
+.PHONY: transformers-protogen-clean
+transformers-protogen-clean:
+	$(MAKE) -C backend/python/transformers protogen-clean
+
+.PHONY: parler-tts-protogen
+parler-tts-protogen:
+	$(MAKE) -C backend/python/parler-tts protogen
+
+.PHONY: parler-tts-protogen-clean
+parler-tts-protogen-clean:
+	$(MAKE) -C backend/python/parler-tts protogen-clean
+
+.PHONY: transformers-musicgen-protogen
+transformers-musicgen-protogen:
+	$(MAKE) -C backend/python/transformers-musicgen protogen
+
+.PHONY: transformers-musicgen-protogen-clean
+transformers-musicgen-protogen-clean:
+	$(MAKE) -C backend/python/transformers-musicgen protogen-clean
+
+.PHONY: vall-e-x-protogen
+vall-e-x-protogen:
+	$(MAKE) -C backend/python/vall-e-x protogen
+
+.PHONY: vall-e-x-protogen-clean
+vall-e-x-protogen-clean:
+	$(MAKE) -C backend/python/vall-e-x protogen-clean
+
+.PHONY: openvoice-protogen
+openvoice-protogen:
+	$(MAKE) -C backend/python/openvoice protogen
+
+.PHONY: openvoice-protogen-clean
+openvoice-protogen-clean:
+	$(MAKE) -C backend/python/openvoice protogen-clean
+
+.PHONY: vllm-protogen
+vllm-protogen:
+	$(MAKE) -C backend/python/vllm protogen
+
+.PHONY: vllm-protogen-clean
+vllm-protogen-clean:
+	$(MAKE) -C backend/python/vllm protogen-clean
 
 ## GRPC
 # Note: it is duplicated in the Dockerfile
-prepare-extra-conda-environments:
+prepare-extra-conda-environments: protogen-python
 	$(MAKE) -C backend/python/autogptq
 	$(MAKE) -C backend/python/bark
 	$(MAKE) -C backend/python/coqui
@@ -433,14 +622,17 @@ prepare-extra-conda-environments:
 	$(MAKE) -C backend/python/vllm
 	$(MAKE) -C backend/python/mamba
 	$(MAKE) -C backend/python/sentencetransformers
+	$(MAKE) -C backend/python/rerankers
 	$(MAKE) -C backend/python/transformers
 	$(MAKE) -C backend/python/transformers-musicgen
+	$(MAKE) -C backend/python/parler-tts
 	$(MAKE) -C backend/python/vall-e-x
+	$(MAKE) -C backend/python/openvoice
 	$(MAKE) -C backend/python/exllama
 	$(MAKE) -C backend/python/petals
 	$(MAKE) -C backend/python/exllama2
 
-prepare-test-extra:
+prepare-test-extra: protogen-python
 	$(MAKE) -C backend/python/transformers
 	$(MAKE) -C backend/python/diffusers
 
@@ -454,92 +646,158 @@ ifeq ($(BUILD_API_ONLY),true)
 	touch backend-assets/keep
 endif
 
-backend-assets/grpc:
+backend-assets/espeak-ng-data: sources/go-piper sources/go-piper/libpiper_binding.a
+	mkdir -p backend-assets/espeak-ng-data
+	@cp -rf sources/go-piper/piper-phonemize/pi/share/espeak-ng-data/. backend-assets/espeak-ng-data
+
+backend-assets/gpt4all: sources/gpt4all sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a
+	mkdir -p backend-assets/gpt4all
+	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.so backend-assets/gpt4all/ || true
+	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dylib backend-assets/gpt4all/ || true
+	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dll backend-assets/gpt4all/ || true
+
+backend-assets/grpc: protogen-go replace
 	mkdir -p backend-assets/grpc
 
-backend-assets/grpc/llama: backend-assets/grpc sources/go-llama/libbinding.a
-	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(CURDIR)/sources/go-llama
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-llama LIBRARY_PATH=$(CURDIR)/sources/go-llama \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama ./backend/go/llm/llama/
-# TODO: every binary should have its own folder instead, so can have different  implementations
-ifeq ($(BUILD_TYPE),metal)
-	cp backend/cpp/llama/llama.cpp/ggml-metal.metal backend-assets/grpc/
-endif
+backend-assets/grpc/bert-embeddings: sources/go-bert.cpp sources/go-bert.cpp/libgobert.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-bert.cpp LIBRARY_PATH=$(CURDIR)/sources/go-bert.cpp \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/bert-embeddings ./backend/go/llm/bert/
 
-## BACKEND CPP LLAMA START
-# Sets the variables in case it has to build the gRPC locally.
-INSTALLED_PACKAGES=$(CURDIR)/backend/cpp/grpc/installed_packages
-INSTALLED_LIB_CMAKE=$(INSTALLED_PACKAGES)/lib/cmake
-ADDED_CMAKE_ARGS=-Dabsl_DIR=${INSTALLED_LIB_CMAKE}/absl \
-                 -DProtobuf_DIR=${INSTALLED_LIB_CMAKE}/protobuf \
-                 -Dutf8_range_DIR=${INSTALLED_LIB_CMAKE}/utf8_range \
-                 -DgRPC_DIR=${INSTALLED_LIB_CMAKE}/grpc \
-                 -DCMAKE_CXX_STANDARD_INCLUDE_DIRECTORIES=${INSTALLED_PACKAGES}/include
-
-backend/cpp/llama/grpc-server:
-ifdef BUILD_GRPC_FOR_BACKEND_LLAMA
-	$(MAKE) -C backend/cpp/grpc build
-	export _PROTOBUF_PROTOC=${INSTALLED_PACKAGES}/bin/proto && \
-	export _GRPC_CPP_PLUGIN_EXECUTABLE=${INSTALLED_PACKAGES}/bin/grpc_cpp_plugin && \
-	export PATH="${INSTALLED_PACKAGES}/bin:${PATH}" && \
-	CMAKE_ARGS="${CMAKE_ARGS} ${ADDED_CMAKE_ARGS}" LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama grpc-server
-else
-	echo "BUILD_GRPC_FOR_BACKEND_LLAMA is not defined."
-	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama grpc-server
-endif
-## BACKEND CPP LLAMA END
-
-##
-backend-assets/grpc/llama-cpp: backend-assets/grpc backend/cpp/llama/grpc-server
-	cp -rfv backend/cpp/llama/grpc-server backend-assets/grpc/llama-cpp
-# TODO: every binary should have its own folder instead, so can have different metal implementations
-ifeq ($(BUILD_TYPE),metal)
-	cp backend/cpp/llama/llama.cpp/build/bin/ggml-metal.metal backend-assets/grpc/
-endif
-
-backend-assets/grpc/llama-ggml: backend-assets/grpc sources/go-llama-ggml/libbinding.a
-	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(CURDIR)/sources/go-llama-ggml
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-llama-ggml LIBRARY_PATH=$(CURDIR)/sources/go-llama-ggml \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama-ggml ./backend/go/llm/llama-ggml/
-
-backend-assets/grpc/gpt4all: backend-assets/grpc backend-assets/gpt4all sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a
+backend-assets/grpc/gpt4all: sources/gpt4all sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a backend-assets/gpt4all backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang/ LIBRARY_PATH=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang/ \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gpt4all ./backend/go/llm/gpt4all/
 
-backend-assets/grpc/rwkv: backend-assets/grpc sources/go-rwkv/librwkv.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-rwkv LIBRARY_PATH=$(CURDIR)/sources/go-rwkv \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/rwkv ./backend/go/llm/rwkv
+backend-assets/grpc/huggingface: backend-assets/grpc
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/huggingface ./backend/go/llm/langchain/
 
-backend-assets/grpc/bert-embeddings: backend-assets/grpc sources/go-bert/libgobert.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-bert LIBRARY_PATH=$(CURDIR)/sources/go-bert \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/bert-embeddings ./backend/go/llm/bert/
+backend/cpp/llama/llama.cpp:
+	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama llama.cpp
 
-backend-assets/grpc/langchain-huggingface: backend-assets/grpc
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/langchain-huggingface ./backend/go/llm/langchain/
+INSTALLED_PACKAGES=$(CURDIR)/backend/cpp/grpc/installed_packages
+INSTALLED_LIB_CMAKE=$(INSTALLED_PACKAGES)/lib/cmake
+ADDED_CMAKE_ARGS=-Dabsl_DIR=${INSTALLED_LIB_CMAKE}/absl \
+				 -DProtobuf_DIR=${INSTALLED_LIB_CMAKE}/protobuf \
+				 -Dutf8_range_DIR=${INSTALLED_LIB_CMAKE}/utf8_range \
+				 -DgRPC_DIR=${INSTALLED_LIB_CMAKE}/grpc \
+				 -DCMAKE_CXX_STANDARD_INCLUDE_DIRECTORIES=${INSTALLED_PACKAGES}/include
+build-llama-cpp-grpc-server:
+# Conditionally build grpc for the llama backend to use if needed
+ifdef BUILD_GRPC_FOR_BACKEND_LLAMA
+	$(MAKE) -C backend/cpp/grpc build
+	_PROTOBUF_PROTOC=${INSTALLED_PACKAGES}/bin/proto \
+	_GRPC_CPP_PLUGIN_EXECUTABLE=${INSTALLED_PACKAGES}/bin/grpc_cpp_plugin \
+	PATH="${INSTALLED_PACKAGES}/bin:${PATH}" \
+	CMAKE_ARGS="${CMAKE_ARGS} ${ADDED_CMAKE_ARGS}" \
+	LLAMA_VERSION=$(CPPLLAMA_VERSION) \
+	$(MAKE) -C backend/cpp/${VARIANT} grpc-server
+else
+	echo "BUILD_GRPC_FOR_BACKEND_LLAMA is not defined."
+	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/${VARIANT} grpc-server
+endif
 
-backend-assets/grpc/stablediffusion: backend-assets/grpc
-	if [ ! -f backend-assets/grpc/stablediffusion ]; then \
-		$(MAKE) sources/go-stable-diffusion; \
-		$(MAKE) sources/go-stable-diffusion/libstablediffusion.a; \
-		CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-stable-diffusion/ LIBRARY_PATH=$(CURDIR)/sources/go-stable-diffusion/ \
-		$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/stablediffusion ./backend/go/image/stablediffusion; \
-	fi
+# This target is for manually building a variant with-auto detected flags
+backend-assets/grpc/llama-cpp: backend-assets/grpc
+	cp -rf backend/cpp/llama backend/cpp/llama-cpp
+	$(MAKE) -C backend/cpp/llama-cpp purge
+	$(info ${GREEN}I llama-cpp build info:avx2${RESET})
+	$(MAKE) VARIANT="llama-cpp" build-llama-cpp-grpc-server
+	cp -rfv backend/cpp/llama-cpp/grpc-server backend-assets/grpc/llama-cpp
 
-backend-assets/grpc/tinydream: backend-assets/grpc sources/go-tiny-dream/libtinydream.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" LIBRARY_PATH=$(CURDIR)/go-tiny-dream \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/tinydream ./backend/go/image/tinydream
+backend-assets/grpc/llama-cpp-avx2: backend-assets/grpc
+	cp -rf backend/cpp/llama backend/cpp/llama-avx2
+	$(MAKE) -C backend/cpp/llama-avx2 purge
+	$(info ${GREEN}I llama-cpp build info:avx2${RESET})
+	CMAKE_ARGS="$(CMAKE_ARGS) -DLLAMA_AVX=on -DLLAMA_AVX2=on -DLLAMA_AVX512=off -DLLAMA_FMA=on -DLLAMA_F16C=on" $(MAKE) VARIANT="llama-avx2" build-llama-cpp-grpc-server
+	cp -rfv backend/cpp/llama-avx2/grpc-server backend-assets/grpc/llama-cpp-avx2
 
-backend-assets/grpc/piper: backend-assets/grpc backend-assets/espeak-ng-data sources/go-piper/libpiper_binding.a
+backend-assets/grpc/llama-cpp-avx: backend-assets/grpc
+	cp -rf backend/cpp/llama backend/cpp/llama-avx
+	$(MAKE) -C backend/cpp/llama-avx purge
+	$(info ${GREEN}I llama-cpp build info:avx${RESET})
+	CMAKE_ARGS="$(CMAKE_ARGS) -DLLAMA_AVX=on -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off" $(MAKE) VARIANT="llama-avx" build-llama-cpp-grpc-server
+	cp -rfv backend/cpp/llama-avx/grpc-server backend-assets/grpc/llama-cpp-avx
+
+backend-assets/grpc/llama-cpp-fallback: backend-assets/grpc
+	cp -rf backend/cpp/llama backend/cpp/llama-fallback
+	$(MAKE) -C backend/cpp/llama-fallback purge
+	$(info ${GREEN}I llama-cpp build info:fallback${RESET})
+	CMAKE_ARGS="$(CMAKE_ARGS) -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off" $(MAKE) VARIANT="llama-fallback" build-llama-cpp-grpc-server
+	cp -rfv backend/cpp/llama-fallback/grpc-server backend-assets/grpc/llama-cpp-fallback
+# TODO: every binary should have its own folder instead, so can have different metal implementations
+ifeq ($(BUILD_TYPE),metal)
+	cp backend/cpp/llama-fallback/llama.cpp/build/bin/default.metallib backend-assets/grpc/
+endif
+
+backend-assets/grpc/llama-cpp-cuda: backend-assets/grpc
+	cp -rf backend/cpp/llama backend/cpp/llama-cuda
+	$(MAKE) -C backend/cpp/llama-cuda purge
+	$(info ${GREEN}I llama-cpp build info:cuda${RESET})
+	CMAKE_ARGS="$(CMAKE_ARGS) -DLLAMA_AVX=on -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off -DLLAMA_CUDA=ON" $(MAKE) VARIANT="llama-cuda" build-llama-cpp-grpc-server
+	cp -rfv backend/cpp/llama-cuda/grpc-server backend-assets/grpc/llama-cpp-cuda
+
+backend-assets/grpc/llama-cpp-hipblas: backend-assets/grpc
+	cp -rf backend/cpp/llama backend/cpp/llama-hipblas
+	$(MAKE) -C backend/cpp/llama-hipblas purge
+	$(info ${GREEN}I llama-cpp build info:hipblas${RESET})
+	BUILD_TYPE="hipblas" $(MAKE) VARIANT="llama-hipblas" build-llama-cpp-grpc-server
+	cp -rfv backend/cpp/llama-hipblas/grpc-server backend-assets/grpc/llama-cpp-hipblas
+
+backend-assets/grpc/llama-cpp-sycl_f16: backend-assets/grpc
+	cp -rf backend/cpp/llama backend/cpp/llama-sycl_f16
+	$(MAKE) -C backend/cpp/llama-sycl_f16 purge
+	$(info ${GREEN}I llama-cpp build info:sycl_f16${RESET})
+	BUILD_TYPE="sycl_f16" $(MAKE) VARIANT="llama-sycl_f16" build-llama-cpp-grpc-server
+	cp -rfv backend/cpp/llama-sycl_f16/grpc-server backend-assets/grpc/llama-cpp-sycl_f16
+
+backend-assets/grpc/llama-cpp-sycl_f32: backend-assets/grpc
+	cp -rf backend/cpp/llama backend/cpp/llama-sycl_f32
+	$(MAKE) -C backend/cpp/llama-sycl_f32 purge
+	$(info ${GREEN}I llama-cpp build info:sycl_f32${RESET})
+	BUILD_TYPE="sycl_f32" $(MAKE) VARIANT="llama-sycl_f32" build-llama-cpp-grpc-server
+	cp -rfv backend/cpp/llama-sycl_f32/grpc-server backend-assets/grpc/llama-cpp-sycl_f32
+
+backend-assets/grpc/llama-cpp-grpc: backend-assets/grpc
+	cp -rf backend/cpp/llama backend/cpp/llama-grpc
+	$(MAKE) -C backend/cpp/llama-grpc purge
+	$(info ${GREEN}I llama-cpp build info:grpc${RESET})
+	CMAKE_ARGS="$(CMAKE_ARGS) -DLLAMA_RPC=ON -DLLAMA_AVX=off -DLLAMA_AVX2=off -DLLAMA_AVX512=off -DLLAMA_FMA=off -DLLAMA_F16C=off" $(MAKE) VARIANT="llama-grpc" build-llama-cpp-grpc-server
+	cp -rfv backend/cpp/llama-grpc/grpc-server backend-assets/grpc/llama-cpp-grpc
+
+backend-assets/util/llama-cpp-rpc-server: backend-assets/grpc/llama-cpp-grpc
+	mkdir -p backend-assets/util/
+	cp -rf backend/cpp/llama-grpc/llama.cpp/build/bin/rpc-server backend-assets/util/llama-cpp-rpc-server
+
+backend-assets/grpc/llama-ggml: sources/go-llama.cpp sources/go-llama.cpp/libbinding.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-llama.cpp LIBRARY_PATH=$(CURDIR)/sources/go-llama.cpp \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama-ggml ./backend/go/llm/llama-ggml/
+
+backend-assets/grpc/piper: sources/go-piper sources/go-piper/libpiper_binding.a backend-assets/grpc backend-assets/espeak-ng-data
 	CGO_CXXFLAGS="$(PIPER_CGO_CXXFLAGS)" CGO_LDFLAGS="$(PIPER_CGO_LDFLAGS)" LIBRARY_PATH=$(CURDIR)/sources/go-piper \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/piper ./backend/go/tts/
 
-backend-assets/grpc/whisper: backend-assets/grpc sources/whisper.cpp/libwhisper.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/whisper.cpp LIBRARY_PATH=$(CURDIR)/sources/whisper.cpp \
+backend-assets/grpc/rwkv: sources/go-rwkv.cpp sources/go-rwkv.cpp/librwkv.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-rwkv.cpp LIBRARY_PATH=$(CURDIR)/sources/go-rwkv.cpp \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/rwkv ./backend/go/llm/rwkv
+
+backend-assets/grpc/stablediffusion: sources/go-stable-diffusion sources/go-stable-diffusion/libstablediffusion.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" CPATH="$(CPATH):$(CURDIR)/sources/go-stable-diffusion/:/usr/include/opencv4" LIBRARY_PATH=$(CURDIR)/sources/go-stable-diffusion/ \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/stablediffusion ./backend/go/image/stablediffusion
+
+backend-assets/grpc/tinydream: sources/go-tiny-dream sources/go-tiny-dream/libtinydream.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" LIBRARY_PATH=$(CURDIR)/go-tiny-dream \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/tinydream ./backend/go/image/tinydream
+
+backend-assets/grpc/whisper: sources/whisper.cpp sources/whisper.cpp/libwhisper.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS) $(CGO_LDFLAGS_WHISPER)" C_INCLUDE_PATH=$(CURDIR)/sources/whisper.cpp LIBRARY_PATH=$(CURDIR)/sources/whisper.cpp \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/whisper ./backend/go/transcribe/
 
+backend-assets/grpc/local-store: backend-assets/grpc
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/local-store ./backend/go/stores/
+
 grpcs: prepare $(GRPC_BACKENDS)
 
 DOCKER_IMAGE?=local-ai
+DOCKER_AIO_IMAGE?=local-ai-aio
 IMAGE_TYPE?=core
 BASE_IMAGE?=ubuntu:22.04
 
@@ -547,20 +805,60 @@ docker:
 	docker build \
 		--build-arg BASE_IMAGE=$(BASE_IMAGE) \
 		--build-arg IMAGE_TYPE=$(IMAGE_TYPE) \
-		--build-arg GO_TAGS=$(GO_TAGS) \
+		--build-arg GO_TAGS="$(GO_TAGS)" \
+		--build-arg MAKEFLAGS="$(DOCKER_MAKEFLAGS)" \
 		--build-arg BUILD_TYPE=$(BUILD_TYPE) \
 		-t $(DOCKER_IMAGE) .
 
+docker-aio:
+	@echo "Building AIO image with base $(BASE_IMAGE) as $(DOCKER_AIO_IMAGE)"
+	docker build \
+		--build-arg BASE_IMAGE=$(BASE_IMAGE) \
+		--build-arg MAKEFLAGS="$(DOCKER_MAKEFLAGS)" \
+		-t $(DOCKER_AIO_IMAGE) -f Dockerfile.aio .
+
+docker-aio-all:
+	$(MAKE) docker-aio DOCKER_AIO_SIZE=cpu
+	$(MAKE) docker-aio DOCKER_AIO_SIZE=cpu
+
 docker-image-intel:
 	docker build \
-		--build-arg BASE_IMAGE=intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04 \
+		--build-arg BASE_IMAGE=intel/oneapi-basekit:2024.1.0-devel-ubuntu22.04 \
 		--build-arg IMAGE_TYPE=$(IMAGE_TYPE) \
 		--build-arg GO_TAGS="none" \
+		--build-arg MAKEFLAGS="$(DOCKER_MAKEFLAGS)" \
 		--build-arg BUILD_TYPE=sycl_f32 -t $(DOCKER_IMAGE) .
 
 docker-image-intel-xpu:
 	docker build \
-		--build-arg BASE_IMAGE=intel/oneapi-basekit:2024.0.1-devel-ubuntu22.04 \
+		--build-arg BASE_IMAGE=intel/oneapi-basekit:2024.1.0-devel-ubuntu22.04 \
 		--build-arg IMAGE_TYPE=$(IMAGE_TYPE) \
 		--build-arg GO_TAGS="none" \
-		--build-arg BUILD_TYPE=sycl_f32 -t $(DOCKER_IMAGE) .
+		--build-arg MAKEFLAGS="$(DOCKER_MAKEFLAGS)" \
+		--build-arg BUILD_TYPE=sycl_f32 -t $(DOCKER_IMAGE) .
+
+.PHONY: swagger
+swagger:
+	swag init -g core/http/app.go --output swagger
+
+.PHONY: gen-assets
+gen-assets:
+	$(GOCMD) run core/dependencies_manager/manager.go embedded/webui_static.yaml core/http/static/assets
+
+## Documentation
+docs/layouts/_default: 
+	mkdir -p docs/layouts/_default
+
+docs/static/gallery.html: docs/layouts/_default
+	$(GOCMD) run ./.github/ci/modelslist.go ./gallery/index.yaml > docs/static/gallery.html
+
+docs/public: docs/layouts/_default docs/static/gallery.html
+	cd docs && hugo --minify
+
+docs-clean:
+	rm -rf docs/public
+	rm -rf docs/static/gallery.html
+
+.PHONY: docs
+docs: docs/static/gallery.html
+	cd docs && hugo serve

README.md

@@ -20,14 +20,14 @@
 </a>
 </p>
 
-[<img src="https://img.shields.io/badge/dockerhub-images-important.svg?logo=Docker">](https://hub.docker.com/r/localai/localai)
-[<img src="https://img.shields.io/badge/quay.io-images-important.svg?">](https://quay.io/repository/go-skynet/local-ai?tab=tags&tag=latest)
-
-> :bulb: Get help - [❓FAQ](https://localai.io/faq/) [💭Discussions](https://github.com/go-skynet/LocalAI/discussions) [:speech_balloon: Discord](https://discord.gg/uJAeKSAGDy) [:book: Documentation website](https://localai.io/)
->
-> [💻 Quickstart](https://localai.io/basics/getting_started/) [📣 News](https://localai.io/basics/news/) [ 🛫 Examples ](https://github.com/go-skynet/LocalAI/tree/master/examples/) [ 🖼️ Models ](https://localai.io/models/) [ 🚀 Roadmap ](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap)
-
-[![tests](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml)[![Build and Release](https://github.com/go-skynet/LocalAI/actions/workflows/release.yaml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/release.yaml)[![build container images](https://github.com/go-skynet/LocalAI/actions/workflows/image.yml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/image.yml)[![Bump dependencies](https://github.com/go-skynet/LocalAI/actions/workflows/bump_deps.yaml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/bump_deps.yaml)[![Artifact Hub](https://img.shields.io/endpoint?url=https://artifacthub.io/badge/repository/localai)](https://artifacthub.io/packages/search?repo=localai)
+<p align="center">
+<a href="https://hub.docker.com/r/localai/localai" target="blank">
+<img src="https://img.shields.io/badge/dockerhub-images-important.svg?logo=Docker" alt="LocalAI Docker hub"/>
+</a>
+<a href="https://quay.io/repository/go-skynet/local-ai?tab=tags&tag=latest" target="blank">
+<img src="https://img.shields.io/badge/quay.io-images-important.svg?" alt="LocalAI Quay.io"/>
+</a>
+</p>
 
 <p align="center">
 <a href="https://twitter.com/LocalAI_API" target="blank">
@@ -36,26 +36,46 @@
 <a href="https://discord.gg/uJAeKSAGDy" target="blank">
 <img src="https://dcbadge.vercel.app/api/server/uJAeKSAGDy?style=flat-square&theme=default-inverted" alt="Join LocalAI Discord Community"/>
 </a>
+</p>
 
-**LocalAI** is the free, Open Source OpenAI alternative. LocalAI act as a drop-in replacement REST API that’s compatible with OpenAI API specifications for local inferencing. It allows you to run LLMs, generate images, audio (and not only) locally or on-prem with consumer grade hardware, supporting multiple model families. Does not require GPU.
+> :bulb: Get help - [❓FAQ](https://localai.io/faq/) [💭Discussions](https://github.com/go-skynet/LocalAI/discussions) [:speech_balloon: Discord](https://discord.gg/uJAeKSAGDy) [:book: Documentation website](https://localai.io/)
+>
+> [💻 Quickstart](https://localai.io/basics/getting_started/) [📣 News](https://localai.io/basics/news/) [ 🛫 Examples ](https://github.com/go-skynet/LocalAI/tree/master/examples/) [ 🖼️ Models ](https://localai.io/models/) [ 🚀 Roadmap ](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap)
+
+[![tests](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml)[![Build and Release](https://github.com/go-skynet/LocalAI/actions/workflows/release.yaml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/release.yaml)[![build container images](https://github.com/go-skynet/LocalAI/actions/workflows/image.yml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/image.yml)[![Bump dependencies](https://github.com/go-skynet/LocalAI/actions/workflows/bump_deps.yaml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/bump_deps.yaml)[![Artifact Hub](https://img.shields.io/endpoint?url=https://artifacthub.io/badge/repository/localai)](https://artifacthub.io/packages/search?repo=localai)
+
+**LocalAI** is the free, Open Source OpenAI alternative. LocalAI act as a drop-in replacement REST API that’s compatible with OpenAI (Elevenlabs, Anthropic... ) API specifications for local AI inferencing. It allows you to run LLMs, generate images, audio (and not only) locally or on-prem with consumer grade hardware, supporting multiple model families. Does not require GPU. It is created and maintained by [Ettore Di Giacinto](https://github.com/mudler).
+
+![screen](https://github.com/mudler/LocalAI/assets/2420543/20b5ccd2-8393-44f0-aaf6-87a23806381e)
+
+```bash
+docker run -ti --name local-ai -p 8080:8080 localai/localai:latest-aio-cpu
+# Alternative images:
+# - if you have an Nvidia GPU:
+# docker run -ti --name local-ai -p 8080:8080 --gpus all localai/localai:latest-aio-gpu-nvidia-cuda-12
+# - without preconfigured models
+# docker run -ti --name local-ai -p 8080:8080 localai/localai:latest
+# - without preconfigured models for Nvidia GPUs
+# docker run -ti --name local-ai -p 8080:8080 --gpus all localai/localai:latest-gpu-nvidia-cuda-12 
+```
+
+[💻 Getting started](https://localai.io/basics/getting_started/index.html)
 
 ## 🔥🔥 Hot topics / Roadmap
 
 [Roadmap](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap)
 
-- Parallel function calling: https://github.com/mudler/LocalAI/pull/1726
-- Upload file API: https://github.com/mudler/LocalAI/pull/1703
-- Tools API support: https://github.com/mudler/LocalAI/pull/1715
-- LLaVa 1.6: https://github.com/mudler/LocalAI/pull/1714
-- ROCm container images: https://github.com/mudler/LocalAI/pull/1595
-- Intel GPU support (sycl, transformers, diffusers): https://github.com/mudler/LocalAI/issues/1653
-- Deprecation of old backends: https://github.com/mudler/LocalAI/issues/1651
-- Mamba support: https://github.com/mudler/LocalAI/pull/1589
-- Start and share models with config file: https://github.com/mudler/LocalAI/pull/1522
-- 🐸 Coqui: https://github.com/mudler/LocalAI/pull/1489
-- Img2vid https://github.com/mudler/LocalAI/pull/1442
+- 🆕 You can browse now the model gallery without LocalAI! Check out https://models.localai.io
+- 🔥🔥 Decentralized llama.cpp:  https://github.com/mudler/LocalAI/pull/2343 (peer2peer llama.cpp!) 👉 Docs  https://localai.io/features/distribute/
+- 🔥🔥 Openvoice: https://github.com/mudler/LocalAI/pull/2334
+- 🆕 Function calls without grammars and mixed mode: https://github.com/mudler/LocalAI/pull/2328
+- 🔥🔥 Distributed inferencing: https://github.com/mudler/LocalAI/pull/2324
+- Chat, TTS, and Image generation in the WebUI: https://github.com/mudler/LocalAI/pull/2222
+- Reranker API: https://github.com/mudler/LocalAI/pull/2121
 
 Hot topics (looking for contributors):
+
+- WebUI improvements: https://github.com/mudler/LocalAI/issues/2156
 - Backends v2: https://github.com/mudler/LocalAI/issues/1126
 - Improving UX v2: https://github.com/mudler/LocalAI/issues/1373
 - Assistant API: https://github.com/mudler/LocalAI/issues/1273
@@ -64,25 +84,19 @@ Hot topics (looking for contributors):
 
 If you want to help and contribute, issues up for grabs: https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3A%22up+for+grabs%22
 
-## 💻 [Getting started](https://localai.io/basics/getting_started/index.html)
-
-For a detailed step-by-step introduction, refer to the [Getting Started](https://localai.io/basics/getting_started/index.html) guide. For those in a hurry, here's a straightforward one-liner to launch a LocalAI instance with [phi-2](https://huggingface.co/microsoft/phi-2) using `docker`:
-
-```
-docker run -ti -p 8080:8080 localai/localai:v2.9.0-ffmpeg-core phi-2
-```
-
 ## 🚀 [Features](https://localai.io/features/)
 
 - 📖 [Text generation with GPTs](https://localai.io/features/text-generation/) (`llama.cpp`, `gpt4all.cpp`, ... [:book: and more](https://localai.io/model-compatibility/index.html#model-compatibility-table))
 - 🗣 [Text to Audio](https://localai.io/features/text-to-audio/)
 - 🔈 [Audio to Text](https://localai.io/features/audio-to-text/) (Audio transcription with `whisper.cpp`)
 - 🎨 [Image generation with stable diffusion](https://localai.io/features/image-generation)
-- 🔥 [OpenAI functions](https://localai.io/features/openai-functions/) 🆕
+- 🔥 [OpenAI-alike tools API](https://localai.io/features/openai-functions/) 
 - 🧠 [Embeddings generation for vector databases](https://localai.io/features/embeddings/)
 - ✍️ [Constrained grammars](https://localai.io/features/constrained_grammars/)
 - 🖼️ [Download Models directly from Huggingface ](https://localai.io/models/)
-- 🆕 [Vision API](https://localai.io/features/gpt-vision/)
+- 🥽 [Vision API](https://localai.io/features/gpt-vision/)
+- 📈 [Reranker API](https://localai.io/features/reranker/)
+- 🆕🖧 [P2P Inferencing](https://localai.io/features/distribute/)
 
 ## 💻 Usage
 
@@ -96,6 +110,7 @@ Build and deploy custom containers:
 WebUIs:
 - https://github.com/Jirubizu/localai-admin
 - https://github.com/go-skynet/LocalAI-frontend
+- QA-Pilot(An interactive chat project that leverages LocalAI LLMs for rapid understanding and navigation of GitHub code repository) https://github.com/reid41/QA-Pilot
 
 Model galleries
 - https://github.com/go-skynet/model-gallery
@@ -103,17 +118,19 @@ Model galleries
 Other:
 - Helm chart https://github.com/go-skynet/helm-charts
 - VSCode extension https://github.com/badgooooor/localai-vscode-plugin
+- Terminal utility https://github.com/djcopley/ShellOracle
 - Local Smart assistant https://github.com/mudler/LocalAGI
-- Home Assistant https://github.com/sammcj/homeassistant-localai / https://github.com/drndos/hass-openai-custom-conversation
+- Home Assistant https://github.com/sammcj/homeassistant-localai / https://github.com/drndos/hass-openai-custom-conversation / https://github.com/valentinfrlch/ha-gpt4vision
 - Discord bot https://github.com/mudler/LocalAGI/tree/main/examples/discord
 - Slack bot https://github.com/mudler/LocalAGI/tree/main/examples/slack
+- Shell-Pilot(Interact with LLM using LocalAI models via pure shell scripts on your Linux or MacOS system) https://github.com/reid41/shell-pilot
 - Telegram bot https://github.com/mudler/LocalAI/tree/master/examples/telegram-bot
 - Examples: https://github.com/mudler/LocalAI/tree/master/examples/
   
 
 ### 🔗 Resources
 
-- 🆕 New! [LLM finetuning guide](https://localai.io/docs/advanced/fine-tuning/)
+- [LLM finetuning guide](https://localai.io/docs/advanced/fine-tuning/)
 - [How to build locally](https://localai.io/basics/build/index.html)
 - [How to install in Kubernetes](https://localai.io/basics/getting_started/index.html#run-localai-in-kubernetes)
 - [Projects integrating LocalAI](https://localai.io/docs/integrations/)
@@ -121,7 +138,8 @@ Other:
 
 ## :book: 🎥 [Media, Blogs, Social](https://localai.io/basics/news/#media-blogs-social)
 
-- [Run LocalAI on AWS EKS with Pulumi](https://www.pulumi.com/ai/answers/tiZMDoZzZV6TLxgDXNBnFE/deploying-helm-charts-on-aws-eks)
+- 🆕 [Run LocalAI on Jetson Nano Devkit](https://mudler.pm/posts/local-ai-jetson-nano-devkit/)
+- [Run LocalAI on AWS EKS with Pulumi](https://www.pulumi.com/blog/low-code-llm-apps-with-local-ai-flowise-and-pulumi/)
 - [Run LocalAI on AWS](https://staleks.hashnode.dev/installing-localai-on-aws-ec2-instance)
 - [Create a slackbot for teams and OSS projects that answer to documentation](https://mudler.pm/posts/smart-slackbot-for-teams/)
 - [LocalAI meets k8sgpt](https://www.youtube.com/watch?v=PKrDNuJ_dfE)
@@ -148,17 +166,16 @@ If you utilize this repository, data in a downstream project, please consider ci
 
 Support the project by becoming [a backer or sponsor](https://github.com/sponsors/mudler). Your logo will show up here with a link to your website.
 
-A huge thank you to our generous sponsors who support this project:
+A huge thank you to our generous sponsors who support this project covering CI expenses, and our [Sponsor list](https://github.com/sponsors/mudler):
 
-| ![Spectro Cloud logo_600x600px_transparent bg](https://github.com/go-skynet/LocalAI/assets/2420543/68a6f3cb-8a65-4a4d-99b5-6417a8905512) |
-|:-----------------------------------------------:|
-|  [Spectro Cloud](https://www.spectrocloud.com/)  |
-|  Spectro Cloud kindly supports LocalAI by providing GPU and computing resources to run tests on lamdalabs!  |
-
-And a huge shout-out to individuals sponsoring the project by donating hardware or backing the project.
-
-- [Sponsor list](https://github.com/sponsors/mudler)
-- JDAM00 (donating HW for the CI)
+<p align="center">
+  <a href="https://www.spectrocloud.com/" target="blank">
+    <img height="200" src="https://github.com/go-skynet/LocalAI/assets/2420543/68a6f3cb-8a65-4a4d-99b5-6417a8905512">
+  </a>
+  <a href="https://www.premai.io/" target="blank">
+    <img height="200" src="https://github.com/mudler/LocalAI/assets/2420543/42e4ca83-661e-4f79-8e46-ae43689683d6"> <br>
+  </a>
+</p>
 
 ## 🌟 Star history
 
@@ -168,7 +185,7 @@ And a huge shout-out to individuals sponsoring the project by donating hardware
 
 LocalAI is a community-driven project created by [Ettore Di Giacinto](https://github.com/mudler/).
 
-MIT - Author Ettore Di Giacinto
+MIT - Author Ettore Di Giacinto <mudler@localai.io>
 
 ## 🙇 Acknowledgements
 

aio/cpu/README.md

@@ -0,0 +1,5 @@
+## AIO CPU size
+
+Use this image with CPU-only.
+
+Please keep using only C++ backends so the base image is as small as possible (without CUDA, cuDNN, python, etc).

aio/cpu/embeddings.yaml

@@ -0,0 +1,12 @@
+name: text-embedding-ada-002
+backend: bert-embeddings
+parameters:
+  model: huggingface://mudler/all-MiniLM-L6-v2/ggml-model-q4_0.bin
+
+usage: |
+    You can test this model with curl like this:
+
+    curl http://localhost:8080/embeddings -X POST -H "Content-Type: application/json" -d '{
+      "input": "Your text string goes here",
+      "model": "text-embedding-ada-002"
+    }'

aio/cpu/image-gen.yaml

@@ -0,0 +1,62 @@
+name: stablediffusion
+backend: stablediffusion
+parameters:
+  model: stablediffusion_assets
+
+license: "BSD-3"
+urls:
+- https://github.com/EdVince/Stable-Diffusion-NCNN
+- https://github.com/EdVince/Stable-Diffusion-NCNN/blob/main/LICENSE
+
+description: |
+     Stable Diffusion in NCNN with c++, supported txt2img and img2img
+
+download_files:
+- filename: "stablediffusion_assets/AutoencoderKL-256-256-fp16-opt.param"
+  sha256: "18ca4b66685e21406bcf64c484b3b680b4949900415536d599cc876579c85c82"
+  uri: "https://raw.githubusercontent.com/EdVince/Stable-Diffusion-NCNN/main/x86/linux/assets/AutoencoderKL-256-256-fp16-opt.param"
+- filename: "stablediffusion_assets/AutoencoderKL-512-512-fp16-opt.param"
+  sha256: "cf45f63aacf3dbbab0f59ed92a6f2c14d9a1801314631cd3abe91e3c85639a20"
+  uri: "https://raw.githubusercontent.com/EdVince/Stable-Diffusion-NCNN/main/x86/linux/assets/AutoencoderKL-512-512-fp16-opt.param"
+- filename: "stablediffusion_assets/AutoencoderKL-base-fp16.param"
+  sha256: "0254a056dce61b0c27dc9ec1b78b53bcf55315c540f55f051eb841aa992701ba"
+  uri: "https://raw.githubusercontent.com/EdVince/Stable-Diffusion-NCNN/main/x86/linux/assets/AutoencoderKL-base-fp16.param"
+- filename: "stablediffusion_assets/AutoencoderKL-encoder-512-512-fp16.bin"
+  sha256: "ddcb79a9951b9f91e05e087739ed69da2c1c4ae30ba4168cce350b49d617c9fa"
+  uri: "https://github.com/EdVince/Stable-Diffusion-NCNN/releases/download/naifu/AutoencoderKL-encoder-512-512-fp16.bin"
+- filename: "stablediffusion_assets/AutoencoderKL-fp16.bin"
+  sha256: "f02e71f80e70252734724bbfaed5c4ddd3a8ed7e61bb2175ff5f53099f0e35dd"
+  uri: "https://github.com/EdVince/Stable-Diffusion-NCNN/releases/download/naifu/AutoencoderKL-fp16.bin"
+- filename: "stablediffusion_assets/FrozenCLIPEmbedder-fp16.bin"
+  sha256: "1c9a12f4e1dd1b295a388045f7f28a2352a4d70c3dc96a542189a3dd7051fdd6"
+  uri: "https://github.com/EdVince/Stable-Diffusion-NCNN/releases/download/naifu/FrozenCLIPEmbedder-fp16.bin"
+- filename: "stablediffusion_assets/FrozenCLIPEmbedder-fp16.param"
+  sha256: "471afbe678dd1fd3fe764ef9c6eccaccb0a7d7e601f27b462aa926b20eb368c9"
+  uri: "https://raw.githubusercontent.com/EdVince/Stable-Diffusion-NCNN/main/x86/linux/assets/FrozenCLIPEmbedder-fp16.param"
+- filename: "stablediffusion_assets/log_sigmas.bin"
+  sha256: "a2089f8aa4c61f9c200feaec541ab3f5c94233b28deb6d5e8bcd974fa79b68ac"
+  uri: "https://github.com/EdVince/Stable-Diffusion-NCNN/raw/main/x86/linux/assets/log_sigmas.bin"
+- filename: "stablediffusion_assets/UNetModel-256-256-MHA-fp16-opt.param"
+  sha256: "a58c380229f09491776df837b7aa7adffc0a87821dc4708b34535da2e36e3da1"
+  uri: "https://raw.githubusercontent.com/EdVince/Stable-Diffusion-NCNN/main/x86/linux/assets/UNetModel-256-256-MHA-fp16-opt.param"
+- filename: "stablediffusion_assets/UNetModel-512-512-MHA-fp16-opt.param"
+  sha256: "f12034067062827bd7f43d1d21888d1f03905401acf6c6eea22be23c259636fa"
+  uri: "https://raw.githubusercontent.com/EdVince/Stable-Diffusion-NCNN/main/x86/linux/assets/UNetModel-512-512-MHA-fp16-opt.param"
+- filename: "stablediffusion_assets/UNetModel-base-MHA-fp16.param"
+  sha256: "696f6975de49f4325b53ce32aff81861a6d6c07cd9ce3f0aae2cc405350af38d"
+  uri: "https://raw.githubusercontent.com/EdVince/Stable-Diffusion-NCNN/main/x86/linux/assets/UNetModel-base-MHA-fp16.param"
+- filename: "stablediffusion_assets/UNetModel-MHA-fp16.bin"
+  sha256: "d618918d011bfc1f644c0f2a33bf84931bd53b28a98492b0a8ed6f3a818852c3"
+  uri: "https://github.com/EdVince/Stable-Diffusion-NCNN/releases/download/naifu/UNetModel-MHA-fp16.bin"
+- filename: "stablediffusion_assets/vocab.txt"
+  sha256: "e30e57b6f1e47616982ef898d8922be24e535b4fa3d0110477b3a6f02ebbae7d"
+  uri: "https://raw.githubusercontent.com/EdVince/Stable-Diffusion-NCNN/main/x86/linux/assets/vocab.txt"
+
+usage: |
+        curl http://localhost:8080/v1/images/generations \
+          -H "Content-Type: application/json" \
+          -d '{
+            "prompt": "<positive prompt>|<negative prompt>",
+            "step": 25,
+            "size": "512x512"
+          }'

aio/cpu/rerank.yaml

@@ -0,0 +1,27 @@
+name: jina-reranker-v1-base-en
+backend: rerankers
+parameters:
+  model: cross-encoder
+
+usage: |
+    You can test this model with curl like this:
+
+    curl http://localhost:8080/v1/rerank \
+      -H "Content-Type: application/json" \
+      -d '{
+      "model": "jina-reranker-v1-base-en",
+      "query": "Organic skincare products for sensitive skin",
+      "documents": [
+        "Eco-friendly kitchenware for modern homes",
+        "Biodegradable cleaning supplies for eco-conscious consumers",
+        "Organic cotton baby clothes for sensitive skin",
+        "Natural organic skincare range for sensitive skin",
+        "Tech gadgets for smart homes: 2024 edition",
+        "Sustainable gardening tools and compost solutions",
+        "Sensitive skin-friendly facial cleansers and toners",
+        "Organic food wraps and storage solutions",
+        "All-natural pet food for dogs with allergies",
+        "Yoga mats made from recycled materials"
+      ],
+      "top_n": 3
+    }'

aio/cpu/speech-to-text.yaml

@@ -0,0 +1,18 @@
+name: whisper-1
+backend: whisper
+parameters:
+  model: ggml-whisper-base.bin
+
+usage: |
+    ## example audio file
+    wget --quiet --show-progress -O gb1.ogg https://upload.wikimedia.org/wikipedia/commons/1/1f/George_W_Bush_Columbia_FINAL.ogg
+
+    ## Send the example audio file to the transcriptions endpoint
+    curl http://localhost:8080/v1/audio/transcriptions \
+         -H "Content-Type: multipart/form-data" \
+         -F file="@$PWD/gb1.ogg" -F model="whisper-1"
+
+download_files:
+- filename: "ggml-whisper-base.bin"
+  sha256: "60ed5bc3dd14eea856493d334349b405782ddcaf0028d4b5df4088345fba2efe"
+  uri: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.bin"

aio/cpu/text-to-speech.yaml

@@ -0,0 +1,15 @@
+name: tts-1
+download_files:
+  - filename: voice-en-us-amy-low.tar.gz
+    uri: https://github.com/rhasspy/piper/releases/download/v0.0.2/voice-en-us-amy-low.tar.gz
+
+parameters:
+  model: en-us-amy-low.onnx
+
+usage: |
+    To test if this model works as expected, you can use the following curl command:
+
+    curl http://localhost:8080/tts -H "Content-Type: application/json" -d '{
+      "model":"voice-en-us-amy-low",
+      "input": "Hi, this is a test."
+    }'

aio/cpu/text-to-text.yaml

@@ -0,0 +1,101 @@
+name: gpt-4
+mmap: true
+parameters:
+  model: huggingface://NousResearch/Hermes-2-Pro-Llama-3-8B-GGUF/Hermes-2-Pro-Llama-3-8B-Q4_K_M.gguf
+context_size: 8192
+
+stopwords:
+- "<|im_end|>"
+- "<dummy32000>"
+- "</tool_call>"
+- "<|eot_id|>"
+- "<|end_of_text|>"
+
+function:
+  # disable injecting the "answer" tool
+  disable_no_action: true
+
+  grammar:
+    # This allows the grammar to also return messages
+    mixed_mode: true
+    # Suffix to add to the grammar
+    #prefix: '<tool_call>\n'
+    # Force parallel calls in the grammar
+    # parallel_calls: true
+
+  return_name_in_function_response: true
+  # Without grammar uncomment the lines below
+  # Warning: this is relying only on the capability of the
+  # LLM model to generate the correct function call.
+  json_regex_match: 
+   - "(?s)<tool_call>(.*?)</tool_call>"
+   - "(?s)<tool_call>(.*?)"
+  replace_llm_results:
+  # Drop the scratchpad content from responses
+  - key: "(?s)<scratchpad>.*</scratchpad>"
+    value: ""
+  replace_function_results: 
+  # Replace everything that is not JSON array or object
+  # 
+  - key: '(?s)^[^{\[]*'
+    value: ""
+  - key: '(?s)[^}\]]*$'
+    value: ""
+  - key: "'([^']*?)'"
+    value: "_DQUOTE_${1}_DQUOTE_"
+  - key: '\\"'
+    value: "__TEMP_QUOTE__"
+  - key: "\'"
+    value: "'"
+  - key: "_DQUOTE_"
+    value: '"'
+  - key: "__TEMP_QUOTE__"
+    value: '"'
+  # Drop the scratchpad content from responses
+  - key: "(?s)<scratchpad>.*</scratchpad>"
+    value: ""
+
+template:
+  chat: |
+    {{.Input -}}
+    <|im_start|>assistant
+  chat_message: |
+    <|im_start|>{{if eq .RoleName "assistant"}}assistant{{else if eq .RoleName "system"}}system{{else if eq .RoleName "tool"}}tool{{else if eq .RoleName "user"}}user{{end}}
+    {{- if .FunctionCall }}
+    <tool_call>
+    {{- else if eq .RoleName "tool" }}
+    <tool_response>
+    {{- end }}
+    {{- if .Content}}
+    {{.Content }}
+    {{- end }}
+    {{- if .FunctionCall}}
+    {{toJson .FunctionCall}}
+    {{- end }}
+    {{- if .FunctionCall }}
+    </tool_call>
+    {{- else if eq .RoleName "tool" }}
+    </tool_response>
+    {{- end }}<|im_end|>
+  completion: |
+    {{.Input}}
+  function: |-
+    <|im_start|>system
+    You are a function calling AI model.
+    Here are the available tools:
+    <tools>
+    {{range .Functions}}
+    {'type': 'function', 'function': {'name': '{{.Name}}', 'description': '{{.Description}}', 'parameters': {{toJson .Parameters}} }}
+    {{end}}
+    </tools>
+    You should call the tools provided to you sequentially
+    Please use <scratchpad> XML tags to record your reasoning and planning before you call the functions as follows:
+    <scratchpad>
+    {step-by-step reasoning and plan in bullet points}
+    </scratchpad>
+    For each function call return a json object with function name and arguments within <tool_call> XML tags as follows:
+    <tool_call>
+    {"arguments": <args-dict>, "name": <function-name>}
+    </tool_call><|im_end|>
+    {{.Input -}}
+    <|im_start|>assistant

aio/cpu/vision.yaml

@@ -0,0 +1,31 @@
+backend: llama-cpp
+context_size: 4096
+f16: true
+mmap: true
+name: gpt-4-vision-preview
+
+roles:
+  user: "USER:"
+  assistant: "ASSISTANT:"
+  system: "SYSTEM:"
+
+mmproj: bakllava-mmproj.gguf
+parameters:
+  model: bakllava.gguf
+
+template:
+  chat: |
+    A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.
+    {{.Input}}
+    ASSISTANT:
+
+download_files:
+- filename: bakllava.gguf
+  uri: huggingface://mys/ggml_bakllava-1/ggml-model-q4_k.gguf
+- filename: bakllava-mmproj.gguf
+  uri: huggingface://mys/ggml_bakllava-1/mmproj-model-f16.gguf
+
+usage: |
+    curl http://localhost:8080/v1/chat/completions -H "Content-Type: application/json" -d '{
+        "model": "gpt-4-vision-preview",
+        "messages": [{"role": "user", "content": [{"type":"text", "text": "What is in the image?"}, {"type": "image_url", "image_url": {"url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg" }}], "temperature": 0.9}]}'

aio/entrypoint.sh

@@ -0,0 +1,138 @@
+#!/bin/bash
+
+echo "===> LocalAI All-in-One (AIO) container starting..."
+
+GPU_ACCELERATION=false
+GPU_VENDOR=""
+
+function check_intel() {
+    if lspci | grep -E 'VGA|3D' | grep -iq intel; then
+        echo "Intel GPU detected"
+        if [ -d /opt/intel ]; then
+            GPU_ACCELERATION=true
+            GPU_VENDOR=intel
+        else
+            echo "Intel GPU detected, but Intel GPU drivers are not installed. GPU acceleration will not be available."
+        fi
+    fi
+}
+
+function check_nvidia_wsl() {
+    if lspci | grep -E 'VGA|3D' | grep -iq "Microsoft Corporation Device 008e"; then
+        # We make the assumption this WSL2 cars is NVIDIA, then check for nvidia-smi
+        # Make sure the container was run with `--gpus all` as the only required parameter
+        echo "NVIDIA GPU detected via WSL2"
+        # nvidia-smi should be installed in the container
+        if nvidia-smi; then
+            GPU_ACCELERATION=true
+            GPU_VENDOR=nvidia
+        else
+            echo "NVIDIA GPU detected via WSL2, but nvidia-smi is not installed. GPU acceleration will not be available."
+        fi
+    fi
+}
+
+function check_amd() {
+    if lspci | grep -E 'VGA|3D' | grep -iq amd; then
+        echo "AMD GPU detected"
+        # Check if ROCm is installed
+        if [ -d /opt/rocm ]; then
+            GPU_ACCELERATION=true
+            GPU_VENDOR=amd
+        else
+            echo "AMD GPU detected, but ROCm is not installed. GPU acceleration will not be available."
+        fi
+    fi
+}
+
+function check_nvidia() {
+    if lspci | grep -E 'VGA|3D' | grep -iq nvidia; then
+        echo "NVIDIA GPU detected"
+        # nvidia-smi should be installed in the container
+        if nvidia-smi; then
+            GPU_ACCELERATION=true
+            GPU_VENDOR=nvidia
+        else
+            echo "NVIDIA GPU detected, but nvidia-smi is not installed. GPU acceleration will not be available."
+        fi
+    fi
+}
+
+function check_metal() {
+    if system_profiler SPDisplaysDataType | grep -iq 'Metal'; then
+        echo "Apple Metal supported GPU detected"
+        GPU_ACCELERATION=true
+        GPU_VENDOR=apple
+    fi
+}
+
+function detect_gpu() {
+    case "$(uname -s)" in
+        Linux)
+            check_nvidia
+            check_amd
+            check_intel
+            check_nvidia_wsl
+            ;;
+        Darwin)
+            check_metal
+            ;;
+    esac
+}
+
+function detect_gpu_size() {
+    # Attempting to find GPU memory size for NVIDIA GPUs
+    if [ "$GPU_ACCELERATION" = true ] && [ "$GPU_VENDOR" = "nvidia" ]; then
+        echo "NVIDIA GPU detected. Attempting to find memory size..."
+        # Using head -n 1 to get the total memory of the 1st NVIDIA GPU detected.
+        # If handling multiple GPUs is required in the future, this is the place to do it
+        nvidia_sm=$(nvidia-smi --query-gpu=memory.total --format=csv,noheader,nounits | head -n 1)
+        if [ ! -z "$nvidia_sm" ]; then
+            echo "Total GPU Memory: $nvidia_sm MiB"
+            # if bigger than 8GB, use 16GB
+            #if [ "$nvidia_sm" -gt 8192 ]; then
+            #    GPU_SIZE=gpu-16g
+            #else
+            GPU_SIZE=gpu-8g
+            #fi
+        else
+            echo "Unable to determine NVIDIA GPU memory size. Falling back to CPU."
+            GPU_SIZE=gpu-8g
+        fi
+    elif [ "$GPU_ACCELERATION" = true ] && [ "$GPU_VENDOR" = "intel" ]; then
+        GPU_SIZE=intel
+    # Default to a generic GPU size until we implement GPU size detection for non NVIDIA GPUs
+    elif [ "$GPU_ACCELERATION" = true ]; then
+        echo "Non-NVIDIA GPU detected. Specific GPU memory size detection is not implemented."
+        GPU_SIZE=gpu-8g
+
+    # default to cpu if GPU_SIZE is not set
+    else
+        echo "GPU acceleration is not enabled or supported. Defaulting to CPU."
+        GPU_SIZE=cpu
+    fi
+}
+
+function check_vars() {
+    if [ -z "$MODELS" ]; then
+        echo "MODELS environment variable is not set. Please set it to a comma-separated list of model YAML files to load."
+        exit 1
+    fi
+
+    if [ -z "$PROFILE" ]; then
+        echo "PROFILE environment variable is not set. Please set it to one of the following: cpu, gpu-8g, gpu-16g, apple"
+        exit 1
+    fi
+}
+
+detect_gpu
+detect_gpu_size
+
+PROFILE="${PROFILE:-$GPU_SIZE}" # default to cpu
+export MODELS="${MODELS:-/aio/${PROFILE}/embeddings.yaml,/aio/${PROFILE}/rerank.yaml,/aio/${PROFILE}/text-to-speech.yaml,/aio/${PROFILE}/image-gen.yaml,/aio/${PROFILE}/text-to-text.yaml,/aio/${PROFILE}/speech-to-text.yaml,/aio/${PROFILE}/vision.yaml}"
+
+check_vars
+
+echo "===> Starting LocalAI[$PROFILE] with the following models: $MODELS"
+
+exec /build/entrypoint.sh "$@"

aio/gpu-8g/embeddings.yaml

@@ -0,0 +1,12 @@
+name: text-embedding-ada-002
+backend: sentencetransformers
+parameters:
+  model: all-MiniLM-L6-v2
+
+usage: |
+    You can test this model with curl like this:
+
+    curl http://localhost:8080/embeddings -X POST -H "Content-Type: application/json" -d '{
+      "input": "Your text string goes here",
+      "model": "text-embedding-ada-002"
+    }'

aio/gpu-8g/image-gen.yaml

@@ -0,0 +1,25 @@
+name: stablediffusion
+parameters:
+  model: DreamShaper_8_pruned.safetensors
+backend: diffusers
+step: 25
+f16: true
+
+diffusers:
+  pipeline_type: StableDiffusionPipeline
+  cuda: true
+  enable_parameters: "negative_prompt,num_inference_steps"
+  scheduler_type: "k_dpmpp_2m"
+
+download_files:
+- filename: DreamShaper_8_pruned.safetensors
+  uri: huggingface://Lykon/DreamShaper/DreamShaper_8_pruned.safetensors
+
+usage: |
+        curl http://localhost:8080/v1/images/generations \
+          -H "Content-Type: application/json" \
+          -d '{
+            "prompt": "<positive prompt>|<negative prompt>",
+            "step": 25,
+            "size": "512x512"
+          }'

aio/gpu-8g/rerank.yaml

@@ -0,0 +1,27 @@
+name: jina-reranker-v1-base-en
+backend: rerankers
+parameters:
+  model: cross-encoder
+
+usage: |
+    You can test this model with curl like this:
+
+    curl http://localhost:8080/v1/rerank \
+      -H "Content-Type: application/json" \
+      -d '{
+      "model": "jina-reranker-v1-base-en",
+      "query": "Organic skincare products for sensitive skin",
+      "documents": [
+        "Eco-friendly kitchenware for modern homes",
+        "Biodegradable cleaning supplies for eco-conscious consumers",
+        "Organic cotton baby clothes for sensitive skin",
+        "Natural organic skincare range for sensitive skin",
+        "Tech gadgets for smart homes: 2024 edition",
+        "Sustainable gardening tools and compost solutions",
+        "Sensitive skin-friendly facial cleansers and toners",
+        "Organic food wraps and storage solutions",
+        "All-natural pet food for dogs with allergies",
+        "Yoga mats made from recycled materials"
+      ],
+      "top_n": 3
+    }'

aio/gpu-8g/speech-to-text.yaml

@@ -0,0 +1,18 @@
+name: whisper-1
+backend: whisper
+parameters:
+  model: ggml-whisper-base.bin
+
+usage: |
+    ## example audio file
+    wget --quiet --show-progress -O gb1.ogg https://upload.wikimedia.org/wikipedia/commons/1/1f/George_W_Bush_Columbia_FINAL.ogg
+
+    ## Send the example audio file to the transcriptions endpoint
+    curl http://localhost:8080/v1/audio/transcriptions \
+         -H "Content-Type: multipart/form-data" \
+         -F file="@$PWD/gb1.ogg" -F model="whisper-1"
+
+download_files:
+- filename: "ggml-whisper-base.bin"
+  sha256: "60ed5bc3dd14eea856493d334349b405782ddcaf0028d4b5df4088345fba2efe"
+  uri: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.bin"

aio/gpu-8g/text-to-speech.yaml

@@ -0,0 +1,15 @@
+name: tts-1
+download_files:
+  - filename: voice-en-us-amy-low.tar.gz
+    uri: https://github.com/rhasspy/piper/releases/download/v0.0.2/voice-en-us-amy-low.tar.gz
+
+parameters:
+  model: en-us-amy-low.onnx
+
+usage: |
+    To test if this model works as expected, you can use the following curl command:
+
+    curl http://localhost:8080/tts -H "Content-Type: application/json" -d '{
+      "model":"tts-1",
+      "input": "Hi, this is a test."
+    }'

aio/gpu-8g/text-to-text.yaml

@@ -0,0 +1,101 @@
+name: gpt-4
+mmap: true
+parameters:
+  model: huggingface://NousResearch/Hermes-2-Pro-Llama-3-8B-GGUF/Hermes-2-Pro-Llama-3-8B-Q4_K_M.gguf
+context_size: 8192
+
+stopwords:
+- "<|im_end|>"
+- "<dummy32000>"
+- "</tool_call>"
+- "<|eot_id|>"
+- "<|end_of_text|>"
+
+function:
+  # disable injecting the "answer" tool
+  disable_no_action: true
+
+  grammar:
+    # This allows the grammar to also return messages
+    mixed_mode: true
+    # Suffix to add to the grammar
+    #prefix: '<tool_call>\n'
+    # Force parallel calls in the grammar
+    # parallel_calls: true
+
+  return_name_in_function_response: true
+  # Without grammar uncomment the lines below
+  # Warning: this is relying only on the capability of the
+  # LLM model to generate the correct function call.
+  json_regex_match: 
+   - "(?s)<tool_call>(.*?)</tool_call>"
+   - "(?s)<tool_call>(.*?)"
+  replace_llm_results:
+  # Drop the scratchpad content from responses
+  - key: "(?s)<scratchpad>.*</scratchpad>"
+    value: ""
+  replace_function_results: 
+  # Replace everything that is not JSON array or object
+  # 
+  - key: '(?s)^[^{\[]*'
+    value: ""
+  - key: '(?s)[^}\]]*$'
+    value: ""
+  - key: "'([^']*?)'"
+    value: "_DQUOTE_${1}_DQUOTE_"
+  - key: '\\"'
+    value: "__TEMP_QUOTE__"
+  - key: "\'"
+    value: "'"
+  - key: "_DQUOTE_"
+    value: '"'
+  - key: "__TEMP_QUOTE__"
+    value: '"'
+  # Drop the scratchpad content from responses
+  - key: "(?s)<scratchpad>.*</scratchpad>"
+    value: ""
+
+template:
+  chat: |
+    {{.Input -}}
+    <|im_start|>assistant
+  chat_message: |
+    <|im_start|>{{if eq .RoleName "assistant"}}assistant{{else if eq .RoleName "system"}}system{{else if eq .RoleName "tool"}}tool{{else if eq .RoleName "user"}}user{{end}}
+    {{- if .FunctionCall }}
+    <tool_call>
+    {{- else if eq .RoleName "tool" }}
+    <tool_response>
+    {{- end }}
+    {{- if .Content}}
+    {{.Content }}
+    {{- end }}
+    {{- if .FunctionCall}}
+    {{toJson .FunctionCall}}
+    {{- end }}
+    {{- if .FunctionCall }}
+    </tool_call>
+    {{- else if eq .RoleName "tool" }}
+    </tool_response>
+    {{- end }}<|im_end|>
+  completion: |
+    {{.Input}}
+  function: |-
+    <|im_start|>system
+    You are a function calling AI model.
+    Here are the available tools:
+    <tools>
+    {{range .Functions}}
+    {'type': 'function', 'function': {'name': '{{.Name}}', 'description': '{{.Description}}', 'parameters': {{toJson .Parameters}} }}
+    {{end}}
+    </tools>
+    You should call the tools provided to you sequentially
+    Please use <scratchpad> XML tags to record your reasoning and planning before you call the functions as follows:
+    <scratchpad>
+    {step-by-step reasoning and plan in bullet points}
+    </scratchpad>
+    For each function call return a json object with function name and arguments within <tool_call> XML tags as follows:
+    <tool_call>
+    {"arguments": <args-dict>, "name": <function-name>}
+    </tool_call><|im_end|>
+    {{.Input -}}
+    <|im_start|>assistant

aio/gpu-8g/vision.yaml

@@ -0,0 +1,35 @@
+backend: llama-cpp
+context_size: 4096
+f16: true
+mmap: true
+name: gpt-4-vision-preview
+
+roles:
+  user: "USER:"
+  assistant: "ASSISTANT:"
+  system: "SYSTEM:"
+
+mmproj: llava-v1.6-7b-mmproj-f16.gguf
+parameters:
+  model: llava-v1.6-mistral-7b.Q5_K_M.gguf
+  temperature: 0.2
+  top_k: 40
+  top_p: 0.95
+  seed: -1
+
+template:
+  chat: |
+    A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.
+    {{.Input}}
+    ASSISTANT:
+
+download_files:
+- filename: llava-v1.6-mistral-7b.Q5_K_M.gguf
+  uri: huggingface://cjpais/llava-1.6-mistral-7b-gguf/llava-v1.6-mistral-7b.Q5_K_M.gguf
+- filename: llava-v1.6-7b-mmproj-f16.gguf
+  uri: huggingface://cjpais/llava-1.6-mistral-7b-gguf/mmproj-model-f16.gguf
+
+usage: |
+    curl http://localhost:8080/v1/chat/completions -H "Content-Type: application/json" -d '{
+        "model": "gpt-4-vision-preview",
+        "messages": [{"role": "user", "content": [{"type":"text", "text": "What is in the image?"}, {"type": "image_url", "image_url": {"url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg" }}], "temperature": 0.9}]}'

aio/intel/embeddings.yaml

@@ -0,0 +1,12 @@
+name: text-embedding-ada-002
+backend: sentencetransformers
+parameters:
+  model: all-MiniLM-L6-v2
+
+usage: |
+    You can test this model with curl like this:
+
+    curl http://localhost:8080/embeddings -X POST -H "Content-Type: application/json" -d '{
+      "input": "Your text string goes here",
+      "model": "text-embedding-ada-002"
+    }'

aio/intel/image-gen.yaml

@@ -0,0 +1,20 @@
+name: stablediffusion
+parameters:
+  model: runwayml/stable-diffusion-v1-5
+backend: diffusers
+step: 25
+f16: true
+diffusers:
+  pipeline_type: StableDiffusionPipeline
+  cuda: true
+  enable_parameters: "negative_prompt,num_inference_steps"
+  scheduler_type: "k_dpmpp_2m"
+
+usage: |
+        curl http://localhost:8080/v1/images/generations \
+          -H "Content-Type: application/json" \
+          -d '{
+            "prompt": "<positive prompt>|<negative prompt>",
+            "step": 25,
+            "size": "512x512"
+          }'

aio/intel/rerank.yaml

@@ -0,0 +1,27 @@
+name: jina-reranker-v1-base-en
+backend: rerankers
+parameters:
+  model: cross-encoder
+
+usage: |
+    You can test this model with curl like this:
+
+    curl http://localhost:8080/v1/rerank \
+      -H "Content-Type: application/json" \
+      -d '{
+      "model": "jina-reranker-v1-base-en",
+      "query": "Organic skincare products for sensitive skin",
+      "documents": [
+        "Eco-friendly kitchenware for modern homes",
+        "Biodegradable cleaning supplies for eco-conscious consumers",
+        "Organic cotton baby clothes for sensitive skin",
+        "Natural organic skincare range for sensitive skin",
+        "Tech gadgets for smart homes: 2024 edition",
+        "Sustainable gardening tools and compost solutions",
+        "Sensitive skin-friendly facial cleansers and toners",
+        "Organic food wraps and storage solutions",
+        "All-natural pet food for dogs with allergies",
+        "Yoga mats made from recycled materials"
+      ],
+      "top_n": 3
+    }'

aio/intel/speech-to-text.yaml

@@ -0,0 +1,18 @@
+name: whisper-1
+backend: whisper
+parameters:
+  model: ggml-whisper-base.bin
+
+usage: |
+    ## example audio file
+    wget --quiet --show-progress -O gb1.ogg https://upload.wikimedia.org/wikipedia/commons/1/1f/George_W_Bush_Columbia_FINAL.ogg
+
+    ## Send the example audio file to the transcriptions endpoint
+    curl http://localhost:8080/v1/audio/transcriptions \
+         -H "Content-Type: multipart/form-data" \
+         -F file="@$PWD/gb1.ogg" -F model="whisper-1"
+
+download_files:
+- filename: "ggml-whisper-base.bin"
+  sha256: "60ed5bc3dd14eea856493d334349b405782ddcaf0028d4b5df4088345fba2efe"
+  uri: "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.bin"

aio/intel/text-to-speech.yaml

@@ -0,0 +1,15 @@
+name: tts-1
+download_files:
+  - filename: voice-en-us-amy-low.tar.gz
+    uri: https://github.com/rhasspy/piper/releases/download/v0.0.2/voice-en-us-amy-low.tar.gz
+
+parameters:
+  model: en-us-amy-low.onnx
+
+usage: |
+    To test if this model works as expected, you can use the following curl command:
+
+    curl http://localhost:8080/tts -H "Content-Type: application/json" -d '{
+      "model":"tts-1",
+      "input": "Hi, this is a test."
+    }'

aio/intel/text-to-text.yaml

@@ -0,0 +1,103 @@
+name: gpt-4
+mmap: false
+context_size: 8192
+
+f16: false
+parameters:
+  model: huggingface://NousResearch/Hermes-2-Pro-Llama-3-8B-GGUF/Hermes-2-Pro-Llama-3-8B-Q4_K_M.gguf
+
+stopwords:
+- "<|im_end|>"
+- "<dummy32000>"
+- "</tool_call>"
+- "<|eot_id|>"
+- "<|end_of_text|>"
+
+function:
+  # disable injecting the "answer" tool
+  disable_no_action: true
+
+  grammar:
+    # This allows the grammar to also return messages
+    mixed_mode: true
+    # Suffix to add to the grammar
+    #prefix: '<tool_call>\n'
+    # Force parallel calls in the grammar
+    # parallel_calls: true
+
+  return_name_in_function_response: true
+  # Without grammar uncomment the lines below
+  # Warning: this is relying only on the capability of the
+  # LLM model to generate the correct function call.
+  json_regex_match: 
+   - "(?s)<tool_call>(.*?)</tool_call>"
+   - "(?s)<tool_call>(.*?)"
+  replace_llm_results:
+  # Drop the scratchpad content from responses
+  - key: "(?s)<scratchpad>.*</scratchpad>"
+    value: ""
+  replace_function_results: 
+  # Replace everything that is not JSON array or object
+  # 
+  - key: '(?s)^[^{\[]*'
+    value: ""
+  - key: '(?s)[^}\]]*$'
+    value: ""
+  - key: "'([^']*?)'"
+    value: "_DQUOTE_${1}_DQUOTE_"
+  - key: '\\"'
+    value: "__TEMP_QUOTE__"
+  - key: "\'"
+    value: "'"
+  - key: "_DQUOTE_"
+    value: '"'
+  - key: "__TEMP_QUOTE__"
+    value: '"'
+  # Drop the scratchpad content from responses
+  - key: "(?s)<scratchpad>.*</scratchpad>"
+    value: ""
+
+template:
+  chat: |
+    {{.Input -}}
+    <|im_start|>assistant
+  chat_message: |
+    <|im_start|>{{if eq .RoleName "assistant"}}assistant{{else if eq .RoleName "system"}}system{{else if eq .RoleName "tool"}}tool{{else if eq .RoleName "user"}}user{{end}}
+    {{- if .FunctionCall }}
+    <tool_call>
+    {{- else if eq .RoleName "tool" }}
+    <tool_response>
+    {{- end }}
+    {{- if .Content}}
+    {{.Content }}
+    {{- end }}
+    {{- if .FunctionCall}}
+    {{toJson .FunctionCall}}
+    {{- end }}
+    {{- if .FunctionCall }}
+    </tool_call>
+    {{- else if eq .RoleName "tool" }}
+    </tool_response>
+    {{- end }}<|im_end|>
+  completion: |
+    {{.Input}}
+  function: |-
+    <|im_start|>system
+    You are a function calling AI model.
+    Here are the available tools:
+    <tools>
+    {{range .Functions}}
+    {'type': 'function', 'function': {'name': '{{.Name}}', 'description': '{{.Description}}', 'parameters': {{toJson .Parameters}} }}
+    {{end}}
+    </tools>
+    You should call the tools provided to you sequentially
+    Please use <scratchpad> XML tags to record your reasoning and planning before you call the functions as follows:
+    <scratchpad>
+    {step-by-step reasoning and plan in bullet points}
+    </scratchpad>
+    For each function call return a json object with function name and arguments within <tool_call> XML tags as follows:
+    <tool_call>
+    {"arguments": <args-dict>, "name": <function-name>}
+    </tool_call><|im_end|>
+    {{.Input -}}
+    <|im_start|>assistant

aio/intel/vision.yaml

@@ -0,0 +1,35 @@
+backend: llama-cpp
+context_size: 4096
+mmap: false
+f16: false
+name: gpt-4-vision-preview
+
+roles:
+  user: "USER:"
+  assistant: "ASSISTANT:"
+  system: "SYSTEM:"
+
+mmproj: llava-v1.6-7b-mmproj-f16.gguf
+parameters:
+  model: llava-v1.6-mistral-7b.Q5_K_M.gguf
+  temperature: 0.2
+  top_k: 40
+  top_p: 0.95
+  seed: -1
+
+template:
+  chat: |
+    A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.
+    {{.Input}}
+    ASSISTANT:
+
+download_files:
+- filename: llava-v1.6-mistral-7b.Q5_K_M.gguf
+  uri: huggingface://cjpais/llava-1.6-mistral-7b-gguf/llava-v1.6-mistral-7b.Q5_K_M.gguf
+- filename: llava-v1.6-7b-mmproj-f16.gguf
+  uri: huggingface://cjpais/llava-1.6-mistral-7b-gguf/mmproj-model-f16.gguf
+
+usage: |
+    curl http://localhost:8080/v1/chat/completions -H "Content-Type: application/json" -d '{
+        "model": "gpt-4-vision-preview",
+        "messages": [{"role": "user", "content": [{"type":"text", "text": "What is in the image?"}, {"type": "image_url", "image_url": {"url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg" }}], "temperature": 0.9}]}'

backend/backend.proto

@@ -18,6 +18,72 @@ service Backend {
   rpc TTS(TTSRequest) returns (Result) {}
   rpc TokenizeString(PredictOptions) returns (TokenizationResponse) {}
   rpc Status(HealthMessage) returns (StatusResponse) {}
+
+  rpc StoresSet(StoresSetOptions) returns (Result) {}
+  rpc StoresDelete(StoresDeleteOptions) returns (Result) {}
+  rpc StoresGet(StoresGetOptions) returns (StoresGetResult) {}
+  rpc StoresFind(StoresFindOptions) returns (StoresFindResult) {}
+
+  rpc Rerank(RerankRequest) returns (RerankResult) {}
+}
+
+message RerankRequest {
+  string query = 1;
+  repeated string documents = 2;
+  int32 top_n = 3;
+}
+
+message RerankResult {
+  Usage usage = 1;
+  repeated DocumentResult results = 2;
+}
+
+message Usage {
+  int32 total_tokens = 1;
+  int32 prompt_tokens = 2;
+}
+
+message DocumentResult {
+  int32 index = 1;
+  string text = 2;
+  float relevance_score = 3;
+}
+
+message StoresKey {
+  repeated float Floats = 1;
+}
+
+message StoresValue {
+  bytes Bytes = 1;
+}
+
+message StoresSetOptions {
+  repeated StoresKey Keys = 1;
+  repeated StoresValue Values = 2;
+}
+
+message StoresDeleteOptions {
+  repeated StoresKey Keys = 1;
+}
+
+message StoresGetOptions {
+  repeated StoresKey Keys = 1;
+}
+
+message StoresGetResult {
+  repeated StoresKey Keys = 1;
+  repeated StoresValue Values = 2;
+}
+
+message StoresFindOptions {
+  StoresKey Key = 1;
+  int32 TopK = 2;
+}
+
+message StoresFindResult {
+  repeated StoresKey Keys = 1;
+  repeated StoresValue Values = 2;
+  repeated float Similarities = 3;
 }
 
 message HealthMessage {}
@@ -65,11 +131,15 @@ message PredictOptions {
   string NegativePrompt = 40;
   int32 NDraft = 41;
   repeated string Images = 42;
+  bool UseTokenizerTemplate = 43;
+  repeated Message Messages = 44;
 }
 
 // The response message containing the result
 message Reply {
   bytes message = 1;
+  int32 tokens = 2;
+  int32 prompt_tokens = 3;
 }
 
 message ModelOptions {
@@ -121,7 +191,7 @@ message ModelOptions {
 
   bool NoMulMatQ = 37;
   string DraftModel = 39;
-  
+
   string AudioPath = 38;
 
   // vllm
@@ -131,6 +201,7 @@ message ModelOptions {
   bool   EnforceEager = 52;
   int32  SwapSpace = 53;
   int32  MaxModelLen = 54;
+  int32  TensorParallelSize = 55;
 
   string MMProj = 41;
 
@@ -141,6 +212,9 @@ message ModelOptions {
   float YarnBetaSlow = 47;
 
   string Type = 49;
+
+  bool FlashAttention = 56;
+  bool NoKVOffload = 57;
 }
 
 message Result {
@@ -192,6 +266,7 @@ message TTSRequest {
   string model = 2;
   string dst = 3;
   string voice = 4;
+  optional string language = 5;
 }
 
 message TokenizationResponse {
@@ -213,4 +288,9 @@ message StatusResponse {
   }
   State state = 1;
   MemoryUsageData memory = 2;
+}
+
+message Message {
+  string role = 1;
+  string content = 2;
 }

backend/backend_grpc.pb.go

@@ -1,457 +0,0 @@
-// Code generated by protoc-gen-go-grpc. DO NOT EDIT.
-// versions:
-// - protoc-gen-go-grpc v1.2.0
-// - protoc             v4.23.4
-// source: backend/backend.proto
-
-package proto
-
-import (
-	context "context"
-	grpc "google.golang.org/grpc"
-	codes "google.golang.org/grpc/codes"
-	status "google.golang.org/grpc/status"
-)
-
-// This is a compile-time assertion to ensure that this generated file
-// is compatible with the grpc package it is being compiled against.
-// Requires gRPC-Go v1.32.0 or later.
-const _ = grpc.SupportPackageIsVersion7
-
-// BackendClient is the client API for Backend service.
-//
-// For semantics around ctx use and closing/ending streaming RPCs, please refer to https://pkg.go.dev/google.golang.org/grpc/?tab=doc#ClientConn.NewStream.
-type BackendClient interface {
-	Health(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*Reply, error)
-	Predict(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*Reply, error)
-	LoadModel(ctx context.Context, in *ModelOptions, opts ...grpc.CallOption) (*Result, error)
-	PredictStream(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (Backend_PredictStreamClient, error)
-	Embedding(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*EmbeddingResult, error)
-	GenerateImage(ctx context.Context, in *GenerateImageRequest, opts ...grpc.CallOption) (*Result, error)
-	AudioTranscription(ctx context.Context, in *TranscriptRequest, opts ...grpc.CallOption) (*TranscriptResult, error)
-	TTS(ctx context.Context, in *TTSRequest, opts ...grpc.CallOption) (*Result, error)
-	TokenizeString(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*TokenizationResponse, error)
-	Status(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*StatusResponse, error)
-}
-
-type backendClient struct {
-	cc grpc.ClientConnInterface
-}
-
-func NewBackendClient(cc grpc.ClientConnInterface) BackendClient {
-	return &backendClient{cc}
-}
-
-func (c *backendClient) Health(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*Reply, error) {
-	out := new(Reply)
-	err := c.cc.Invoke(ctx, "/backend.Backend/Health", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) Predict(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*Reply, error) {
-	out := new(Reply)
-	err := c.cc.Invoke(ctx, "/backend.Backend/Predict", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) LoadModel(ctx context.Context, in *ModelOptions, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, "/backend.Backend/LoadModel", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) PredictStream(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (Backend_PredictStreamClient, error) {
-	stream, err := c.cc.NewStream(ctx, &Backend_ServiceDesc.Streams[0], "/backend.Backend/PredictStream", opts...)
-	if err != nil {
-		return nil, err
-	}
-	x := &backendPredictStreamClient{stream}
-	if err := x.ClientStream.SendMsg(in); err != nil {
-		return nil, err
-	}
-	if err := x.ClientStream.CloseSend(); err != nil {
-		return nil, err
-	}
-	return x, nil
-}
-
-type Backend_PredictStreamClient interface {
-	Recv() (*Reply, error)
-	grpc.ClientStream
-}
-
-type backendPredictStreamClient struct {
-	grpc.ClientStream
-}
-
-func (x *backendPredictStreamClient) Recv() (*Reply, error) {
-	m := new(Reply)
-	if err := x.ClientStream.RecvMsg(m); err != nil {
-		return nil, err
-	}
-	return m, nil
-}
-
-func (c *backendClient) Embedding(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*EmbeddingResult, error) {
-	out := new(EmbeddingResult)
-	err := c.cc.Invoke(ctx, "/backend.Backend/Embedding", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) GenerateImage(ctx context.Context, in *GenerateImageRequest, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, "/backend.Backend/GenerateImage", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) AudioTranscription(ctx context.Context, in *TranscriptRequest, opts ...grpc.CallOption) (*TranscriptResult, error) {
-	out := new(TranscriptResult)
-	err := c.cc.Invoke(ctx, "/backend.Backend/AudioTranscription", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) TTS(ctx context.Context, in *TTSRequest, opts ...grpc.CallOption) (*Result, error) {
-	out := new(Result)
-	err := c.cc.Invoke(ctx, "/backend.Backend/TTS", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) TokenizeString(ctx context.Context, in *PredictOptions, opts ...grpc.CallOption) (*TokenizationResponse, error) {
-	out := new(TokenizationResponse)
-	err := c.cc.Invoke(ctx, "/backend.Backend/TokenizeString", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-func (c *backendClient) Status(ctx context.Context, in *HealthMessage, opts ...grpc.CallOption) (*StatusResponse, error) {
-	out := new(StatusResponse)
-	err := c.cc.Invoke(ctx, "/backend.Backend/Status", in, out, opts...)
-	if err != nil {
-		return nil, err
-	}
-	return out, nil
-}
-
-// BackendServer is the server API for Backend service.
-// All implementations must embed UnimplementedBackendServer
-// for forward compatibility
-type BackendServer interface {
-	Health(context.Context, *HealthMessage) (*Reply, error)
-	Predict(context.Context, *PredictOptions) (*Reply, error)
-	LoadModel(context.Context, *ModelOptions) (*Result, error)
-	PredictStream(*PredictOptions, Backend_PredictStreamServer) error
-	Embedding(context.Context, *PredictOptions) (*EmbeddingResult, error)
-	GenerateImage(context.Context, *GenerateImageRequest) (*Result, error)
-	AudioTranscription(context.Context, *TranscriptRequest) (*TranscriptResult, error)
-	TTS(context.Context, *TTSRequest) (*Result, error)
-	TokenizeString(context.Context, *PredictOptions) (*TokenizationResponse, error)
-	Status(context.Context, *HealthMessage) (*StatusResponse, error)
-	mustEmbedUnimplementedBackendServer()
-}
-
-// UnimplementedBackendServer must be embedded to have forward compatible implementations.
-type UnimplementedBackendServer struct {
-}
-
-func (UnimplementedBackendServer) Health(context.Context, *HealthMessage) (*Reply, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Health not implemented")
-}
-func (UnimplementedBackendServer) Predict(context.Context, *PredictOptions) (*Reply, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Predict not implemented")
-}
-func (UnimplementedBackendServer) LoadModel(context.Context, *ModelOptions) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method LoadModel not implemented")
-}
-func (UnimplementedBackendServer) PredictStream(*PredictOptions, Backend_PredictStreamServer) error {
-	return status.Errorf(codes.Unimplemented, "method PredictStream not implemented")
-}
-func (UnimplementedBackendServer) Embedding(context.Context, *PredictOptions) (*EmbeddingResult, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Embedding not implemented")
-}
-func (UnimplementedBackendServer) GenerateImage(context.Context, *GenerateImageRequest) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method GenerateImage not implemented")
-}
-func (UnimplementedBackendServer) AudioTranscription(context.Context, *TranscriptRequest) (*TranscriptResult, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method AudioTranscription not implemented")
-}
-func (UnimplementedBackendServer) TTS(context.Context, *TTSRequest) (*Result, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method TTS not implemented")
-}
-func (UnimplementedBackendServer) TokenizeString(context.Context, *PredictOptions) (*TokenizationResponse, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method TokenizeString not implemented")
-}
-func (UnimplementedBackendServer) Status(context.Context, *HealthMessage) (*StatusResponse, error) {
-	return nil, status.Errorf(codes.Unimplemented, "method Status not implemented")
-}
-func (UnimplementedBackendServer) mustEmbedUnimplementedBackendServer() {}
-
-// UnsafeBackendServer may be embedded to opt out of forward compatibility for this service.
-// Use of this interface is not recommended, as added methods to BackendServer will
-// result in compilation errors.
-type UnsafeBackendServer interface {
-	mustEmbedUnimplementedBackendServer()
-}
-
-func RegisterBackendServer(s grpc.ServiceRegistrar, srv BackendServer) {
-	s.RegisterService(&Backend_ServiceDesc, srv)
-}
-
-func _Backend_Health_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(HealthMessage)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Health(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/Health",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Health(ctx, req.(*HealthMessage))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_Predict_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(PredictOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Predict(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/Predict",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Predict(ctx, req.(*PredictOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_LoadModel_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(ModelOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).LoadModel(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/LoadModel",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).LoadModel(ctx, req.(*ModelOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_PredictStream_Handler(srv interface{}, stream grpc.ServerStream) error {
-	m := new(PredictOptions)
-	if err := stream.RecvMsg(m); err != nil {
-		return err
-	}
-	return srv.(BackendServer).PredictStream(m, &backendPredictStreamServer{stream})
-}
-
-type Backend_PredictStreamServer interface {
-	Send(*Reply) error
-	grpc.ServerStream
-}
-
-type backendPredictStreamServer struct {
-	grpc.ServerStream
-}
-
-func (x *backendPredictStreamServer) Send(m *Reply) error {
-	return x.ServerStream.SendMsg(m)
-}
-
-func _Backend_Embedding_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(PredictOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Embedding(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/Embedding",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Embedding(ctx, req.(*PredictOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_GenerateImage_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(GenerateImageRequest)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).GenerateImage(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/GenerateImage",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).GenerateImage(ctx, req.(*GenerateImageRequest))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_AudioTranscription_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(TranscriptRequest)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).AudioTranscription(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/AudioTranscription",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).AudioTranscription(ctx, req.(*TranscriptRequest))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_TTS_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(TTSRequest)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).TTS(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/TTS",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).TTS(ctx, req.(*TTSRequest))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_TokenizeString_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(PredictOptions)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).TokenizeString(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/TokenizeString",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).TokenizeString(ctx, req.(*PredictOptions))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-func _Backend_Status_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
-	in := new(HealthMessage)
-	if err := dec(in); err != nil {
-		return nil, err
-	}
-	if interceptor == nil {
-		return srv.(BackendServer).Status(ctx, in)
-	}
-	info := &grpc.UnaryServerInfo{
-		Server:     srv,
-		FullMethod: "/backend.Backend/Status",
-	}
-	handler := func(ctx context.Context, req interface{}) (interface{}, error) {
-		return srv.(BackendServer).Status(ctx, req.(*HealthMessage))
-	}
-	return interceptor(ctx, in, info, handler)
-}
-
-// Backend_ServiceDesc is the grpc.ServiceDesc for Backend service.
-// It's only intended for direct use with grpc.RegisterService,
-// and not to be introspected or modified (even as a copy)
-var Backend_ServiceDesc = grpc.ServiceDesc{
-	ServiceName: "backend.Backend",
-	HandlerType: (*BackendServer)(nil),
-	Methods: []grpc.MethodDesc{
-		{
-			MethodName: "Health",
-			Handler:    _Backend_Health_Handler,
-		},
-		{
-			MethodName: "Predict",
-			Handler:    _Backend_Predict_Handler,
-		},
-		{
-			MethodName: "LoadModel",
-			Handler:    _Backend_LoadModel_Handler,
-		},
-		{
-			MethodName: "Embedding",
-			Handler:    _Backend_Embedding_Handler,
-		},
-		{
-			MethodName: "GenerateImage",
-			Handler:    _Backend_GenerateImage_Handler,
-		},
-		{
-			MethodName: "AudioTranscription",
-			Handler:    _Backend_AudioTranscription_Handler,
-		},
-		{
-			MethodName: "TTS",
-			Handler:    _Backend_TTS_Handler,
-		},
-		{
-			MethodName: "TokenizeString",
-			Handler:    _Backend_TokenizeString_Handler,
-		},
-		{
-			MethodName: "Status",
-			Handler:    _Backend_Status_Handler,
-		},
-	},
-	Streams: []grpc.StreamDesc{
-		{
-			StreamName:    "PredictStream",
-			Handler:       _Backend_PredictStream_Handler,
-			ServerStreams: true,
-		},
-	},
-	Metadata: "backend/backend.proto",
-}

backend/cpp/grpc/Makefile

@@ -5,7 +5,6 @@ SYSTEM ?= $(HOST_SYSTEM)
 TAG_LIB_GRPC?=v1.59.0
 GIT_REPO_LIB_GRPC?=https://github.com/grpc/grpc.git
 GIT_CLONE_DEPTH?=1
-NUM_BUILD_THREADS?=$(shell nproc --ignore=1)
 
 INSTALLED_PACKAGES=installed_packages
 GRPC_REPO=grpc_repo
@@ -48,11 +47,11 @@ $(INSTALLED_PACKAGES): grpc_build
 
 $(GRPC_REPO):
 	git clone --depth $(GIT_CLONE_DEPTH) -b $(TAG_LIB_GRPC) $(GIT_REPO_LIB_GRPC) $(GRPC_REPO)/grpc
-	cd $(GRPC_REPO)/grpc && git submodule update --init --recursive --depth $(GIT_CLONE_DEPTH)
+	cd $(GRPC_REPO)/grpc && git submodule update --jobs 2 --init --recursive --depth $(GIT_CLONE_DEPTH)
 
 $(GRPC_BUILD): $(GRPC_REPO)
 	mkdir -p $(GRPC_BUILD)
-	cd $(GRPC_BUILD) && cmake $(CMAKE_ARGS) ../$(GRPC_REPO)/grpc && cmake --build . -- -j ${NUM_BUILD_THREADS} && cmake --build . --target install -- -j ${NUM_BUILD_THREADS}
+	cd $(GRPC_BUILD) && cmake $(CMAKE_ARGS) ../$(GRPC_REPO)/grpc && cmake --build . && cmake --build . --target install
 
 build: $(INSTALLED_PACKAGES)
 

backend/cpp/llama/Makefile

@@ -18,6 +18,12 @@ else ifeq ($(BUILD_TYPE),clblas)
 # If it's hipblas we do have also to set CC=/opt/rocm/llvm/bin/clang CXX=/opt/rocm/llvm/bin/clang++ 
 else ifeq ($(BUILD_TYPE),hipblas)
 	CMAKE_ARGS+=-DLLAMA_HIPBLAS=ON
+# If it's OSX, DO NOT embed the metal library - -DLLAMA_METAL_EMBED_LIBRARY=ON requires further investigation
+# But if it's OSX without metal, disable it here
+else ifeq ($(OS),darwin)
+	ifneq ($(BUILD_TYPE),metal)
+		CMAKE_ARGS+=-DLLAMA_METAL=OFF
+	endif
 endif
 
 ifeq ($(BUILD_TYPE),sycl_f16)
@@ -35,37 +41,29 @@ llama.cpp:
 	fi
 	cd llama.cpp && git checkout -b build $(LLAMA_VERSION) && git submodule update --init --recursive --depth 1
 
-llama.cpp/examples/grpc-server:
+llama.cpp/examples/grpc-server: llama.cpp
 	mkdir -p llama.cpp/examples/grpc-server
-	cp -r $(abspath ./)/CMakeLists.txt llama.cpp/examples/grpc-server/
-	cp -r $(abspath ./)/grpc-server.cpp llama.cpp/examples/grpc-server/
-	cp -rfv $(abspath ./)/json.hpp llama.cpp/examples/grpc-server/
-	cp -rfv $(abspath ./)/utils.hpp llama.cpp/examples/grpc-server/
-	echo "add_subdirectory(grpc-server)" >> llama.cpp/examples/CMakeLists.txt
-## XXX: In some versions of CMake clip wasn't being built before llama.
-## This is an hack for now, but it should be fixed in the future.
-	cp -rfv llama.cpp/examples/llava/clip.h llama.cpp/examples/grpc-server/clip.h
-	cp -rfv llama.cpp/examples/llava/llava.cpp llama.cpp/examples/grpc-server/llava.cpp
-	echo '#include "llama.h"' > llama.cpp/examples/grpc-server/llava.h
-	cat llama.cpp/examples/llava/llava.h >> llama.cpp/examples/grpc-server/llava.h
-	cp -rfv llama.cpp/examples/llava/clip.cpp llama.cpp/examples/grpc-server/clip.cpp
+	bash prepare.sh
 
 rebuild:
-	cp -rfv $(abspath ./)/CMakeLists.txt llama.cpp/examples/grpc-server/
-	cp -rfv $(abspath ./)/grpc-server.cpp llama.cpp/examples/grpc-server/
-	cp -rfv $(abspath ./)/json.hpp llama.cpp/examples/grpc-server/
+	bash prepare.sh
 	rm -rf grpc-server
 	$(MAKE) grpc-server
 
-clean:
-	rm -rf llama.cpp
+purge:
+	rm -rf llama.cpp/build
+	rm -rf llama.cpp/examples/grpc-server
 	rm -rf grpc-server
 
+clean: purge
+	rm -rf llama.cpp
+
 grpc-server: llama.cpp llama.cpp/examples/grpc-server
+	@echo "Building grpc-server with $(BUILD_TYPE) build type and $(CMAKE_ARGS)"
 ifneq (,$(findstring sycl,$(BUILD_TYPE)))
 	bash -c "source $(ONEAPI_VARS); \
-	cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && cmake --build . --config Release"	
+	cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && $(MAKE)"
 else
-	cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && cmake --build . --config Release
+	cd llama.cpp && mkdir -p build && cd build && cmake .. $(CMAKE_ARGS) && $(MAKE)
 endif
 	cp llama.cpp/build/bin/grpc-server .

backend/cpp/llama/grpc-server.cpp

@@ -791,7 +791,7 @@ struct llama_server_context
                     sampler_names.emplace_back(sampler_name);
                 }
             }
-            slot->sparams.samplers_sequence = sampler_types_from_names(sampler_names, false);
+            slot->sparams.samplers_sequence = llama_sampling_types_from_names(sampler_names, false);
         }
         else
         {
@@ -1084,7 +1084,7 @@ struct llama_server_context
             slot.has_next_token = false;
         }
 
-        if (!slot.cache_tokens.empty() && result.tok == llama_token_eos(model))
+        if (result.tok == llama_token_eos(model))
         {
             slot.stopped_eos = true;
             slot.has_next_token = false;
@@ -1146,7 +1146,7 @@ struct llama_server_context
         std::vector<std::string> samplers_sequence;
         for (const auto &sampler_type : slot.sparams.samplers_sequence)
         {
-            samplers_sequence.emplace_back(sampler_type_to_name_string(sampler_type));
+            samplers_sequence.emplace_back(llama_sampling_type_to_str(sampler_type));
         }
 
         return json {
@@ -2217,6 +2217,12 @@ static void params_parse(const backend::ModelOptions* request,
     } else {
         params.n_parallel = 1;
     }
+
+    const char *llama_grpc_servers = std::getenv("LLAMACPP_GRPC_SERVERS");
+    if (llama_grpc_servers != NULL) {
+        params.rpc_servers = std::string(llama_grpc_servers);
+    }
+    
     // TODO: Add yarn
 
     if (!request->tensorsplit().empty()) {
@@ -2254,6 +2260,9 @@ static void params_parse(const backend::ModelOptions* request,
     }
     params.use_mlock = request->mlock();
     params.use_mmap = request->mmap();
+    params.flash_attn = request->flashattention();
+    params.no_kv_offload = request->nokvoffload();
+
     params.embedding = request->embeddings();
 
     if (request->ropescaling() == "none")   { params.rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_NONE; }
@@ -2332,6 +2341,10 @@ public:
                 std::string completion_text = result.result_json.value("content", "");
 
                 reply.set_message(completion_text);
+                int32_t tokens_predicted = result.result_json.value("tokens_predicted", 0);
+                reply.set_tokens(tokens_predicted);
+                int32_t tokens_evaluated = result.result_json.value("tokens_evaluated", 0);
+                reply.set_prompt_tokens(tokens_evaluated);
 
                 // Send the reply
                 writer->Write(reply);
@@ -2357,6 +2370,10 @@ public:
         task_result result = llama.queue_results.recv(task_id);
         if (!result.error && result.stop) {
             completion_text = result.result_json.value("content", "");
+            int32_t tokens_predicted = result.result_json.value("tokens_predicted", 0);
+            int32_t tokens_evaluated = result.result_json.value("tokens_evaluated", 0);
+            reply->set_prompt_tokens(tokens_evaluated);
+            reply->set_tokens(tokens_predicted);
             reply->set_message(completion_text);
         }
         else

backend/cpp/llama/prepare.sh

@@ -0,0 +1,20 @@
+#!/bin/bash
+
+cp -r CMakeLists.txt llama.cpp/examples/grpc-server/
+cp -r grpc-server.cpp llama.cpp/examples/grpc-server/
+cp -rfv json.hpp llama.cpp/examples/grpc-server/
+cp -rfv utils.hpp llama.cpp/examples/grpc-server/
+    
+if grep -q "grpc-server" llama.cpp/examples/CMakeLists.txt; then
+    echo "grpc-server already added"
+else
+    echo "add_subdirectory(grpc-server)" >> llama.cpp/examples/CMakeLists.txt
+fi
+
+## XXX: In some versions of CMake clip wasn't being built before llama.
+## This is an hack for now, but it should be fixed in the future.
+cp -rfv llama.cpp/examples/llava/clip.h llama.cpp/examples/grpc-server/clip.h
+cp -rfv llama.cpp/examples/llava/llava.cpp llama.cpp/examples/grpc-server/llava.cpp
+echo '#include "llama.h"' > llama.cpp/examples/grpc-server/llava.h
+cat llama.cpp/examples/llava/llava.h >> llama.cpp/examples/grpc-server/llava.h
+cp -rfv llama.cpp/examples/llava/clip.cpp llama.cpp/examples/grpc-server/clip.cpp

backend/go/llm/langchain/langchain.go

@@ -4,6 +4,7 @@ package main
 // It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)
 import (
 	"fmt"
+	"os"
 
 	"github.com/go-skynet/LocalAI/pkg/grpc/base"
 	pb "github.com/go-skynet/LocalAI/pkg/grpc/proto"
@@ -18,9 +19,14 @@ type LLM struct {
 }
 
 func (llm *LLM) Load(opts *pb.ModelOptions) error {
-	llm.langchain, _ = langchain.NewHuggingFace(opts.Model)
+	var err error
+	hfToken := os.Getenv("HUGGINGFACEHUB_API_TOKEN")
+	if hfToken == "" {
+		return fmt.Errorf("no huggingface token provided")
+	}
+	llm.langchain, err = langchain.NewHuggingFace(opts.Model, hfToken)
 	llm.model = opts.Model
-	return nil
+	return err
 }
 
 func (llm *LLM) Predict(opts *pb.PredictOptions) (string, error) {

backend/go/stores/debug.go

@@ -0,0 +1,14 @@
+//go:build debug
+// +build debug
+
+package main
+
+import (
+	"github.com/rs/zerolog/log"
+)
+
+func assert(cond bool, msg string) {
+	if !cond {
+		log.Fatal().Stack().Msg(msg)
+	}
+}

backend/go/stores/main.go

@@ -0,0 +1,26 @@
+package main
+
+// Note: this is started internally by LocalAI and a server is allocated for each store
+
+import (
+	"flag"
+	"os"
+
+	grpc "github.com/go-skynet/LocalAI/pkg/grpc"
+	"github.com/rs/zerolog"
+	"github.com/rs/zerolog/log"
+)
+
+var (
+	addr = flag.String("addr", "localhost:50051", "the address to connect to")
+)
+
+func main() {
+	log.Logger = log.Output(zerolog.ConsoleWriter{Out: os.Stderr})
+
+	flag.Parse()
+
+	if err := grpc.StartServer(*addr, NewStore()); err != nil {
+		panic(err)
+	}
+}

backend/go/stores/production.go

@@ -0,0 +1,7 @@
+//go:build !debug
+// +build !debug
+
+package main
+
+func assert(cond bool, msg string) {
+}

backend/go/stores/store.go

@@ -0,0 +1,507 @@
+package main
+
+// This is a wrapper to statisfy the GRPC service interface
+// It is meant to be used by the main executable that is the server for the specific backend type (falcon, gpt3, etc)
+import (
+	"container/heap"
+	"fmt"
+	"math"
+	"slices"
+
+	"github.com/go-skynet/LocalAI/pkg/grpc/base"
+	pb "github.com/go-skynet/LocalAI/pkg/grpc/proto"
+
+	"github.com/rs/zerolog/log"
+)
+
+type Store struct {
+	base.SingleThread
+
+	// The sorted keys
+	keys [][]float32
+	// The sorted values
+	values [][]byte
+
+	// If for every K it holds that ||k||^2 = 1, then we can use the normalized distance functions
+	// TODO: Should we normalize incoming keys if they are not instead?
+	keysAreNormalized bool
+	// The first key decides the length of the keys
+	keyLen int
+}
+
+// TODO: Only used for sorting using Go's builtin implementation. The interfaces are columnar because
+// that's theoretically best for memory layout and cache locality, but this isn't optimized yet.
+type Pair struct {
+	Key   []float32
+	Value []byte
+}
+
+func NewStore() *Store {
+	return &Store{
+		keys:              make([][]float32, 0),
+		values:            make([][]byte, 0),
+		keysAreNormalized: true,
+		keyLen:            -1,
+	}
+}
+
+func compareSlices(k1, k2 []float32) int {
+	assert(len(k1) == len(k2), fmt.Sprintf("compareSlices: len(k1) = %d, len(k2) = %d", len(k1), len(k2)))
+
+	return slices.Compare(k1, k2)
+}
+
+func hasKey(unsortedSlice [][]float32, target []float32) bool {
+	return slices.ContainsFunc(unsortedSlice, func(k []float32) bool {
+		return compareSlices(k, target) == 0
+	})
+}
+
+func findInSortedSlice(sortedSlice [][]float32, target []float32) (int, bool) {
+	return slices.BinarySearchFunc(sortedSlice, target, func(k, t []float32) int {
+		return compareSlices(k, t)
+	})
+}
+
+func isSortedPairs(kvs []Pair) bool {
+	for i := 1; i < len(kvs); i++ {
+		if compareSlices(kvs[i-1].Key, kvs[i].Key) > 0 {
+			return false
+		}
+	}
+
+	return true
+}
+
+func isSortedKeys(keys [][]float32) bool {
+	for i := 1; i < len(keys); i++ {
+		if compareSlices(keys[i-1], keys[i]) > 0 {
+			return false
+		}
+	}
+
+	return true
+}
+
+func sortIntoKeySlicese(keys []*pb.StoresKey) [][]float32 {
+	ks := make([][]float32, len(keys))
+
+	for i, k := range keys {
+		ks[i] = k.Floats
+	}
+
+	slices.SortFunc(ks, compareSlices)
+
+	assert(len(ks) == len(keys), fmt.Sprintf("len(ks) = %d, len(keys) = %d", len(ks), len(keys)))
+	assert(isSortedKeys(ks), "keys are not sorted")
+
+	return ks
+}
+
+func (s *Store) Load(opts *pb.ModelOptions) error {
+	return nil
+}
+
+// Sort the incoming kvs and merge them with the existing sorted kvs
+func (s *Store) StoresSet(opts *pb.StoresSetOptions) error {
+	if len(opts.Keys) == 0 {
+		return fmt.Errorf("no keys to add")
+	}
+
+	if len(opts.Keys) != len(opts.Values) {
+		return fmt.Errorf("len(keys) = %d, len(values) = %d", len(opts.Keys), len(opts.Values))
+	}
+
+	if s.keyLen == -1 {
+		s.keyLen = len(opts.Keys[0].Floats)
+	} else {
+		if len(opts.Keys[0].Floats) != s.keyLen {
+			return fmt.Errorf("Try to add key with length %d when existing length is %d", len(opts.Keys[0].Floats), s.keyLen)
+		}
+	}
+
+	kvs := make([]Pair, len(opts.Keys))
+
+	for i, k := range opts.Keys {
+		if s.keysAreNormalized && !isNormalized(k.Floats) {
+			s.keysAreNormalized = false
+			var sample []float32
+			if len(s.keys) > 5 {
+				sample = k.Floats[:5]
+			} else {
+				sample = k.Floats
+			}
+			log.Debug().Msgf("Key is not normalized: %v", sample)
+		}
+
+		kvs[i] = Pair{
+			Key:   k.Floats,
+			Value: opts.Values[i].Bytes,
+		}
+	}
+
+	slices.SortFunc(kvs, func(a, b Pair) int {
+		return compareSlices(a.Key, b.Key)
+	})
+
+	assert(len(kvs) == len(opts.Keys), fmt.Sprintf("len(kvs) = %d, len(opts.Keys) = %d", len(kvs), len(opts.Keys)))
+	assert(isSortedPairs(kvs), "keys are not sorted")
+
+	l := len(kvs) + len(s.keys)
+	merge_ks := make([][]float32, 0, l)
+	merge_vs := make([][]byte, 0, l)
+
+	i, j := 0, 0
+	for {
+		if i+j >= l {
+			break
+		}
+
+		if i >= len(kvs) {
+			merge_ks = append(merge_ks, s.keys[j])
+			merge_vs = append(merge_vs, s.values[j])
+			j++
+			continue
+		}
+
+		if j >= len(s.keys) {
+			merge_ks = append(merge_ks, kvs[i].Key)
+			merge_vs = append(merge_vs, kvs[i].Value)
+			i++
+			continue
+		}
+
+		c := compareSlices(kvs[i].Key, s.keys[j])
+		if c < 0 {
+			merge_ks = append(merge_ks, kvs[i].Key)
+			merge_vs = append(merge_vs, kvs[i].Value)
+			i++
+		} else if c > 0 {
+			merge_ks = append(merge_ks, s.keys[j])
+			merge_vs = append(merge_vs, s.values[j])
+			j++
+		} else {
+			merge_ks = append(merge_ks, kvs[i].Key)
+			merge_vs = append(merge_vs, kvs[i].Value)
+			i++
+			j++
+		}
+	}
+
+	assert(len(merge_ks) == l, fmt.Sprintf("len(merge_ks) = %d, l = %d", len(merge_ks), l))
+	assert(isSortedKeys(merge_ks), "merge keys are not sorted")
+
+	s.keys = merge_ks
+	s.values = merge_vs
+
+	return nil
+}
+
+func (s *Store) StoresDelete(opts *pb.StoresDeleteOptions) error {
+	if len(opts.Keys) == 0 {
+		return fmt.Errorf("no keys to delete")
+	}
+
+	if len(opts.Keys) == 0 {
+		return fmt.Errorf("no keys to add")
+	}
+
+	if s.keyLen == -1 {
+		s.keyLen = len(opts.Keys[0].Floats)
+	} else {
+		if len(opts.Keys[0].Floats) != s.keyLen {
+			return fmt.Errorf("Trying to delete key with length %d when existing length is %d", len(opts.Keys[0].Floats), s.keyLen)
+		}
+	}
+
+	ks := sortIntoKeySlicese(opts.Keys)
+
+	l := len(s.keys) - len(ks)
+	merge_ks := make([][]float32, 0, l)
+	merge_vs := make([][]byte, 0, l)
+
+	tail_ks := s.keys
+	tail_vs := s.values
+	for _, k := range ks {
+		j, found := findInSortedSlice(tail_ks, k)
+
+		if found {
+			merge_ks = append(merge_ks, tail_ks[:j]...)
+			merge_vs = append(merge_vs, tail_vs[:j]...)
+			tail_ks = tail_ks[j+1:]
+			tail_vs = tail_vs[j+1:]
+		} else {
+			assert(!hasKey(s.keys, k), fmt.Sprintf("Key exists, but was not found: t=%d, %v", len(tail_ks), k))
+		}
+
+		log.Debug().Msgf("Delete: found = %v, t = %d, j = %d, len(merge_ks) = %d, len(merge_vs) = %d", found, len(tail_ks), j, len(merge_ks), len(merge_vs))
+	}
+
+	merge_ks = append(merge_ks, tail_ks...)
+	merge_vs = append(merge_vs, tail_vs...)
+
+	assert(len(merge_ks) <= len(s.keys), fmt.Sprintf("len(merge_ks) = %d, len(s.keys) = %d", len(merge_ks), len(s.keys)))
+
+	s.keys = merge_ks
+	s.values = merge_vs
+
+	assert(len(s.keys) >= l, fmt.Sprintf("len(s.keys) = %d, l = %d", len(s.keys), l))
+	assert(isSortedKeys(s.keys), "keys are not sorted")
+	assert(func() bool {
+		for _, k := range ks {
+			if _, found := findInSortedSlice(s.keys, k); found {
+				return false
+			}
+		}
+		return true
+	}(), "Keys to delete still present")
+
+	if len(s.keys) != l {
+		log.Debug().Msgf("Delete: Some keys not found: len(s.keys) = %d, l = %d", len(s.keys), l)
+	}
+
+	return nil
+}
+
+func (s *Store) StoresGet(opts *pb.StoresGetOptions) (pb.StoresGetResult, error) {
+	pbKeys := make([]*pb.StoresKey, 0, len(opts.Keys))
+	pbValues := make([]*pb.StoresValue, 0, len(opts.Keys))
+	ks := sortIntoKeySlicese(opts.Keys)
+
+	if len(s.keys) == 0 {
+		log.Debug().Msgf("Get: No keys in store")
+	}
+
+	if s.keyLen == -1 {
+		s.keyLen = len(opts.Keys[0].Floats)
+	} else {
+		if len(opts.Keys[0].Floats) != s.keyLen {
+			return pb.StoresGetResult{}, fmt.Errorf("Try to get a key with length %d when existing length is %d", len(opts.Keys[0].Floats), s.keyLen)
+		}
+	}
+
+	tail_k := s.keys
+	tail_v := s.values
+	for i, k := range ks {
+		j, found := findInSortedSlice(tail_k, k)
+
+		if found {
+			pbKeys = append(pbKeys, &pb.StoresKey{
+				Floats: k,
+			})
+			pbValues = append(pbValues, &pb.StoresValue{
+				Bytes: tail_v[j],
+			})
+
+			tail_k = tail_k[j+1:]
+			tail_v = tail_v[j+1:]
+		} else {
+			assert(!hasKey(s.keys, k), fmt.Sprintf("Key exists, but was not found: i=%d, %v", i, k))
+		}
+	}
+
+	if len(pbKeys) != len(opts.Keys) {
+		log.Debug().Msgf("Get: Some keys not found: len(pbKeys) = %d, len(opts.Keys) = %d, len(s.Keys) = %d", len(pbKeys), len(opts.Keys), len(s.keys))
+	}
+
+	return pb.StoresGetResult{
+		Keys:   pbKeys,
+		Values: pbValues,
+	}, nil
+}
+
+func isNormalized(k []float32) bool {
+	var sum float32
+	for _, v := range k {
+		sum += v
+	}
+
+	return sum == 1.0
+}
+
+// TODO: This we could replace with handwritten SIMD code
+func normalizedCosineSimilarity(k1, k2 []float32) float32 {
+	assert(len(k1) == len(k2), fmt.Sprintf("normalizedCosineSimilarity: len(k1) = %d, len(k2) = %d", len(k1), len(k2)))
+
+	var dot float32
+	for i := 0; i < len(k1); i++ {
+		dot += k1[i] * k2[i]
+	}
+
+	assert(dot >= -1 && dot <= 1, fmt.Sprintf("dot = %f", dot))
+
+	// 2.0 * (1.0 - dot) would be the Euclidean distance
+	return dot
+}
+
+type PriorityItem struct {
+	Similarity float32
+	Key        []float32
+	Value      []byte
+}
+
+type PriorityQueue []*PriorityItem
+
+func (pq PriorityQueue) Len() int { return len(pq) }
+
+func (pq PriorityQueue) Less(i, j int) bool {
+	// Inverted because the most similar should be at the top
+	return pq[i].Similarity < pq[j].Similarity
+}
+
+func (pq PriorityQueue) Swap(i, j int) {
+	pq[i], pq[j] = pq[j], pq[i]
+}
+
+func (pq *PriorityQueue) Push(x any) {
+	item := x.(*PriorityItem)
+	*pq = append(*pq, item)
+}
+
+func (pq *PriorityQueue) Pop() any {
+	old := *pq
+	n := len(old)
+	item := old[n-1]
+	*pq = old[0 : n-1]
+	return item
+}
+
+func (s *Store) StoresFindNormalized(opts *pb.StoresFindOptions) (pb.StoresFindResult, error) {
+	tk := opts.Key.Floats
+	top_ks := make(PriorityQueue, 0, int(opts.TopK))
+	heap.Init(&top_ks)
+
+	for i, k := range s.keys {
+		sim := normalizedCosineSimilarity(tk, k)
+		heap.Push(&top_ks, &PriorityItem{
+			Similarity: sim,
+			Key:        k,
+			Value:      s.values[i],
+		})
+
+		if top_ks.Len() > int(opts.TopK) {
+			heap.Pop(&top_ks)
+		}
+	}
+
+	similarities := make([]float32, top_ks.Len())
+	pbKeys := make([]*pb.StoresKey, top_ks.Len())
+	pbValues := make([]*pb.StoresValue, top_ks.Len())
+
+	for i := top_ks.Len() - 1; i >= 0; i-- {
+		item := heap.Pop(&top_ks).(*PriorityItem)
+
+		similarities[i] = item.Similarity
+		pbKeys[i] = &pb.StoresKey{
+			Floats: item.Key,
+		}
+		pbValues[i] = &pb.StoresValue{
+			Bytes: item.Value,
+		}
+	}
+
+	return pb.StoresFindResult{
+		Keys:         pbKeys,
+		Values:       pbValues,
+		Similarities: similarities,
+	}, nil
+}
+
+func cosineSimilarity(k1, k2 []float32, mag1 float64) float32 {
+	assert(len(k1) == len(k2), fmt.Sprintf("cosineSimilarity: len(k1) = %d, len(k2) = %d", len(k1), len(k2)))
+
+	var dot, mag2 float64
+	for i := 0; i < len(k1); i++ {
+		dot += float64(k1[i] * k2[i])
+		mag2 += float64(k2[i] * k2[i])
+	}
+
+	sim := float32(dot / (mag1 * math.Sqrt(mag2)))
+
+	assert(sim >= -1 && sim <= 1, fmt.Sprintf("sim = %f", sim))
+
+	return sim
+}
+
+func (s *Store) StoresFindFallback(opts *pb.StoresFindOptions) (pb.StoresFindResult, error) {
+	tk := opts.Key.Floats
+	top_ks := make(PriorityQueue, 0, int(opts.TopK))
+	heap.Init(&top_ks)
+
+	var mag1 float64
+	for _, v := range tk {
+		mag1 += float64(v * v)
+	}
+	mag1 = math.Sqrt(mag1)
+
+	for i, k := range s.keys {
+		dist := cosineSimilarity(tk, k, mag1)
+		heap.Push(&top_ks, &PriorityItem{
+			Similarity: dist,
+			Key:        k,
+			Value:      s.values[i],
+		})
+
+		if top_ks.Len() > int(opts.TopK) {
+			heap.Pop(&top_ks)
+		}
+	}
+
+	similarities := make([]float32, top_ks.Len())
+	pbKeys := make([]*pb.StoresKey, top_ks.Len())
+	pbValues := make([]*pb.StoresValue, top_ks.Len())
+
+	for i := top_ks.Len() - 1; i >= 0; i-- {
+		item := heap.Pop(&top_ks).(*PriorityItem)
+
+		similarities[i] = item.Similarity
+		pbKeys[i] = &pb.StoresKey{
+			Floats: item.Key,
+		}
+		pbValues[i] = &pb.StoresValue{
+			Bytes: item.Value,
+		}
+	}
+
+	return pb.StoresFindResult{
+		Keys:         pbKeys,
+		Values:       pbValues,
+		Similarities: similarities,
+	}, nil
+}
+
+func (s *Store) StoresFind(opts *pb.StoresFindOptions) (pb.StoresFindResult, error) {
+	tk := opts.Key.Floats
+
+	if len(tk) != s.keyLen {
+		return pb.StoresFindResult{}, fmt.Errorf("Try to find key with length %d when existing length is %d", len(tk), s.keyLen)
+	}
+
+	if opts.TopK < 1 {
+		return pb.StoresFindResult{}, fmt.Errorf("opts.TopK = %d, must be >= 1", opts.TopK)
+	}
+
+	if s.keyLen == -1 {
+		s.keyLen = len(opts.Key.Floats)
+	} else {
+		if len(opts.Key.Floats) != s.keyLen {
+			return pb.StoresFindResult{}, fmt.Errorf("Try to add key with length %d when existing length is %d", len(opts.Key.Floats), s.keyLen)
+		}
+	}
+
+	if s.keysAreNormalized && isNormalized(tk) {
+		return s.StoresFindNormalized(opts)
+	} else {
+		if s.keysAreNormalized {
+			var sample []float32
+			if len(s.keys) > 5 {
+				sample = tk[:5]
+			} else {
+				sample = tk
+			}
+			log.Debug().Msgf("Trying to compare non-normalized key with normalized keys: %v", sample)
+		}
+
+		return s.StoresFindFallback(opts)
+	}
+}

backend/go/transcribe/transcript.go

@@ -11,8 +11,8 @@ import (
 	"github.com/go-skynet/LocalAI/core/schema"
 )
 
-func runCommand(command []string) (string, error) {
-	cmd := exec.Command(command[0], command[1:]...)
+func ffmpegCommand(args []string) (string, error) {
+	cmd := exec.Command("ffmpeg", args...) // Constrain this to ffmpeg to permit security scanner to see that the command is safe.
 	cmd.Env = os.Environ()
 	out, err := cmd.CombinedOutput()
 	return string(out), err
@@ -21,16 +21,16 @@ func runCommand(command []string) (string, error) {
 // AudioToWav converts audio to wav for transcribe.
 // TODO: use https://github.com/mccoyst/ogg?
 func audioToWav(src, dst string) error {
-    command := []string{"ffmpeg", "-i", src, "-format", "s16le", "-ar", "16000", "-ac", "1", "-acodec", "pcm_s16le", dst}
-	out, err := runCommand(command)
+	commandArgs := []string{"-i", src, "-format", "s16le", "-ar", "16000", "-ac", "1", "-acodec", "pcm_s16le", dst}
+	out, err := ffmpegCommand(commandArgs)
 	if err != nil {
 		return fmt.Errorf("error: %w out: %s", err, out)
 	}
 	return nil
 }
 
-func Transcript(model whisper.Model, audiopath, language string, threads uint) (schema.Result, error) {
-	res := schema.Result{}
+func Transcript(model whisper.Model, audiopath, language string, threads uint) (schema.TranscriptionResult, error) {
+	res := schema.TranscriptionResult{}
 
 	dir, err := os.MkdirTemp("", "whisper")
 	if err != nil {

backend/go/transcribe/whisper.go

@@ -21,6 +21,6 @@ func (sd *Whisper) Load(opts *pb.ModelOptions) error {
 	return err
 }
 
-func (sd *Whisper) AudioTranscription(opts *pb.TranscriptRequest) (schema.Result, error) {
+func (sd *Whisper) AudioTranscription(opts *pb.TranscriptRequest) (schema.TranscriptionResult, error) {
 	return Transcript(sd.whisper, opts.Dst, opts.Language, uint(opts.Threads))
 }

backend/python/autogptq/Makefile

@@ -1,4 +1,17 @@
 .PHONY: autogptq
-autogptq:
-	$(MAKE) -C ../common-env/transformers
+autogptq: protogen
+	bash install.sh
 
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
+
+.PHONY: clean
+clean: protogen-clean
+	rm -rf venv __pycache__

backend/python/autogptq/autogptq.yml

@@ -1,86 +0,0 @@
-name: autogptq
-channels:
-  - defaults
-dependencies:
-  - _libgcc_mutex=0.1=main
-  - _openmp_mutex=5.1=1_gnu
-  - bzip2=1.0.8=h7b6447c_0
-  - ca-certificates=2023.08.22=h06a4308_0
-  - ld_impl_linux-64=2.38=h1181459_1
-  - libffi=3.4.4=h6a678d5_0
-  - libgcc-ng=11.2.0=h1234567_1
-  - libgomp=11.2.0=h1234567_1
-  - libstdcxx-ng=11.2.0=h1234567_1
-  - libuuid=1.41.5=h5eee18b_0
-  - ncurses=6.4=h6a678d5_0
-  - openssl=3.0.11=h7f8727e_2
-  - pip=23.2.1=py311h06a4308_0
-  - python=3.11.5=h955ad1f_0
-  - readline=8.2=h5eee18b_0
-  - setuptools=68.0.0=py311h06a4308_0
-  - sqlite=3.41.2=h5eee18b_0
-  - tk=8.6.12=h1ccaba5_0
-  - wheel=0.41.2=py311h06a4308_0
-  - xz=5.4.2=h5eee18b_0
-  - zlib=1.2.13=h5eee18b_0
-  - pip:
-      - accelerate==0.23.0
-      - aiohttp==3.8.5
-      - aiosignal==1.3.1
-      - async-timeout==4.0.3
-      - attrs==23.1.0
-      - auto-gptq==0.4.2
-      - certifi==2023.7.22
-      - charset-normalizer==3.3.0
-      - datasets==2.14.5
-      - dill==0.3.7
-      - filelock==3.12.4
-      - frozenlist==1.4.0
-      - fsspec==2023.6.0
-      - grpcio==1.59.0
-      - huggingface-hub==0.16.4
-      - idna==3.4
-      - jinja2==3.1.2
-      - markupsafe==2.1.3
-      - mpmath==1.3.0
-      - multidict==6.0.4
-      - multiprocess==0.70.15
-      - networkx==3.1
-      - numpy==1.26.0
-      - nvidia-cublas-cu12==12.1.3.1
-      - nvidia-cuda-cupti-cu12==12.1.105
-      - nvidia-cuda-nvrtc-cu12==12.1.105
-      - nvidia-cuda-runtime-cu12==12.1.105
-      - nvidia-cudnn-cu12==8.9.2.26
-      - nvidia-cufft-cu12==11.0.2.54
-      - nvidia-curand-cu12==10.3.2.106
-      - nvidia-cusolver-cu12==11.4.5.107
-      - nvidia-cusparse-cu12==12.1.0.106
-      - nvidia-nccl-cu12==2.18.1
-      - nvidia-nvjitlink-cu12==12.2.140
-      - nvidia-nvtx-cu12==12.1.105
-      - packaging==23.2
-      - pandas==2.1.1
-      - peft==0.5.0
-      - protobuf==4.24.4
-      - psutil==5.9.5
-      - pyarrow==13.0.0
-      - python-dateutil==2.8.2
-      - pytz==2023.3.post1
-      - pyyaml==6.0.1
-      - regex==2023.10.3
-      - requests==2.31.0
-      - rouge==1.0.1
-      - safetensors>=0.3.3
-      - six==1.16.0
-      - sympy==1.12
-      - tokenizers==0.14.0
-      - torch==2.1.0
-      - tqdm==4.66.1
-      - transformers==4.34.0
-      - triton==2.1.0
-      - typing-extensions==4.8.0
-      - tzdata==2023.3
-      - urllib3==2.0.6
-      - xxhash==3.4.1
-      - yarl==1.9.2

backend/python/autogptq/backend.py

@@ -5,12 +5,14 @@ import signal
 import sys
 import os
 import time
+import base64
 
 import grpc
 import backend_pb2
 import backend_pb2_grpc
+
 from auto_gptq import AutoGPTQForCausalLM
-from transformers import AutoTokenizer
+from transformers import AutoTokenizer, AutoModelForCausalLM
 from transformers import TextGenerationPipeline
 
 _ONE_DAY_IN_SECONDS = 60 * 60 * 24
@@ -28,9 +30,18 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
             if request.Device != "":
                 device = request.Device
 
-            tokenizer = AutoTokenizer.from_pretrained(request.Model, use_fast=request.UseFastTokenizer)
+            # support loading local model files
+            model_path = os.path.join(os.environ.get('MODELS_PATH', './'), request.Model)
+            tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=True, trust_remote_code=request.TrustRemoteCode)
 
-            model = AutoGPTQForCausalLM.from_quantized(request.Model,
+            # support model `Qwen/Qwen-VL-Chat-Int4`
+            if "qwen-vl" in request.Model.lower():
+                self.model_name = "Qwen-VL-Chat"
+                model = AutoModelForCausalLM.from_pretrained(model_path, 
+                    trust_remote_code=request.TrustRemoteCode,
+                    device_map="auto").eval()
+            else:
+                model = AutoGPTQForCausalLM.from_quantized(model_path,
                     model_basename=request.ModelBaseName,
                     use_safetensors=True,
                     trust_remote_code=request.TrustRemoteCode,
@@ -55,6 +66,11 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
         if request.TopP != 0.0:
             top_p = request.TopP
 
+        
+        prompt_images = self.recompile_vl_prompt(request)
+        compiled_prompt = prompt_images[0]
+        print(f"Prompt: {compiled_prompt}", file=sys.stderr)
+
         # Implement Predict RPC
         pipeline = TextGenerationPipeline(
             model=self.model, 
@@ -64,10 +80,17 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
             top_p=top_p,
             repetition_penalty=penalty,
             )
-        t = pipeline(request.Prompt)[0]["generated_text"]
-        # Remove prompt from response if present
-        if request.Prompt in t:
-            t = t.replace(request.Prompt, "")
+        t = pipeline(compiled_prompt)[0]["generated_text"]
+        print(f"generated_text: {t}", file=sys.stderr)
+        
+        if compiled_prompt in t:
+            t = t.replace(compiled_prompt, "")
+        # house keeping. Remove the image files from /tmp folder
+        for img_path in prompt_images[1]:
+            try:
+                os.remove(img_path)
+            except Exception as e:
+                print(f"Error removing image file: {img_path}, {e}", file=sys.stderr)
 
         return backend_pb2.Result(message=bytes(t, encoding='utf-8'))
 
@@ -78,6 +101,24 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
         # Not implemented yet
         return self.Predict(request, context)
 
+    def recompile_vl_prompt(self, request):
+        prompt = request.Prompt
+        image_paths = []
+
+        if "qwen-vl" in self.model_name.lower():
+            # request.Images is an array which contains base64 encoded images. Iterate the request.Images array, decode and save each image to /tmp folder with a random filename.
+            # Then, save the image file paths to an array "image_paths".
+            # read "request.Prompt", replace "[img-%d]" with the image file paths in the order they appear in "image_paths". Save the new prompt to "prompt".
+            for i, img in enumerate(request.Images):
+                timestamp = str(int(time.time() * 1000))  # Generate timestamp
+                img_path = f"/tmp/vl-{timestamp}.jpg"  # Use timestamp in filename
+                with open(img_path, "wb") as f:
+                    f.write(base64.b64decode(img))
+                image_paths.append(img_path)
+                prompt = prompt.replace(f"[img-{i}]", "<img>" + img_path + "</img>,")
+        else:
+            prompt = request.Prompt
+        return (prompt, image_paths)
 
 def serve(address):
     server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS))

backend/python/autogptq/backend_pb2_grpc.py

@@ -1,363 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)

backend/python/autogptq/install.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+set -e
+
+source $(dirname $0)/../common/libbackend.sh
+
+# This is here because the Intel pip index is broken and returns 200 status codes for every package name, it just doesn't return any package links.
+# This makes uv think that the package exists in the Intel pip index, and by default it stops looking at other pip indexes once it finds a match.
+# We need uv to continue falling through to the pypi default index to find optimum[openvino] in the pypi index
+# the --upgrade actually allows us to *downgrade* torch to the version provided in the Intel pip index
+if [ "x${BUILD_PROFILE}" == "xintel" ]; then
+    EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
+fi
+
+installRequirements

backend/python/autogptq/requirements-hipblas.txt

@@ -0,0 +1,2 @@
+--extra-index-url https://download.pytorch.org/whl/rocm6.0
+torch

backend/python/autogptq/requirements-intel.txt

@@ -0,0 +1,5 @@
+--extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
+intel-extension-for-pytorch
+torch
+optimum[openvino]
+setuptools==69.5.1 # https://github.com/mudler/LocalAI/issues/2406

backend/python/autogptq/requirements.txt

@@ -0,0 +1,7 @@
+accelerate
+auto-gptq==0.7.1
+grpcio==1.64.0
+protobuf
+torch
+certifi
+transformers

backend/python/autogptq/run.sh

@@ -1,14 +1,4 @@
 #!/bin/bash
+source $(dirname $0)/../common/libbackend.sh
 
-##
-## A bash script wrapper that runs the autogptq server with conda
-
-export PATH=$PATH:/opt/conda/bin
-
-# Activate conda environment
-source activate transformers
-
-# get the directory where the bash script is located
-DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
-
-python $DIR/autogptq.py $@
+startBackend $@

backend/python/autogptq/test.sh

@@ -0,0 +1,6 @@
+#!/bin/bash
+set -e
+
+source $(dirname $0)/../common/libbackend.sh
+
+runUnittests

backend/python/bark/Makefile

@@ -1,15 +1,29 @@
 .PHONY: ttsbark
-ttsbark:
-	$(MAKE) -C ../common-env/transformers
+ttsbark: protogen
+	bash install.sh
 
 .PHONY: run
-run:
+run: protogen
 	@echo "Running bark..."
 	bash run.sh
 	@echo "bark run."
 
 .PHONY: test
-test:
+test: protogen
 	@echo "Testing bark..."
 	bash test.sh
 	@echo "bark tested."
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
+
+.PHONY: clean
+clean: protogen-clean
+	rm -rf venv __pycache__

backend/python/bark/backend_pb2_grpc.py

@@ -1,363 +0,0 @@
-# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
-"""Client and server classes corresponding to protobuf-defined services."""
-import grpc
-
-import backend_pb2 as backend__pb2
-
-
-class BackendStub(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def __init__(self, channel):
-        """Constructor.
-
-        Args:
-            channel: A grpc.Channel.
-        """
-        self.Health = channel.unary_unary(
-                '/backend.Backend/Health',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Predict = channel.unary_unary(
-                '/backend.Backend/Predict',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.LoadModel = channel.unary_unary(
-                '/backend.Backend/LoadModel',
-                request_serializer=backend__pb2.ModelOptions.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.PredictStream = channel.unary_stream(
-                '/backend.Backend/PredictStream',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.Reply.FromString,
-                )
-        self.Embedding = channel.unary_unary(
-                '/backend.Backend/Embedding',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.EmbeddingResult.FromString,
-                )
-        self.GenerateImage = channel.unary_unary(
-                '/backend.Backend/GenerateImage',
-                request_serializer=backend__pb2.GenerateImageRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.AudioTranscription = channel.unary_unary(
-                '/backend.Backend/AudioTranscription',
-                request_serializer=backend__pb2.TranscriptRequest.SerializeToString,
-                response_deserializer=backend__pb2.TranscriptResult.FromString,
-                )
-        self.TTS = channel.unary_unary(
-                '/backend.Backend/TTS',
-                request_serializer=backend__pb2.TTSRequest.SerializeToString,
-                response_deserializer=backend__pb2.Result.FromString,
-                )
-        self.TokenizeString = channel.unary_unary(
-                '/backend.Backend/TokenizeString',
-                request_serializer=backend__pb2.PredictOptions.SerializeToString,
-                response_deserializer=backend__pb2.TokenizationResponse.FromString,
-                )
-        self.Status = channel.unary_unary(
-                '/backend.Backend/Status',
-                request_serializer=backend__pb2.HealthMessage.SerializeToString,
-                response_deserializer=backend__pb2.StatusResponse.FromString,
-                )
-
-
-class BackendServicer(object):
-    """Missing associated documentation comment in .proto file."""
-
-    def Health(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Predict(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def LoadModel(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def PredictStream(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Embedding(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def GenerateImage(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def AudioTranscription(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TTS(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def TokenizeString(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def Status(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-
-def add_BackendServicer_to_server(servicer, server):
-    rpc_method_handlers = {
-            'Health': grpc.unary_unary_rpc_method_handler(
-                    servicer.Health,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Predict': grpc.unary_unary_rpc_method_handler(
-                    servicer.Predict,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'LoadModel': grpc.unary_unary_rpc_method_handler(
-                    servicer.LoadModel,
-                    request_deserializer=backend__pb2.ModelOptions.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'PredictStream': grpc.unary_stream_rpc_method_handler(
-                    servicer.PredictStream,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.Reply.SerializeToString,
-            ),
-            'Embedding': grpc.unary_unary_rpc_method_handler(
-                    servicer.Embedding,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.EmbeddingResult.SerializeToString,
-            ),
-            'GenerateImage': grpc.unary_unary_rpc_method_handler(
-                    servicer.GenerateImage,
-                    request_deserializer=backend__pb2.GenerateImageRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'AudioTranscription': grpc.unary_unary_rpc_method_handler(
-                    servicer.AudioTranscription,
-                    request_deserializer=backend__pb2.TranscriptRequest.FromString,
-                    response_serializer=backend__pb2.TranscriptResult.SerializeToString,
-            ),
-            'TTS': grpc.unary_unary_rpc_method_handler(
-                    servicer.TTS,
-                    request_deserializer=backend__pb2.TTSRequest.FromString,
-                    response_serializer=backend__pb2.Result.SerializeToString,
-            ),
-            'TokenizeString': grpc.unary_unary_rpc_method_handler(
-                    servicer.TokenizeString,
-                    request_deserializer=backend__pb2.PredictOptions.FromString,
-                    response_serializer=backend__pb2.TokenizationResponse.SerializeToString,
-            ),
-            'Status': grpc.unary_unary_rpc_method_handler(
-                    servicer.Status,
-                    request_deserializer=backend__pb2.HealthMessage.FromString,
-                    response_serializer=backend__pb2.StatusResponse.SerializeToString,
-            ),
-    }
-    generic_handler = grpc.method_handlers_generic_handler(
-            'backend.Backend', rpc_method_handlers)
-    server.add_generic_rpc_handlers((generic_handler,))
-
-
- # This class is part of an EXPERIMENTAL API.
-class Backend(object):
-    """Missing associated documentation comment in .proto file."""
-
-    @staticmethod
-    def Health(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Health',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Predict(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Predict',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def LoadModel(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/LoadModel',
-            backend__pb2.ModelOptions.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def PredictStream(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_stream(request, target, '/backend.Backend/PredictStream',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.Reply.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Embedding(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Embedding',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.EmbeddingResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def GenerateImage(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/GenerateImage',
-            backend__pb2.GenerateImageRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def AudioTranscription(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/AudioTranscription',
-            backend__pb2.TranscriptRequest.SerializeToString,
-            backend__pb2.TranscriptResult.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TTS(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TTS',
-            backend__pb2.TTSRequest.SerializeToString,
-            backend__pb2.Result.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def TokenizeString(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/TokenizeString',
-            backend__pb2.PredictOptions.SerializeToString,
-            backend__pb2.TokenizationResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def Status(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/backend.Backend/Status',
-            backend__pb2.HealthMessage.SerializeToString,
-            backend__pb2.StatusResponse.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)

backend/python/bark/install.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+set -e
+
+source $(dirname $0)/../common/libbackend.sh
+
+# This is here because the Intel pip index is broken and returns 200 status codes for every package name, it just doesn't return any package links.
+# This makes uv think that the package exists in the Intel pip index, and by default it stops looking at other pip indexes once it finds a match.
+# We need uv to continue falling through to the pypi default index to find optimum[openvino] in the pypi index
+# the --upgrade actually allows us to *downgrade* torch to the version provided in the Intel pip index
+if [ "x${BUILD_PROFILE}" == "xintel" ]; then
+    EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
+fi
+
+installRequirements

backend/python/bark/requirements-hipblas.txt

@@ -0,0 +1,3 @@
+--extra-index-url https://download.pytorch.org/whl/rocm6.0
+torch
+torchaudio

backend/python/bark/requirements-intel.txt

@@ -0,0 +1,6 @@
+--extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
+intel-extension-for-pytorch
+torch
+torchaudio
+optimum[openvino]
+setuptools==69.5.1 # https://github.com/mudler/LocalAI/issues/2406

backend/python/bark/requirements.txt

@@ -0,0 +1,6 @@
+accelerate
+bark==0.1.5
+grpcio==1.64.0
+protobuf
+certifi
+transformers

backend/python/bark/run.sh

@@ -1,14 +1,4 @@
 #!/bin/bash
+source $(dirname $0)/../common/libbackend.sh
 
-##
-## A bash script wrapper that runs the ttsbark server with conda
-
-export PATH=$PATH:/opt/conda/bin
-
-# Activate conda environment
-source activate transformers
-
-# get the directory where the bash script is located
-DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
-
-python $DIR/ttsbark.py $@
+startBackend $@

backend/python/bark/test.py

@@ -18,7 +18,7 @@ class TestBackendServicer(unittest.TestCase):
         """
         This method sets up the gRPC service by starting the server
         """
-        self.service = subprocess.Popen(["python3", "ttsbark.py", "--addr", "localhost:50051"])
+        self.service = subprocess.Popen(["python3", "backend.py", "--addr", "localhost:50051"])
         time.sleep(10)
 
     def tearDown(self) -> None:

backend/python/bark/test.sh

@@ -1,11 +1,6 @@
 #!/bin/bash
-##
-## A bash script wrapper that runs the bark server with conda
+set -e
 
-# Activate conda environment
-source activate transformers
+source $(dirname $0)/../common/libbackend.sh
 
-# get the directory where the bash script is located
-DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
-
-python -m unittest $DIR/test.py
+runUnittests

backend/python/common-env/transformers/Makefile

@@ -1,21 +0,0 @@
-CONDA_ENV_PATH = "transformers.yml"
-
-ifeq ($(BUILD_TYPE), cublas)
-	CONDA_ENV_PATH = "transformers-nvidia.yml"
-endif
-
-ifeq ($(BUILD_TYPE), hipblas)
-	CONDA_ENV_PATH = "transformers-rocm.yml"
-endif
-
-# Intel GPU are supposed to have dependencies installed in the main python
-# environment, so we skip conda installation for SYCL builds.
-# https://github.com/intel/intel-extension-for-pytorch/issues/538
-ifneq (,$(findstring sycl,$(BUILD_TYPE)))
-export SKIP_CONDA=1
-endif
-
-.PHONY: transformers
-transformers:
-	@echo "Installing $(CONDA_ENV_PATH)..."
-	bash install.sh $(CONDA_ENV_PATH)

backend/python/common-env/transformers/install.sh

@@ -1,38 +0,0 @@
-#!/bin/bash
-set -ex
-
-SKIP_CONDA=${SKIP_CONDA:-0}
-
-# Check if environment exist
-conda_env_exists(){
-    ! conda list --name "${@}" >/dev/null 2>/dev/null
-}
-
-if [ $SKIP_CONDA -eq 1 ]; then
-    echo "Skipping conda environment installation"
-else
-    export PATH=$PATH:/opt/conda/bin
-    if conda_env_exists "transformers" ; then
-        echo "Creating virtual environment..."
-        conda env create --name transformers --file $1
-        echo "Virtual environment created."
-    else 
-        echo "Virtual environment already exists."
-    fi
-fi
-
-if [ -d "/opt/intel" ]; then
-    # Intel GPU: If the directory exists, we assume we are using the intel image
-    # (no conda env)
-    # https://github.com/intel/intel-extension-for-pytorch/issues/538
-    pip install intel-extension-for-transformers datasets sentencepiece tiktoken neural_speed
-fi
-
-if [ "$PIP_CACHE_PURGE" = true ] ; then
-    if [ $SKIP_CONDA -eq 0 ]; then
-        # Activate conda environment
-        source activate transformers
-    fi
-
-    pip cache purge
-fi

backend/python/common-env/transformers/transformers-nvidia.yml

@@ -1,119 +0,0 @@
-name: transformers
-channels:
-  - defaults
-dependencies:
-  - _libgcc_mutex=0.1=main
-  - _openmp_mutex=5.1=1_gnu
-  - bzip2=1.0.8=h7b6447c_0
-  - ca-certificates=2023.08.22=h06a4308_0
-  - ld_impl_linux-64=2.38=h1181459_1
-  - libffi=3.4.4=h6a678d5_0
-  - libgcc-ng=11.2.0=h1234567_1
-  - libgomp=11.2.0=h1234567_1
-  - libstdcxx-ng=11.2.0=h1234567_1
-  - libuuid=1.41.5=h5eee18b_0
-  - ncurses=6.4=h6a678d5_0
-  - openssl=3.0.11=h7f8727e_2
-  - pip=23.2.1=py311h06a4308_0
-  - python=3.11.5=h955ad1f_0
-  - readline=8.2=h5eee18b_0
-  - setuptools=68.0.0=py311h06a4308_0
-  - sqlite=3.41.2=h5eee18b_0
-  - tk=8.6.12=h1ccaba5_0
-  - wheel=0.41.2=py311h06a4308_0
-  - xz=5.4.2=h5eee18b_0
-  - zlib=1.2.13=h5eee18b_0
-  - pip:
-      - accelerate==0.23.0
-      - aiohttp==3.8.5
-      - aiosignal==1.3.1
-      - async-timeout==4.0.3
-      - attrs==23.1.0
-      - bark==0.1.5
-      - boto3==1.28.61
-      - botocore==1.31.61
-      - certifi==2023.7.22
-      - TTS==0.22.0
-      - charset-normalizer==3.3.0
-      - datasets==2.14.5
-      - sentence-transformers==2.5.1 # Updated Version
-      - sentencepiece==0.1.99
-      - dill==0.3.7
-      - einops==0.7.0
-      - encodec==0.1.1
-      - filelock==3.12.4
-      - frozenlist==1.4.0
-      - fsspec==2023.6.0
-      - funcy==2.0
-      - grpcio==1.59.0
-      - huggingface-hub
-      - idna==3.4
-      - jinja2==3.1.2
-      - jmespath==1.0.1
-      - markupsafe==2.1.3
-      - mpmath==1.3.0
-      - multidict==6.0.4
-      - multiprocess==0.70.15
-      - networkx
-      - numpy==1.26.0
-      - nvidia-cublas-cu12==12.1.3.1
-      - nvidia-cuda-cupti-cu12==12.1.105
-      - nvidia-cuda-nvrtc-cu12==12.1.105
-      - nvidia-cuda-runtime-cu12==12.1.105
-      - nvidia-cudnn-cu12==8.9.2.26
-      - nvidia-cufft-cu12==11.0.2.54
-      - nvidia-curand-cu12==10.3.2.106
-      - nvidia-cusolver-cu12==11.4.5.107
-      - nvidia-cusparse-cu12==12.1.0.106
-      - nvidia-nccl-cu12==2.18.1
-      - nvidia-nvjitlink-cu12==12.2.140
-      - nvidia-nvtx-cu12==12.1.105
-      - packaging==23.2
-      - pandas
-      - peft==0.5.0
-      - protobuf==4.24.4
-      - psutil==5.9.5
-      - pyarrow==13.0.0
-      - python-dateutil==2.8.2
-      - pytz==2023.3.post1
-      - pyyaml==6.0.1
-      - regex==2023.10.3
-      - requests==2.31.0
-      - rouge==1.0.1
-      - s3transfer==0.7.0
-      - safetensors>=0.4.1
-      - scipy==1.12.0 # Updated Version
-      - six==1.16.0
-      - sympy==1.12
-      - tokenizers
-      - torch==2.1.2
-      - torchaudio==2.1.2
-      - tqdm==4.66.1
-      - triton==2.1.0
-      - typing-extensions==4.8.0
-      - tzdata==2023.3
-      - urllib3==1.26.17
-      - xxhash==3.4.1
-      - auto-gptq==0.6.0
-      - yarl==1.9.2
-      - soundfile
-      - langid
-      - wget
-      - unidecode
-      - pyopenjtalk-prebuilt
-      - pypinyin
-      - inflect
-      - cn2an
-      - jieba
-      - eng_to_ipa
-      - openai-whisper
-      - matplotlib
-      - gradio==3.41.2
-      - nltk
-      - sudachipy
-      - sudachidict_core
-      - vocos
-      - vllm==0.3.2
-      - transformers>=4.38.2  # Updated Version
-      - xformers==0.0.23.post1  
-prefix: /opt/conda/envs/transformers

backend/python/common-env/transformers/transformers-rocm.yml

@@ -1,109 +0,0 @@
-name: transformers
-channels:
-  - defaults
-dependencies:
-  - _libgcc_mutex=0.1=main
-  - _openmp_mutex=5.1=1_gnu
-  - bzip2=1.0.8=h7b6447c_0
-  - ca-certificates=2023.08.22=h06a4308_0
-  - ld_impl_linux-64=2.38=h1181459_1
-  - libffi=3.4.4=h6a678d5_0
-  - libgcc-ng=11.2.0=h1234567_1
-  - libgomp=11.2.0=h1234567_1
-  - libstdcxx-ng=11.2.0=h1234567_1
-  - libuuid=1.41.5=h5eee18b_0
-  - ncurses=6.4=h6a678d5_0
-  - openssl=3.0.11=h7f8727e_2
-  - pip=23.2.1=py311h06a4308_0
-  - python=3.11.5=h955ad1f_0
-  - readline=8.2=h5eee18b_0
-  - setuptools=68.0.0=py311h06a4308_0
-  - sqlite=3.41.2=h5eee18b_0
-  - tk=8.6.12=h1ccaba5_0
-  - wheel=0.41.2=py311h06a4308_0
-  - xz=5.4.2=h5eee18b_0
-  - zlib=1.2.13=h5eee18b_0
-  - pip:
-      - --pre
-      - --extra-index-url https://download.pytorch.org/whl/nightly/
-      - accelerate==0.23.0
-      - aiohttp==3.8.5
-      - aiosignal==1.3.1
-      - async-timeout==4.0.3
-      - attrs==23.1.0
-      - bark==0.1.5
-      - boto3==1.28.61
-      - botocore==1.31.61
-      - certifi==2023.7.22
-      - TTS==0.22.0
-      - charset-normalizer==3.3.0
-      - datasets==2.14.5
-      - sentence-transformers==2.5.1 # Updated Version
-      - sentencepiece==0.1.99
-      - dill==0.3.7
-      - einops==0.7.0
-      - encodec==0.1.1
-      - filelock==3.12.4
-      - frozenlist==1.4.0
-      - fsspec==2023.6.0
-      - funcy==2.0
-      - grpcio==1.59.0
-      - huggingface-hub
-      - idna==3.4
-      - jinja2==3.1.2
-      - jmespath==1.0.1
-      - markupsafe==2.1.3
-      - mpmath==1.3.0
-      - multidict==6.0.4
-      - multiprocess==0.70.15
-      - networkx
-      - numpy==1.26.0
-      - packaging==23.2
-      - pandas
-      - peft==0.5.0
-      - protobuf==4.24.4
-      - psutil==5.9.5
-      - pyarrow==13.0.0
-      - python-dateutil==2.8.2
-      - pytz==2023.3.post1
-      - pyyaml==6.0.1
-      - regex==2023.10.3
-      - requests==2.31.0
-      - rouge==1.0.1
-      - s3transfer==0.7.0
-      - safetensors>=0.4.1
-      - scipy==1.12.0 # Updated Version
-      - six==1.16.0
-      - sympy==1.12
-      - tokenizers
-      - torch
-      - torchaudio
-      - tqdm==4.66.1
-      - triton==2.1.0
-      - typing-extensions==4.8.0
-      - tzdata==2023.3
-      - auto-gptq==0.6.0
-      - urllib3==1.26.17
-      - xxhash==3.4.1
-      - yarl==1.9.2
-      - soundfile
-      - langid
-      - wget
-      - unidecode
-      - pyopenjtalk-prebuilt
-      - pypinyin
-      - inflect
-      - cn2an
-      - jieba
-      - eng_to_ipa
-      - openai-whisper
-      - matplotlib
-      - gradio==3.41.2
-      - nltk
-      - sudachipy
-      - sudachidict_core
-      - vocos
-      - vllm==0.3.2
-      - transformers>=4.38.2  # Updated Version
-      - xformers==0.0.23.post1
-prefix: /opt/conda/envs/transformers

backend/python/common-env/transformers/transformers.yml

@@ -1,107 +0,0 @@
-name: transformers
-channels:
-  - defaults
-dependencies:
-  - _libgcc_mutex=0.1=main
-  - _openmp_mutex=5.1=1_gnu
-  - bzip2=1.0.8=h7b6447c_0
-  - ca-certificates=2023.08.22=h06a4308_0
-  - ld_impl_linux-64=2.38=h1181459_1
-  - libffi=3.4.4=h6a678d5_0
-  - libgcc-ng=11.2.0=h1234567_1
-  - libgomp=11.2.0=h1234567_1
-  - libstdcxx-ng=11.2.0=h1234567_1
-  - libuuid=1.41.5=h5eee18b_0
-  - ncurses=6.4=h6a678d5_0
-  - openssl=3.0.11=h7f8727e_2
-  - pip=23.2.1=py311h06a4308_0
-  - python=3.11.5=h955ad1f_0
-  - readline=8.2=h5eee18b_0
-  - setuptools=68.0.0=py311h06a4308_0
-  - sqlite=3.41.2=h5eee18b_0
-  - tk=8.6.12=h1ccaba5_0
-  - wheel=0.41.2=py311h06a4308_0
-  - xz=5.4.2=h5eee18b_0
-  - zlib=1.2.13=h5eee18b_0
-  - pip:
-      - accelerate==0.23.0
-      - aiohttp==3.8.5
-      - aiosignal==1.3.1
-      - async-timeout==4.0.3
-      - attrs==23.1.0
-      - bark==0.1.5
-      - boto3==1.28.61
-      - botocore==1.31.61
-      - certifi==2023.7.22
-      - TTS==0.22.0
-      - charset-normalizer==3.3.0
-      - datasets==2.14.5
-      - sentence-transformers==2.5.1 # Updated Version
-      - sentencepiece==0.1.99
-      - dill==0.3.7
-      - einops==0.7.0
-      - encodec==0.1.1
-      - filelock==3.12.4
-      - frozenlist==1.4.0
-      - fsspec==2023.6.0
-      - funcy==2.0
-      - grpcio==1.59.0
-      - huggingface-hub
-      - idna==3.4
-      - jinja2==3.1.2
-      - jmespath==1.0.1
-      - markupsafe==2.1.3
-      - mpmath==1.3.0
-      - multidict==6.0.4
-      - multiprocess==0.70.15
-      - networkx
-      - numpy==1.26.0
-      - packaging==23.2
-      - pandas
-      - peft==0.5.0
-      - protobuf==4.24.4
-      - psutil==5.9.5
-      - pyarrow==13.0.0
-      - python-dateutil==2.8.2
-      - pytz==2023.3.post1
-      - pyyaml==6.0.1
-      - regex==2023.10.3
-      - requests==2.31.0
-      - rouge==1.0.1
-      - s3transfer==0.7.0
-      - safetensors>=0.4.1
-      - scipy==1.12.0 # Updated Version
-      - six==1.16.0
-      - sympy==1.12
-      - tokenizers
-      - torch==2.1.2
-      - torchaudio==2.1.2
-      - tqdm==4.66.1
-      - triton==2.1.0
-      - typing-extensions==4.8.0
-      - tzdata==2023.3
-      - auto-gptq==0.6.0
-      - urllib3==1.26.17
-      - xxhash==3.4.1
-      - yarl==1.9.2
-      - soundfile
-      - langid
-      - wget
-      - unidecode
-      - pyopenjtalk-prebuilt
-      - pypinyin
-      - inflect
-      - cn2an
-      - jieba
-      - eng_to_ipa
-      - openai-whisper
-      - matplotlib
-      - gradio==3.41.2
-      - nltk
-      - sudachipy
-      - sudachidict_core
-      - vocos
-      - vllm==0.3.2
-      - transformers>=4.38.2  # Updated Version
-      - xformers==0.0.23.post1  
-prefix: /opt/conda/envs/transformers

backend/python/common/libbackend.sh

@@ -0,0 +1,213 @@
+
+
+# init handles the setup of the library
+# 
+# use the library by adding the following line to a script:
+# source $(dirname $0)/../common/libbackend.sh
+#
+# If you want to limit what targets a backend can be used on, set the variable LIMIT_TARGETS to a
+# space separated list of valid targets BEFORE sourcing the library, for example to only allow a backend
+# to be used on CUDA and CPU backends:
+#
+# LIMIT_TARGETS="cublas cpu"
+# source $(dirname $0)/../common/libbackend.sh
+#
+# You can use any valid BUILD_TYPE or BUILD_PROFILE, if you need to limit a backend to CUDA 12 only:
+#
+# LIMIT_TARGETS="cublas12"
+# source $(dirname $0)/../common/libbackend.sh
+#
+function init() {
+    BACKEND_NAME=${PWD##*/}
+    MY_DIR=$(realpath `dirname $0`)
+    BUILD_PROFILE=$(getBuildProfile)
+
+    # If a backend has defined a list of valid build profiles...
+    if [ ! -z "${LIMIT_TARGETS}" ]; then
+        isValidTarget=$(checkTargets ${LIMIT_TARGETS})
+        if [ ${isValidTarget} != true ]; then
+            echo "${BACKEND_NAME} can only be used on the following targets: ${LIMIT_TARGETS}"
+            exit 0
+        fi
+    fi
+
+    echo "Initializing libbackend for ${BACKEND_NAME}"
+}
+
+# getBuildProfile will inspect the system to determine which build profile is appropriate:
+# returns one of the following:
+# - cublas11
+# - cublas12
+# - hipblas
+# - intel
+function getBuildProfile() {
+    # First check if we are a cublas build, and if so report the correct build profile
+    if [ x"${BUILD_TYPE}" == "xcublas" ]; then
+        if [ ! -z ${CUDA_MAJOR_VERSION} ]; then
+            # If we have been given a CUDA version, we trust it
+            echo ${BUILD_TYPE}${CUDA_MAJOR_VERSION}
+        else
+            # We don't know what version of cuda we are, so we report ourselves as a generic cublas
+            echo ${BUILD_TYPE}
+        fi
+        return 0
+    fi
+
+    # If /opt/intel exists, then we are doing an intel/ARC build
+    if [ -d "/opt/intel" ]; then
+        echo "intel"
+        return 0
+    fi
+
+    # If for any other values of BUILD_TYPE, we don't need any special handling/discovery
+    if [ ! -z ${BUILD_TYPE} ]; then
+        echo ${BUILD_TYPE}
+        return 0
+    fi
+
+    # If there is no BUILD_TYPE set at all, set a build-profile value of CPU, we aren't building for any GPU targets
+    echo "cpu"
+}
+
+# ensureVenv makes sure that the venv for the backend both exists, and is activated.
+#
+# This function is idempotent, so you can call it as many times as you want and it will
+# always result in an activated virtual environment
+function ensureVenv() {
+    if [ ! -d "${MY_DIR}/venv" ]; then
+        uv venv ${MY_DIR}/venv
+        echo "virtualenv created"
+    fi
+    
+    if [ "x${VIRTUAL_ENV}" != "x${MY_DIR}/venv" ]; then
+        source ${MY_DIR}/venv/bin/activate
+        echo "virtualenv activated"
+    fi
+
+    echo "activated virtualenv has been ensured"
+}
+
+# installRequirements looks for several requirements files and if they exist runs the install for them in order
+#
+#  - requirements-install.txt
+#  - requirements.txt
+#  - requirements-${BUILD_TYPE}.txt
+#  - requirements-${BUILD_PROFILE}.txt
+#
+# BUILD_PROFILE is a pore specific version of BUILD_TYPE, ex: cuda11 or cuda12
+# it can also include some options that we do not have BUILD_TYPES for, ex: intel
+#
+# NOTE: for BUILD_PROFILE==intel, this function does NOT automatically use the Intel python package index.
+# you may want to add the following line to a requirements-intel.txt if you use one:
+#
+# --index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
+#
+# If you need to add extra flags into the pip install command you can do so by setting the variable EXTRA_PIP_INSTALL_FLAGS
+# before calling installRequirements.  For example:
+#
+# source $(dirname $0)/../common/libbackend.sh
+# EXTRA_PIP_INSTALL_FLAGS="--no-build-isolation"
+# installRequirements
+function installRequirements() {
+    ensureVenv
+
+    # These are the requirements files we will attempt to install, in order
+    declare -a requirementFiles=(
+        "${MY_DIR}/requirements-install.txt"
+        "${MY_DIR}/requirements.txt"
+        "${MY_DIR}/requirements-${BUILD_TYPE}.txt"
+    )
+
+    if [ "x${BUILD_TYPE}" != "x${BUILD_PROFILE}" ]; then
+        requirementFiles+=("${MY_DIR}/requirements-${BUILD_PROFILE}.txt")
+    fi
+
+    for reqFile in ${requirementFiles[@]}; do
+        if [ -f ${reqFile} ]; then
+            echo "starting requirements install for ${reqFile}"
+            uv pip install ${EXTRA_PIP_INSTALL_FLAGS} --requirement ${reqFile}
+            echo "finished requirements install for ${reqFile}"
+        fi
+    done
+}
+
+# startBackend discovers and runs the backend GRPC server
+#
+# You can specify a specific backend file to execute by setting BACKEND_FILE before calling startBackend.
+# example:
+#
+# source ../common/libbackend.sh
+# BACKEND_FILE="${MY_DIR}/source/backend.py"
+# startBackend $@
+#
+# valid filenames for autodiscovered backend servers are:
+#  - server.py
+#  - backend.py
+#  - ${BACKEND_NAME}.py
+function startBackend() {
+    ensureVenv
+
+    if [ ! -z ${BACKEND_FILE} ]; then
+        python ${BACKEND_FILE} $@
+    elif [ -e "${MY_DIR}/server.py" ]; then
+        python ${MY_DIR}/server.py $@
+    elif [ -e "${MY_DIR}/backend.py" ]; then
+        python ${MY_DIR}/backend.py $@
+    elif [ -e "${MY_DIR}/${BACKEND_NAME}.py" ]; then
+        python ${MY_DIR}/${BACKEND_NAME}.py $@
+    fi
+}
+
+# runUnittests discovers and runs python unittests
+#
+# You can specify a specific test file to use by setting TEST_FILE before calling runUnittests.
+# example:
+#
+# source ../common/libbackend.sh
+# TEST_FILE="${MY_DIR}/source/test.py"
+# runUnittests $@
+#
+# be default a file named test.py in the backends directory will be used
+function runUnittests() {
+    ensureVenv
+
+    if [ ! -z ${TEST_FILE} ]; then
+        testDir=$(dirname `realpath ${TEST_FILE}`)
+        testFile=$(basename ${TEST_FILE})
+        pushd ${testDir}
+        python -m unittest ${testFile}
+        popd
+    elif [ -f "${MY_DIR}/test.py" ]; then
+        pushd ${MY_DIR}
+        python -m unittest test.py
+        popd
+    else
+        echo "no tests defined for ${BACKEND_NAME}"
+    fi
+}
+
+##################################################################################
+# Below here are helper functions not intended to be used outside of the library #
+##################################################################################
+
+# checkTargets determines if the current BUILD_TYPE or BUILD_PROFILE is in a list of valid targets
+function checkTargets() {
+    # Collect all provided targets into a variable and...
+    targets=$@
+    # ...convert it into an array
+    declare -a targets=($targets)
+
+    for target in ${targets[@]}; do
+        if [ "x${BUILD_TYPE}" == "x${target}" ]; then
+            echo true
+            return 0
+        fi
+        if [ "x${BUILD_PROFILE}" == "x${target}" ]; then
+            echo true
+            return 0
+        fi
+    done
+    echo false
+}
+
+init

backend/python/common/template/Makefile

@@ -0,0 +1,19 @@
+.DEFAULT_GOAL := install
+
+.PHONY: install
+install: protogen
+	bash install.sh
+
+.PHONY: protogen
+protogen: backend_pb2_grpc.py backend_pb2.py
+
+.PHONY: protogen-clean
+protogen-clean:
+	$(RM) backend_pb2_grpc.py backend_pb2.py
+
+backend_pb2_grpc.py backend_pb2.py:
+	python3 -m grpc_tools.protoc -I../.. --python_out=. --grpc_python_out=. backend.proto
+
+.PHONY: clean
+clean: protogen-clean
+	rm -rf venv __pycache__

backend/python/common/template/backend.py

@@ -0,0 +1,4 @@
+#!/usr/bin/env python3
+import grpc
+import backend_pb2
+import backend_pb2_grpc