fix: do not break on newlines on function returns (#864)

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

.gitattributes

@@ -0,0 +1 @@
+*.sh text eol=lf

Makefile

@@ -4,11 +4,11 @@ GOVET=$(GOCMD) vet
 BINARY_NAME=local-ai
 
 # llama.cpp versions
-GOLLAMA_VERSION?=6ba16de8e965e5aa0f32d25ef9d6149bb6586565
+GOLLAMA_VERSION?=50cee7712066d9e38306eccadcfbb44ea87df4b7
 
 # gpt4all version
 GPT4ALL_REPO?=https://github.com/nomic-ai/gpt4all
-GPT4ALL_VERSION?=9100b2ef6fb95a1c49a3e033980d555818f01901
+GPT4ALL_VERSION?=c449b71b56de1ced375a64a986381cf70cec3080
 
 # go-ggml-transformers version
 GOGGMLTRANSFORMERS_VERSION?=ffb09d7dd71e2cbc6c5d7d05357d230eea6f369a

README.md

@@ -1,12 +1,15 @@
 <h1 align="center">
   <br>
-  <img height="300" src="https://user-images.githubusercontent.com/2420543/233147843-88697415-6dbf-4368-a862-ab217f9f7342.jpeg"> <br>
+  <img height="300" src="https://github.com/go-skynet/LocalAI/assets/2420543/0966aa2a-166e-4f99-a3e5-6c915fc997dd"> <br>
     LocalAI
 <br>
 </h1>
 
 [![tests](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml/badge.svg)](https://github.com/go-skynet/LocalAI/actions/workflows/test.yml) [![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)
 
+[![Artifact Hub](https://img.shields.io/endpoint?url=https://artifacthub.io/badge/repository/localai)](https://artifacthub.io/packages/search?repo=localai)
+
+
 [![](https://dcbadge.vercel.app/api/server/uJAeKSAGDy?style=flat-square&theme=default-inverted)](https://discord.gg/uJAeKSAGDy) 
 
 [Documentation website](https://localai.io/)
@@ -24,6 +27,7 @@ In a nutshell:
   - 🔈 [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/) 🆕
+  - 🧠 [Embeddings generation for vector databases](https://localai.io/features/embeddings/)
 - 🏃 Once loaded the first time, it keep models loaded in memory for faster inference
 - ⚡ Doesn't shell-out, but uses C++ bindings for a faster inference and better performance. 
 
@@ -202,6 +206,8 @@ Feel free to open up a PR to get your project listed!
 - [autogpt4all](https://github.com/aorumbayev/autogpt4all)
 - [Mods](https://github.com/charmbracelet/mods)
 - [Flowise](https://github.com/FlowiseAI/Flowise)
+- [BMO Chatbot](https://github.com/longy2k/obsidian-bmo-chatbot)
+- [Mattermost OpenOps](https://openops.mattermost.com)
 
 ## Sponsors
 

api/api_test.go

@@ -8,7 +8,6 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"io/ioutil"
 	"net/http"
 	"os"
 	"path/filepath"
@@ -45,7 +44,7 @@ func getModelStatus(url string) (response map[string]interface{}) {
 	}
 	defer resp.Body.Close()
 
-	body, err := ioutil.ReadAll(resp.Body)
+	body, err := io.ReadAll(resp.Body)
 	if err != nil {
 		fmt.Println("Error reading response body:", err)
 		return
@@ -97,7 +96,7 @@ func postModelApplyRequest(url string, request modelApplyRequest) (response map[
 	}
 	defer resp.Body.Close()
 
-	body, err := ioutil.ReadAll(resp.Body)
+	body, err := io.ReadAll(resp.Body)
 	if err != nil {
 		fmt.Println("Error reading response body:", err)
 		return
@@ -153,7 +152,7 @@ var _ = Describe("API test", func() {
 			}
 			out, err := yaml.Marshal(g)
 			Expect(err).ToNot(HaveOccurred())
-			err = ioutil.WriteFile(filepath.Join(tmpdir, "gallery_simple.yaml"), out, 0644)
+			err = os.WriteFile(filepath.Join(tmpdir, "gallery_simple.yaml"), out, 0644)
 			Expect(err).ToNot(HaveOccurred())
 
 			galleries := []gallery.Gallery{
@@ -535,7 +534,7 @@ var _ = Describe("API test", func() {
 		It("returns the models list", func() {
 			models, err := client.ListModels(context.TODO())
 			Expect(err).ToNot(HaveOccurred())
-			Expect(len(models.Models)).To(Equal(11))
+			Expect(len(models.Models)).To(Equal(6)) // If "config.yaml" should be included, this should be 8?
 		})
 		It("can generate completions", func() {
 			resp, err := client.CreateCompletion(context.TODO(), openai.CompletionRequest{Model: "testmodel", Prompt: "abcdedfghikl"})
@@ -739,19 +738,14 @@ var _ = Describe("API test", func() {
 			cancel()
 			app.Shutdown()
 		})
-		It("can generate chat completions from config file", func() {
-			models, err := client.ListModels(context.TODO())
-			Expect(err).ToNot(HaveOccurred())
-			Expect(len(models.Models)).To(Equal(13))
-		})
-		It("can generate chat completions from config file", func() {
-			resp, err := client.CreateChatCompletion(context.TODO(), openai.ChatCompletionRequest{Model: "list1", Messages: []openai.ChatCompletionMessage{openai.ChatCompletionMessage{Role: "user", Content: "abcdedfghikl"}}})
+		It("can generate chat completions from config file (list1)", func() {
+			resp, err := client.CreateChatCompletion(context.TODO(), openai.ChatCompletionRequest{Model: "list1", Messages: []openai.ChatCompletionMessage{{Role: "user", Content: "abcdedfghikl"}}})
 			Expect(err).ToNot(HaveOccurred())
 			Expect(len(resp.Choices)).To(Equal(1))
 			Expect(resp.Choices[0].Message.Content).ToNot(BeEmpty())
 		})
-		It("can generate chat completions from config file", func() {
-			resp, err := client.CreateChatCompletion(context.TODO(), openai.ChatCompletionRequest{Model: "list2", Messages: []openai.ChatCompletionMessage{openai.ChatCompletionMessage{Role: "user", Content: "abcdedfghikl"}}})
+		It("can generate chat completions from config file (list2)", func() {
+			resp, err := client.CreateChatCompletion(context.TODO(), openai.ChatCompletionRequest{Model: "list2", Messages: []openai.ChatCompletionMessage{{Role: "user", Content: "abcdedfghikl"}}})
 			Expect(err).ToNot(HaveOccurred())
 			Expect(len(resp.Choices)).To(Equal(1))
 			Expect(resp.Choices[0].Message.Content).ToNot(BeEmpty())

api/backend/options.go

@@ -18,6 +18,8 @@ func gRPCModelOpts(c config.Config) *pb.ModelOptions {
 		ContextSize:   int32(c.ContextSize),
 		Seed:          int32(c.Seed),
 		NBatch:        int32(b),
+		NGQA:          c.NGQA,
+		RMSNormEps:    c.RMSNormEps,
 		F16Memory:     c.F16,
 		MLock:         c.MMlock,
 		RopeFreqBase:  c.RopeFreqBase,

api/config/config.go

@@ -51,6 +51,9 @@ type Config struct {
 	FunctionsConfig Functions `yaml:"function"`
 
 	SystemPrompt string `yaml:"system_prompt"`
+
+	RMSNormEps float32 `yaml:"rms_norm_eps"`
+	NGQA       int32   `yaml:"ngqa"`
 }
 
 type Functions struct {
@@ -172,6 +175,16 @@ func (cm *ConfigLoader) GetConfig(m string) (Config, bool) {
 	return v, exists
 }
 
+func (cm *ConfigLoader) GetAllConfigs() []Config {
+	cm.Lock()
+	defer cm.Unlock()
+	var res []Config
+	for _, v := range cm.configs {
+		res = append(res, v)
+	}
+	return res
+}
+
 func (cm *ConfigLoader) ListConfigs() []string {
 	cm.Lock()
 	defer cm.Unlock()

api/localai/gallery.go

@@ -8,6 +8,7 @@ import (
 	"sync"
 
 	json "github.com/json-iterator/go"
+	"gopkg.in/yaml.v3"
 
 	config "github.com/go-skynet/LocalAI/api/config"
 	"github.com/go-skynet/LocalAI/pkg/gallery"
@@ -128,8 +129,27 @@ func (g *galleryApplier) Start(c context.Context, cm *config.ConfigLoader) {
 }
 
 type galleryModel struct {
-	gallery.GalleryModel
-	ID string `json:"id"`
+	gallery.GalleryModel `yaml:",inline"` // https://github.com/go-yaml/yaml/issues/63
+	ID                   string           `json:"id"`
+}
+
+func processRequests(modelPath, s string, cm *config.ConfigLoader, galleries []gallery.Gallery, requests []galleryModel) error {
+	var err error
+	for _, r := range requests {
+		utils.ResetDownloadTimers()
+		if r.ID == "" {
+			err = prepareModel(modelPath, r.GalleryModel, cm, utils.DisplayDownloadFunction)
+		} else {
+			if strings.Contains(r.ID, "@") {
+				err = gallery.InstallModelFromGallery(
+					galleries, r.ID, modelPath, r.GalleryModel, utils.DisplayDownloadFunction)
+			} else {
+				err = gallery.InstallModelFromGalleryByName(
+					galleries, r.ID, modelPath, r.GalleryModel, utils.DisplayDownloadFunction)
+			}
+		}
+	}
+	return err
 }
 
 func ApplyGalleryFromFile(modelPath, s string, cm *config.ConfigLoader, galleries []gallery.Gallery) error {
@@ -137,7 +157,13 @@ func ApplyGalleryFromFile(modelPath, s string, cm *config.ConfigLoader, gallerie
 	if err != nil {
 		return err
 	}
-	return ApplyGalleryFromString(modelPath, string(dat), cm, galleries)
+	var requests []galleryModel
+
+	if err := yaml.Unmarshal(dat, &requests); err != nil {
+		return err
+	}
+
+	return processRequests(modelPath, s, cm, galleries, requests)
 }
 
 func ApplyGalleryFromString(modelPath, s string, cm *config.ConfigLoader, galleries []gallery.Gallery) error {
@@ -147,16 +173,7 @@ func ApplyGalleryFromString(modelPath, s string, cm *config.ConfigLoader, galler
 		return err
 	}
 
-	for _, r := range requests {
-		utils.ResetDownloadTimers()
-		if r.ID == "" {
-			err = prepareModel(modelPath, r.GalleryModel, cm, utils.DisplayDownloadFunction)
-		} else {
-			err = gallery.InstallModelFromGallery(galleries, r.ID, modelPath, r.GalleryModel, utils.DisplayDownloadFunction)
-		}
-	}
-
-	return err
+	return processRequests(modelPath, s, cm, galleries, requests)
 }
 
 /// Endpoints

api/openai/chat.go

@@ -12,6 +12,7 @@ import (
 	"github.com/go-skynet/LocalAI/api/options"
 	"github.com/go-skynet/LocalAI/pkg/grammar"
 	model "github.com/go-skynet/LocalAI/pkg/model"
+	"github.com/go-skynet/LocalAI/pkg/utils"
 	"github.com/gofiber/fiber/v2"
 	"github.com/rs/zerolog/log"
 	"github.com/valyala/fasthttp"
@@ -109,6 +110,7 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 
 		var predInput string
 
+		suppressConfigSystemPrompt := false
 		mess := []string{}
 		for messageIndex, i := range input.Messages {
 			var content string
@@ -146,7 +148,7 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 					content = templatedChatMessage
 				}
 			}
-			// If this model doesn't have such a template, or if
+			// If this model doesn't have such a template, or if that template fails to return a value, template at the message level.
 			if content == "" {
 				if r != "" {
 					if contentExists {
@@ -177,6 +179,10 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 						}
 					}
 				}
+				// Special Handling: System. We care if it was printed at all, not the r branch, so check seperately
+				if contentExists && role == "system" {
+					suppressConfigSystemPrompt = true
+				}
 			}
 
 			mess = append(mess, content)
@@ -207,8 +213,10 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 
 		// A model can have a "file.bin.tmpl" file associated with a prompt template prefix
 		templatedInput, err := o.Loader.EvaluateTemplateForPrompt(model.ChatPromptTemplate, templateFile, model.PromptTemplateData{
-			Input:     predInput,
-			Functions: funcs,
+			SystemPrompt:         config.SystemPrompt,
+			SuppressSystemPrompt: suppressConfigSystemPrompt,
+			Input:                predInput,
+			Functions:            funcs,
 		})
 		if err == nil {
 			predInput = templatedInput
@@ -267,6 +275,8 @@ func ChatEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 			if processFunctions {
 				// As we have to change the result before processing, we can't stream the answer (yet?)
 				ss := map[string]interface{}{}
+				// This prevent newlines to break JSON parsing for clients
+				s = utils.EscapeNewLines(s)
 				json.Unmarshal([]byte(s), &ss)
 				log.Debug().Msgf("Function return: %s %+v", s, ss)
 

api/openai/completion.go

@@ -123,7 +123,8 @@ func CompletionEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fibe
 		for k, i := range config.PromptStrings {
 			// A model can have a "file.bin.tmpl" file associated with a prompt template prefix
 			templatedInput, err := o.Loader.EvaluateTemplateForPrompt(model.CompletionPromptTemplate, templateFile, model.PromptTemplateData{
-				Input: i,
+				SystemPrompt: config.SystemPrompt,
+				Input:        i,
 			})
 			if err == nil {
 				i = templatedInput

api/openai/edit.go

@@ -35,8 +35,9 @@ func EditEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx)
 		for _, i := range config.InputStrings {
 			// A model can have a "file.bin.tmpl" file associated with a prompt template prefix
 			templatedInput, err := o.Loader.EvaluateTemplateForPrompt(model.EditPromptTemplate, templateFile, model.PromptTemplateData{
-				Input:       i,
-				Instruction: input.Instruction,
+				Input:        i,
+				Instruction:  input.Instruction,
+				SystemPrompt: config.SystemPrompt,
 			})
 			if err == nil {
 				i = templatedInput

api/openai/image.go

@@ -4,7 +4,6 @@ import (
 	"encoding/base64"
 	"encoding/json"
 	"fmt"
-	"io/ioutil"
 	"os"
 	"path/filepath"
 	"strconv"
@@ -105,7 +104,7 @@ func ImageEndpoint(cm *config.ConfigLoader, o *options.Option) func(c *fiber.Ctx
 					tempDir = o.ImageDir
 				}
 				// Create a temporary file
-				outputFile, err := ioutil.TempFile(tempDir, "b64")
+				outputFile, err := os.CreateTemp(tempDir, "b64")
 				if err != nil {
 					return err
 				}

api/openai/list.go

@@ -1,6 +1,8 @@
 package openai
 
 import (
+	"regexp"
+
 	config "github.com/go-skynet/LocalAI/api/config"
 	model "github.com/go-skynet/LocalAI/pkg/model"
 	"github.com/gofiber/fiber/v2"
@@ -15,14 +17,43 @@ func ListModelsEndpoint(loader *model.ModelLoader, cm *config.ConfigLoader) func
 		var mm map[string]interface{} = map[string]interface{}{}
 
 		dataModels := []OpenAIModel{}
-		for _, m := range models {
-			mm[m] = nil
-			dataModels = append(dataModels, OpenAIModel{ID: m, Object: "model"})
+
+		var filterFn func(name string) bool
+		filter := c.Query("filter")
+
+		// If filter is not specified, do not filter the list by model name
+		if filter == "" {
+			filterFn = func(_ string) bool { return true }
+		} else {
+			// If filter _IS_ specified, we compile it to a regex which is used to create the filterFn
+			rxp, err := regexp.Compile(filter)
+			if err != nil {
+				return err
+			}
+			filterFn = func(name string) bool {
+				return rxp.MatchString(name)
+			}
 		}
 
-		for _, k := range cm.ListConfigs() {
-			if _, exists := mm[k]; !exists {
-				dataModels = append(dataModels, OpenAIModel{ID: k, Object: "model"})
+		// By default, exclude any loose files that are already referenced by a configuration file.
+		excludeConfigured := c.QueryBool("excludeConfigured", true)
+
+		// Start with the known configurations
+		for _, c := range cm.GetAllConfigs() {
+			if excludeConfigured {
+				mm[c.Model] = nil
+			}
+
+			if filterFn(c.Name) {
+				dataModels = append(dataModels, OpenAIModel{ID: c.Name, Object: "model"})
+			}
+		}
+
+		// Then iterate through the loose files:
+		for _, m := range models {
+			// And only adds them if they shouldn't be skipped.
+			if _, exists := mm[m]; !exists && filterFn(m) {
+				dataModels = append(dataModels, OpenAIModel{ID: m, Object: "model"})
 			}
 		}
 

examples/k8sgpt/README.md

@@ -38,7 +38,7 @@ helm install local-ai go-skynet/local-ai --create-namespace --namespace local-ai
 # Install k8sgpt
 helm repo add k8sgpt https://charts.k8sgpt.ai/
 helm repo update
-helm install release k8sgpt/k8sgpt-operator -n k8sgpt-operator-system --create-namespace
+helm install release k8sgpt/k8sgpt-operator -n k8sgpt-operator-system --create-namespace --version 0.0.17
 ```
 
 Apply the k8sgpt-operator configuration:
@@ -55,7 +55,6 @@ spec:
   baseUrl: http://local-ai.local-ai.svc.cluster.local:8080/v1
   noCache: false
   model: gpt-3.5-turbo
-  noCache: false
   version: v0.3.0
   enableAI: true
 EOF
@@ -67,4 +66,7 @@ Apply a broken pod:
 
 ```
 kubectl apply -f broken-pod.yaml
-```
+```
+
+## ArgoCD Deployment Example
+[Deploy K8sgpt + localai with Argocd](https://github.com/tyler-harpool/gitops/tree/main/infra/k8gpt)

examples/k8sgpt/values.yaml

@@ -2,12 +2,13 @@ replicaCount: 1
 
 deployment:
   # https://quay.io/repository/go-skynet/local-ai?tab=tags
-  image: quay.io/go-skynet/local-ai:latest
+  image: quay.io/go-skynet/local-ai:v1.23.0
   env:
     threads: 4
     debug: "true"
     context_size: 512
-    preload_models: '[{ "url": "github:go-skynet/model-gallery/wizard.yaml", "name": "gpt-3.5-turbo", "overrides": { "parameters": { "model": "WizardLM-7B-uncensored.ggmlv3.q5_1" }},"files": [ { "uri": "https://huggingface.co//WizardLM-7B-uncensored-GGML/resolve/main/WizardLM-7B-uncensored.ggmlv3.q5_1.bin", "sha256": "d92a509d83a8ea5e08ba4c2dbaf08f29015932dc2accd627ce0665ac72c2bb2b", "filename": "WizardLM-7B-uncensored.ggmlv3.q5_1" }]}]'
+    galleries: '[{"name":"model-gallery", "url":"github:go-skynet/model-gallery/index.yaml"}, {"url": "github:go-skynet/model-gallery/huggingface.yaml","name":"huggingface"}]'
+    preload_models: '[{ "id": "huggingface@thebloke__open-llama-13b-open-instruct-ggml__open-llama-13b-open-instruct.ggmlv3.q3_k_m.bin", "name": "gpt-3.5-turbo", "overrides": { "f16": true, "mmap": true }}]'
   modelsPath: "/models"
 
 resources:

extra/grpc/huggingface/backend_pb2.py

@@ -13,7 +13,7 @@ _sym_db = _symbol_database.Default()
 
 
 
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x0f\n\rHealthMessage\"\x86\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xd9\x02\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\x9e\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\"6\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t2\xeb\x03\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\rbackend.proto\x12\x07\x62\x61\x63kend\"\x0f\n\rHealthMessage\"\x86\x06\n\x0ePredictOptions\x12\x0e\n\x06Prompt\x18\x01 \x01(\t\x12\x0c\n\x04Seed\x18\x02 \x01(\x05\x12\x0f\n\x07Threads\x18\x03 \x01(\x05\x12\x0e\n\x06Tokens\x18\x04 \x01(\x05\x12\x0c\n\x04TopK\x18\x05 \x01(\x05\x12\x0e\n\x06Repeat\x18\x06 \x01(\x05\x12\r\n\x05\x42\x61tch\x18\x07 \x01(\x05\x12\r\n\x05NKeep\x18\x08 \x01(\x05\x12\x13\n\x0bTemperature\x18\t \x01(\x02\x12\x0f\n\x07Penalty\x18\n \x01(\x02\x12\r\n\x05\x46\x31\x36KV\x18\x0b \x01(\x08\x12\x11\n\tDebugMode\x18\x0c \x01(\x08\x12\x13\n\x0bStopPrompts\x18\r \x03(\t\x12\x11\n\tIgnoreEOS\x18\x0e \x01(\x08\x12\x19\n\x11TailFreeSamplingZ\x18\x0f \x01(\x02\x12\x10\n\x08TypicalP\x18\x10 \x01(\x02\x12\x18\n\x10\x46requencyPenalty\x18\x11 \x01(\x02\x12\x17\n\x0fPresencePenalty\x18\x12 \x01(\x02\x12\x10\n\x08Mirostat\x18\x13 \x01(\x05\x12\x13\n\x0bMirostatETA\x18\x14 \x01(\x02\x12\x13\n\x0bMirostatTAU\x18\x15 \x01(\x02\x12\x12\n\nPenalizeNL\x18\x16 \x01(\x08\x12\x11\n\tLogitBias\x18\x17 \x01(\t\x12\r\n\x05MLock\x18\x19 \x01(\x08\x12\x0c\n\x04MMap\x18\x1a \x01(\x08\x12\x16\n\x0ePromptCacheAll\x18\x1b \x01(\x08\x12\x15\n\rPromptCacheRO\x18\x1c \x01(\x08\x12\x0f\n\x07Grammar\x18\x1d \x01(\t\x12\x0f\n\x07MainGPU\x18\x1e \x01(\t\x12\x13\n\x0bTensorSplit\x18\x1f \x01(\t\x12\x0c\n\x04TopP\x18  \x01(\x02\x12\x17\n\x0fPromptCachePath\x18! \x01(\t\x12\r\n\x05\x44\x65\x62ug\x18\" \x01(\x08\x12\x17\n\x0f\x45mbeddingTokens\x18# \x03(\x05\x12\x12\n\nEmbeddings\x18$ \x01(\t\x12\x14\n\x0cRopeFreqBase\x18% \x01(\x02\x12\x15\n\rRopeFreqScale\x18& \x01(\x02\x12\x1b\n\x13NegativePromptScale\x18\' \x01(\x02\x12\x16\n\x0eNegativePrompt\x18( \x01(\t\"\x18\n\x05Reply\x12\x0f\n\x07message\x18\x01 \x01(\x0c\"\xfb\x02\n\x0cModelOptions\x12\r\n\x05Model\x18\x01 \x01(\t\x12\x13\n\x0b\x43ontextSize\x18\x02 \x01(\x05\x12\x0c\n\x04Seed\x18\x03 \x01(\x05\x12\x0e\n\x06NBatch\x18\x04 \x01(\x05\x12\x11\n\tF16Memory\x18\x05 \x01(\x08\x12\r\n\x05MLock\x18\x06 \x01(\x08\x12\x0c\n\x04MMap\x18\x07 \x01(\x08\x12\x11\n\tVocabOnly\x18\x08 \x01(\x08\x12\x0f\n\x07LowVRAM\x18\t \x01(\x08\x12\x12\n\nEmbeddings\x18\n \x01(\x08\x12\x0c\n\x04NUMA\x18\x0b \x01(\x08\x12\x12\n\nNGPULayers\x18\x0c \x01(\x05\x12\x0f\n\x07MainGPU\x18\r \x01(\t\x12\x13\n\x0bTensorSplit\x18\x0e \x01(\t\x12\x0f\n\x07Threads\x18\x0f \x01(\x05\x12\x19\n\x11LibrarySearchPath\x18\x10 \x01(\t\x12\x14\n\x0cRopeFreqBase\x18\x11 \x01(\x02\x12\x15\n\rRopeFreqScale\x18\x12 \x01(\x02\x12\x12\n\nRMSNormEps\x18\x13 \x01(\x02\x12\x0c\n\x04NGQA\x18\x14 \x01(\x05\"*\n\x06Result\x12\x0f\n\x07message\x18\x01 \x01(\t\x12\x0f\n\x07success\x18\x02 \x01(\x08\"%\n\x0f\x45mbeddingResult\x12\x12\n\nembeddings\x18\x01 \x03(\x02\"C\n\x11TranscriptRequest\x12\x0b\n\x03\x64st\x18\x02 \x01(\t\x12\x10\n\x08language\x18\x03 \x01(\t\x12\x0f\n\x07threads\x18\x04 \x01(\r\"N\n\x10TranscriptResult\x12,\n\x08segments\x18\x01 \x03(\x0b\x32\x1a.backend.TranscriptSegment\x12\x0c\n\x04text\x18\x02 \x01(\t\"Y\n\x11TranscriptSegment\x12\n\n\x02id\x18\x01 \x01(\x05\x12\r\n\x05start\x18\x02 \x01(\x03\x12\x0b\n\x03\x65nd\x18\x03 \x01(\x03\x12\x0c\n\x04text\x18\x04 \x01(\t\x12\x0e\n\x06tokens\x18\x05 \x03(\x05\"\x9e\x01\n\x14GenerateImageRequest\x12\x0e\n\x06height\x18\x01 \x01(\x05\x12\r\n\x05width\x18\x02 \x01(\x05\x12\x0c\n\x04mode\x18\x03 \x01(\x05\x12\x0c\n\x04step\x18\x04 \x01(\x05\x12\x0c\n\x04seed\x18\x05 \x01(\x05\x12\x17\n\x0fpositive_prompt\x18\x06 \x01(\t\x12\x17\n\x0fnegative_prompt\x18\x07 \x01(\t\x12\x0b\n\x03\x64st\x18\x08 \x01(\t\"6\n\nTTSRequest\x12\x0c\n\x04text\x18\x01 \x01(\t\x12\r\n\x05model\x18\x02 \x01(\t\x12\x0b\n\x03\x64st\x18\x03 \x01(\t2\xeb\x03\n\x07\x42\x61\x63kend\x12\x32\n\x06Health\x12\x16.backend.HealthMessage\x1a\x0e.backend.Reply\"\x00\x12\x34\n\x07Predict\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x12\x35\n\tLoadModel\x12\x15.backend.ModelOptions\x1a\x0f.backend.Result\"\x00\x12<\n\rPredictStream\x12\x17.backend.PredictOptions\x1a\x0e.backend.Reply\"\x00\x30\x01\x12@\n\tEmbedding\x12\x17.backend.PredictOptions\x1a\x18.backend.EmbeddingResult\"\x00\x12\x41\n\rGenerateImage\x12\x1d.backend.GenerateImageRequest\x1a\x0f.backend.Result\"\x00\x12M\n\x12\x41udioTranscription\x12\x1a.backend.TranscriptRequest\x1a\x19.backend.TranscriptResult\"\x00\x12-\n\x03TTS\x12\x13.backend.TTSRequest\x1a\x0f.backend.Result\"\x00\x42Z\n\x19io.skynet.localai.backendB\x0eLocalAIBackendP\x01Z+github.com/go-skynet/LocalAI/pkg/grpc/protob\x06proto3')
 
 _globals = globals()
 _builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
@@ -29,21 +29,21 @@ if _descriptor._USE_C_DESCRIPTORS == False:
   _globals['_REPLY']._serialized_start=820
   _globals['_REPLY']._serialized_end=844
   _globals['_MODELOPTIONS']._serialized_start=847
-  _globals['_MODELOPTIONS']._serialized_end=1192
-  _globals['_RESULT']._serialized_start=1194
-  _globals['_RESULT']._serialized_end=1236
-  _globals['_EMBEDDINGRESULT']._serialized_start=1238
-  _globals['_EMBEDDINGRESULT']._serialized_end=1275
-  _globals['_TRANSCRIPTREQUEST']._serialized_start=1277
-  _globals['_TRANSCRIPTREQUEST']._serialized_end=1344
-  _globals['_TRANSCRIPTRESULT']._serialized_start=1346
-  _globals['_TRANSCRIPTRESULT']._serialized_end=1424
-  _globals['_TRANSCRIPTSEGMENT']._serialized_start=1426
-  _globals['_TRANSCRIPTSEGMENT']._serialized_end=1515
-  _globals['_GENERATEIMAGEREQUEST']._serialized_start=1518
-  _globals['_GENERATEIMAGEREQUEST']._serialized_end=1676
-  _globals['_TTSREQUEST']._serialized_start=1678
-  _globals['_TTSREQUEST']._serialized_end=1732
-  _globals['_BACKEND']._serialized_start=1735
-  _globals['_BACKEND']._serialized_end=2226
+  _globals['_MODELOPTIONS']._serialized_end=1226
+  _globals['_RESULT']._serialized_start=1228
+  _globals['_RESULT']._serialized_end=1270
+  _globals['_EMBEDDINGRESULT']._serialized_start=1272
+  _globals['_EMBEDDINGRESULT']._serialized_end=1309
+  _globals['_TRANSCRIPTREQUEST']._serialized_start=1311
+  _globals['_TRANSCRIPTREQUEST']._serialized_end=1378
+  _globals['_TRANSCRIPTRESULT']._serialized_start=1380
+  _globals['_TRANSCRIPTRESULT']._serialized_end=1458
+  _globals['_TRANSCRIPTSEGMENT']._serialized_start=1460
+  _globals['_TRANSCRIPTSEGMENT']._serialized_end=1549
+  _globals['_GENERATEIMAGEREQUEST']._serialized_start=1552
+  _globals['_GENERATEIMAGEREQUEST']._serialized_end=1710
+  _globals['_TTSREQUEST']._serialized_start=1712
+  _globals['_TTSREQUEST']._serialized_end=1766
+  _globals['_BACKEND']._serialized_start=1769
+  _globals['_BACKEND']._serialized_end=2260
 # @@protoc_insertion_point(module_scope)

go.mod

@@ -9,7 +9,7 @@ require (
 	github.com/go-skynet/bloomz.cpp v0.0.0-20230529155654-1834e77b83fa
 	github.com/go-skynet/go-bert.cpp v0.0.0-20230716133540-6abe312cded1
 	github.com/go-skynet/go-ggml-transformers.cpp v0.0.0-20230714203132-ffb09d7dd71e
-	github.com/go-skynet/go-llama.cpp v0.0.0-20230727163958-6ba16de8e965
+	github.com/go-skynet/go-llama.cpp v0.0.0-20230802220037-50cee7712066
 	github.com/gofiber/fiber/v2 v2.48.0
 	github.com/google/uuid v1.3.0
 	github.com/hashicorp/go-multierror v1.1.1
@@ -20,14 +20,14 @@ require (
 	github.com/mudler/go-ggllm.cpp v0.0.0-20230709223052-862477d16eef
 	github.com/mudler/go-processmanager v0.0.0-20220724164624-c45b5c61312d
 	github.com/mudler/go-stable-diffusion v0.0.0-20230605122230-d89260f598af
-	github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230727161923-39acbc837816
+	github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230802145814-c449b71b56de
 	github.com/onsi/ginkgo/v2 v2.11.0
 	github.com/onsi/gomega v1.27.10
 	github.com/otiai10/openaigo v1.5.2
 	github.com/phayes/freeport v0.0.0-20220201140144-74d24b5ae9f5
-	github.com/rs/zerolog v1.29.1
+	github.com/rs/zerolog v1.30.0
 	github.com/sashabaranov/go-openai v1.14.1
-	github.com/tmc/langchaingo v0.0.0-20230726025230-7d5f9fd5e90a
+	github.com/tmc/langchaingo v0.0.0-20230802030916-271e9bd7e7c5
 	github.com/urfave/cli/v2 v2.25.7
 	github.com/valyala/fasthttp v1.48.0
 	google.golang.org/grpc v1.57.0

go.sum

@@ -45,6 +45,10 @@ github.com/go-skynet/go-llama.cpp v0.0.0-20230724222459-562d2b5a7119 h1:FeUSk5yM
 github.com/go-skynet/go-llama.cpp v0.0.0-20230724222459-562d2b5a7119/go.mod h1:fiJBto+Le1XLtD/cID5SAKs8cKE7wFXJKfTT3wvPQRA=
 github.com/go-skynet/go-llama.cpp v0.0.0-20230727163958-6ba16de8e965 h1:2MO/rABKpkXnnKQ3Ar90aqhnlMEejE9gnKG6bafv+ow=
 github.com/go-skynet/go-llama.cpp v0.0.0-20230727163958-6ba16de8e965/go.mod h1:fiJBto+Le1XLtD/cID5SAKs8cKE7wFXJKfTT3wvPQRA=
+github.com/go-skynet/go-llama.cpp v0.0.0-20230729200103-8c51308e42d7 h1:1uBwholTaJ8Lva8ySJjT4jNaCDAh+MJXtsbZBbQq9lA=
+github.com/go-skynet/go-llama.cpp v0.0.0-20230729200103-8c51308e42d7/go.mod h1:fiJBto+Le1XLtD/cID5SAKs8cKE7wFXJKfTT3wvPQRA=
+github.com/go-skynet/go-llama.cpp v0.0.0-20230802220037-50cee7712066 h1:v4Js+yEdgY9IV7n35M+5MELLxlOMp3qC5whZm5YTLjI=
+github.com/go-skynet/go-llama.cpp v0.0.0-20230802220037-50cee7712066/go.mod h1:fiJBto+Le1XLtD/cID5SAKs8cKE7wFXJKfTT3wvPQRA=
 github.com/go-task/slim-sprig v0.0.0-20210107165309-348f09dbbbc0/go.mod h1:fyg7847qk6SyHyPtNmDHnmrv/HOrqktSC+C9fM+CJOE=
 github.com/go-task/slim-sprig v0.0.0-20230315185526-52ccab3ef572 h1:tfuBGBXKqDEevZMzYi5KSi8KkcZtzBcTgAUUtapy0OI=
 github.com/go-task/slim-sprig v0.0.0-20230315185526-52ccab3ef572/go.mod h1:9Pwr4B2jHnOSGXyyzV8ROjYa2ojvAY6HCGYYfMoC3Ls=
@@ -128,6 +132,10 @@ github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230725212419-9100b2
 github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230725212419-9100b2ef6fb9/go.mod h1:4T3CHXyrt+7FQHXaxULZfPjHbD8/99WuDDJa0YVZARI=
 github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230727161923-39acbc837816 h1:hRi7hpDUuaO0dB4NZ8eyaeD2fRar6CPyNAARsO5DhzA=
 github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230727161923-39acbc837816/go.mod h1:4T3CHXyrt+7FQHXaxULZfPjHbD8/99WuDDJa0YVZARI=
+github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230731161838-cbdcde8b7586 h1:WVEMSZMyHFe68PN204c3Fdk5g2lZouPvbU9/2zkPpWc=
+github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230731161838-cbdcde8b7586/go.mod h1:4T3CHXyrt+7FQHXaxULZfPjHbD8/99WuDDJa0YVZARI=
+github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230802145814-c449b71b56de h1:E5EGczxEAcbaO8yqj074MQxU609QbtB6in3qTOW1EFo=
+github.com/nomic-ai/gpt4all/gpt4all-bindings/golang v0.0.0-20230802145814-c449b71b56de/go.mod h1:4T3CHXyrt+7FQHXaxULZfPjHbD8/99WuDDJa0YVZARI=
 github.com/nwaples/rardecode v1.1.0 h1:vSxaY8vQhOcVr4mm5e8XllHWTiM4JF507A0Katqw7MQ=
 github.com/nwaples/rardecode v1.1.0/go.mod h1:5DzqNKiOdpKKBH87u8VlvAnPZMXcGRhxWkRpHbbfGS0=
 github.com/nxadm/tail v1.4.4/go.mod h1:kenIhsEOeOJmVchQTgglprH7qJGnHDVpk1VPCcaMI8A=
@@ -161,8 +169,11 @@ github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZN
 github.com/rivo/uniseg v0.2.0 h1:S1pD9weZBuJdFmowNwbpi7BJ8TNftyUImj/0WQi72jY=
 github.com/rivo/uniseg v0.2.0/go.mod h1:J6wj4VEh+S6ZtnVlnTBMWIodfgj8LQOQFoIToxlJtxc=
 github.com/rs/xid v1.4.0/go.mod h1:trrq9SKmegXys3aeAKXMUTdJsYXVwGY3RLcfgqegfbg=
+github.com/rs/xid v1.5.0/go.mod h1:trrq9SKmegXys3aeAKXMUTdJsYXVwGY3RLcfgqegfbg=
 github.com/rs/zerolog v1.29.1 h1:cO+d60CHkknCbvzEWxP0S9K6KqyTjrCNUy1LdQLCGPc=
 github.com/rs/zerolog v1.29.1/go.mod h1:Le6ESbR7hc+DP6Lt1THiV8CQSdkkNrd3R0XbEgp3ZBU=
+github.com/rs/zerolog v1.30.0 h1:SymVODrcRsaRaSInD9yQtKbtWqwsfoPcRff/oRXLj4c=
+github.com/rs/zerolog v1.30.0/go.mod h1:/tk+P47gFdPXq4QYjvCmT5/Gsug2nagsFWBWhAiSi1w=
 github.com/russross/blackfriday/v2 v2.1.0 h1:JIOH55/0cWyOuilr9/qlrm0BSXldqnqwMsf35Ld67mk=
 github.com/russross/blackfriday/v2 v2.1.0/go.mod h1:+Rmxgy9KzJVeS9/2gXHxylqXiyQDYRxCVz55jmeOWTM=
 github.com/sashabaranov/go-openai v1.14.0 h1:D1yAB+DHElgbJFdYyjxfTWMFzhddn+PwZmkQ039L7mQ=
@@ -178,6 +189,12 @@ github.com/tmc/langchaingo v0.0.0-20230713201705-dcf7ecdc8ac8 h1:wdJigYmmIRCuXhC
 github.com/tmc/langchaingo v0.0.0-20230713201705-dcf7ecdc8ac8/go.mod h1:mTzgQfAGwmBz2hhQELZfu2bwsbHwyKHA6IHOa+9LDFg=
 github.com/tmc/langchaingo v0.0.0-20230726025230-7d5f9fd5e90a h1:I/2JSuYXkWaVVLSZmrPfrgbvvvPR0IaulZcB0Iu8oVI=
 github.com/tmc/langchaingo v0.0.0-20230726025230-7d5f9fd5e90a/go.mod h1:8T+nNIGBr3nYQEYFmF/YaT8t8YTKLvFYZBuVZOAYn5E=
+github.com/tmc/langchaingo v0.0.0-20230729232647-7df4fe5fb8fe h1:+XVrCjh3rPibfISkUFG2Ck5NLKODQ9cFdmraFye1bGA=
+github.com/tmc/langchaingo v0.0.0-20230729232647-7df4fe5fb8fe/go.mod h1:8T+nNIGBr3nYQEYFmF/YaT8t8YTKLvFYZBuVZOAYn5E=
+github.com/tmc/langchaingo v0.0.0-20230731024823-8f101609f600 h1:SABuIthjhIXEsxnokuA16CZOxxdW9XohIHQqd/go8Nc=
+github.com/tmc/langchaingo v0.0.0-20230731024823-8f101609f600/go.mod h1:8T+nNIGBr3nYQEYFmF/YaT8t8YTKLvFYZBuVZOAYn5E=
+github.com/tmc/langchaingo v0.0.0-20230802030916-271e9bd7e7c5 h1:js7vYDJGzUGVSt0YlIusUc5BXYVECu3LUI/asby5Ggo=
+github.com/tmc/langchaingo v0.0.0-20230802030916-271e9bd7e7c5/go.mod h1:8T+nNIGBr3nYQEYFmF/YaT8t8YTKLvFYZBuVZOAYn5E=
 github.com/ulikunitz/xz v0.5.8/go.mod h1:nbz6k7qbPmH4IRqmfOplQw/tblSgqTqBwxkY0oWt/14=
 github.com/ulikunitz/xz v0.5.9 h1:RsKRIA2MO8x56wkkcd3LbtcE/uMszhb6DpRf+3uwa3I=
 github.com/ulikunitz/xz v0.5.9/go.mod h1:nbz6k7qbPmH4IRqmfOplQw/tblSgqTqBwxkY0oWt/14=

pkg/gallery/gallery.go

@@ -19,6 +19,8 @@ type Gallery struct {
 // Installs a model from the gallery (galleryname@modelname)
 func InstallModelFromGallery(galleries []Gallery, name string, basePath string, req GalleryModel, downloadStatus func(string, string, string, float64)) error {
 	applyModel := func(model *GalleryModel) error {
+		name = strings.ReplaceAll(name, string(os.PathSeparator), "__")
+
 		config, err := GetGalleryConfigFromURL(model.URL)
 		if err != nil {
 			return err
@@ -51,7 +53,11 @@ func InstallModelFromGallery(galleries []Gallery, name string, basePath string,
 
 	model, err := FindGallery(models, name)
 	if err != nil {
-		return err
+		var err2 error
+		model, err2 = FindGallery(models, strings.ToLower(name))
+		if err2 != nil {
+			return err
+		}
 	}
 
 	return applyModel(model)
@@ -79,7 +85,7 @@ func InstallModelFromGalleryByName(galleries []Gallery, name string, basePath st
 	name = strings.ReplaceAll(name, string(os.PathSeparator), "__")
 	var model *GalleryModel
 	for _, m := range models {
-		if name == m.Name {
+		if name == m.Name || name == strings.ToLower(m.Name) {
 			model = m
 		}
 	}

pkg/gallery/models_test.go

@@ -1,7 +1,6 @@
 package gallery_test
 
 import (
-	"io/ioutil"
 	"os"
 	"path/filepath"
 
@@ -50,7 +49,7 @@ var _ = Describe("Model test", func() {
 			}}
 			out, err := yaml.Marshal(gallery)
 			Expect(err).ToNot(HaveOccurred())
-			err = ioutil.WriteFile(filepath.Join(tempdir, "gallery_simple.yaml"), out, 0644)
+			err = os.WriteFile(filepath.Join(tempdir, "gallery_simple.yaml"), out, 0644)
 			Expect(err).ToNot(HaveOccurred())
 
 			galleries := []Gallery{

pkg/grammar/json_schema.go

@@ -15,10 +15,13 @@ var (
 
 	PRIMITIVE_RULES = map[string]string{
 		"boolean": `("true" | "false") space`,
-		"number":  `[0-9]+ space`,                    // TODO complete
-		"integer": `[0-9]+ space`,                    // TODO complete
-		"string":  `"\"" [ \t!#-\[\]-~]* "\"" space`, // TODO complete
-		"null":    `"null" space`,
+		"number":  `("-"? ([0-9] | [1-9] [0-9]*)) ("." [0-9]+)? ([eE] [-+]? [0-9]+)? space`,
+		"integer": `("-"? ([0-9] | [1-9] [0-9]*)) space`,
+		"string": `"\"" (
+			[^"\\] |
+			"\\" (["\\/bfnrt] | "u" [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F])
+		  )* "\"" space`,
+		"null": `"null" space`,
 	}
 
 	INVALID_RULE_CHARS_RE     = regexp.MustCompile(`[^a-zA-Z0-9-]+`)
@@ -176,6 +179,9 @@ func (sc *JSONSchemaConverter) visit(schema map[string]interface{}, name string,
 		if !exists {
 			panic(fmt.Sprintf("Unrecognized schema: %v", schema))
 		}
+		if ruleName == "root" {
+			schemaType = "root"
+		}
 		return sc.addRule(schemaType, primitiveRule)
 	}
 }

pkg/grammar/json_schema_test.go

@@ -48,7 +48,10 @@ root ::= root-0 | root-1
 space ::= " "?
 root-0-arguments ::= "{" space "\"date\"" space ":" space string "," space "\"time\"" space ":" space string "," space "\"title\"" space ":" space string "}" space
 root-1 ::= "{" space "\"arguments\"" space ":" space root-1-arguments "," space "\"function\"" space ":" space root-1-function "}" space
-string ::= "\"" [ \t!#-\[\]-~]* "\"" space
+string ::= "\"" (
+	[^"\\] |
+	"\\" (["\\/bfnrt] | "u" [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F] [0-9a-fA-F])
+)* "\"" space
 root-1-function ::= "\"search\""`
 )
 

pkg/grpc/llm/llama/llama.go

@@ -33,6 +33,14 @@ func (llm *LLM) Load(opts *pb.ModelOptions) error {
 		llama.WithRopeFreqScale(ropeFreqScale),
 	}
 
+	if opts.NGQA != 0 {
+		llamaOpts = append(llamaOpts, llama.WithGQA(int(opts.NGQA)))
+	}
+
+	if opts.RMSNormEps != 0 {
+		llamaOpts = append(llamaOpts, llama.WithRMSNormEPS(opts.RMSNormEps))
+	}
+
 	if opts.ContextSize != 0 {
 		llamaOpts = append(llamaOpts, llama.SetContext(int(opts.ContextSize)))
 	}

pkg/grpc/proto/backend.pb.go

@@ -481,6 +481,8 @@ type ModelOptions struct {
 	LibrarySearchPath string  `protobuf:"bytes,16,opt,name=LibrarySearchPath,proto3" json:"LibrarySearchPath,omitempty"`
 	RopeFreqBase      float32 `protobuf:"fixed32,17,opt,name=RopeFreqBase,proto3" json:"RopeFreqBase,omitempty"`
 	RopeFreqScale     float32 `protobuf:"fixed32,18,opt,name=RopeFreqScale,proto3" json:"RopeFreqScale,omitempty"`
+	RMSNormEps        float32 `protobuf:"fixed32,19,opt,name=RMSNormEps,proto3" json:"RMSNormEps,omitempty"`
+	NGQA              int32   `protobuf:"varint,20,opt,name=NGQA,proto3" json:"NGQA,omitempty"`
 }
 
 func (x *ModelOptions) Reset() {
@@ -641,6 +643,20 @@ func (x *ModelOptions) GetRopeFreqScale() float32 {
 	return 0
 }
 
+func (x *ModelOptions) GetRMSNormEps() float32 {
+	if x != nil {
+		return x.RMSNormEps
+	}
+	return 0
+}
+
+func (x *ModelOptions) GetNGQA() int32 {
+	if x != nil {
+		return x.NGQA
+	}
+	return 0
+}
+
 type Result struct {
 	state         protoimpl.MessageState
 	sizeCache     protoimpl.SizeCache
@@ -1191,7 +1207,7 @@ var file_pkg_grpc_proto_backend_proto_rawDesc = []byte{
 	0x0e, 0x4e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x22,
 	0x21, 0x0a, 0x05, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x12, 0x18, 0x0a, 0x07, 0x6d, 0x65, 0x73, 0x73,
 	0x61, 0x67, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x0c, 0x52, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61,
-	0x67, 0x65, 0x22, 0x94, 0x04, 0x0a, 0x0c, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69,
+	0x67, 0x65, 0x22, 0xc8, 0x04, 0x0a, 0x0c, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69,
 	0x6f, 0x6e, 0x73, 0x12, 0x14, 0x0a, 0x05, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x18, 0x01, 0x20, 0x01,
 	0x28, 0x09, 0x52, 0x05, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x20, 0x0a, 0x0b, 0x43, 0x6f, 0x6e,
 	0x74, 0x65, 0x78, 0x74, 0x53, 0x69, 0x7a, 0x65, 0x18, 0x02, 0x20, 0x01, 0x28, 0x05, 0x52, 0x0b,
@@ -1224,90 +1240,93 @@ var file_pkg_grpc_proto_backend_proto_rawDesc = []byte{
 	0x11, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0c, 0x52, 0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x42,
 	0x61, 0x73, 0x65, 0x12, 0x24, 0x0a, 0x0d, 0x52, 0x6f, 0x70, 0x65, 0x46, 0x72, 0x65, 0x71, 0x53,
 	0x63, 0x61, 0x6c, 0x65, 0x18, 0x12, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0d, 0x52, 0x6f, 0x70, 0x65,
-	0x46, 0x72, 0x65, 0x71, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x22, 0x3c, 0x0a, 0x06, 0x52, 0x65, 0x73,
-	0x75, 0x6c, 0x74, 0x12, 0x18, 0x0a, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x18, 0x01,
-	0x20, 0x01, 0x28, 0x09, 0x52, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x12, 0x18, 0x0a,
-	0x07, 0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x18, 0x02, 0x20, 0x01, 0x28, 0x08, 0x52, 0x07,
-	0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x22, 0x31, 0x0a, 0x0f, 0x45, 0x6d, 0x62, 0x65, 0x64,
-	0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12, 0x1e, 0x0a, 0x0a, 0x65, 0x6d,
-	0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, 0x18, 0x01, 0x20, 0x03, 0x28, 0x02, 0x52, 0x0a,
-	0x65, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, 0x22, 0x5b, 0x0a, 0x11, 0x54, 0x72,
-	0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12,
-	0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73,
-	0x74, 0x12, 0x1a, 0x0a, 0x08, 0x6c, 0x61, 0x6e, 0x67, 0x75, 0x61, 0x67, 0x65, 0x18, 0x03, 0x20,
-	0x01, 0x28, 0x09, 0x52, 0x08, 0x6c, 0x61, 0x6e, 0x67, 0x75, 0x61, 0x67, 0x65, 0x12, 0x18, 0x0a,
-	0x07, 0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x18, 0x04, 0x20, 0x01, 0x28, 0x0d, 0x52, 0x07,
-	0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x22, 0x5e, 0x0a, 0x10, 0x54, 0x72, 0x61, 0x6e, 0x73,
-	0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12, 0x36, 0x0a, 0x08, 0x73,
-	0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x73, 0x18, 0x01, 0x20, 0x03, 0x28, 0x0b, 0x32, 0x1a, 0x2e,
-	0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69,
-	0x70, 0x74, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, 0x08, 0x73, 0x65, 0x67, 0x6d, 0x65,
-	0x6e, 0x74, 0x73, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28,
-	0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x22, 0x77, 0x0a, 0x11, 0x54, 0x72, 0x61, 0x6e, 0x73,
-	0x63, 0x72, 0x69, 0x70, 0x74, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x12, 0x0e, 0x0a, 0x02,
-	0x69, 0x64, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52, 0x02, 0x69, 0x64, 0x12, 0x14, 0x0a, 0x05,
-	0x73, 0x74, 0x61, 0x72, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, 0x03, 0x52, 0x05, 0x73, 0x74, 0x61,
-	0x72, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x65, 0x6e, 0x64, 0x18, 0x03, 0x20, 0x01, 0x28, 0x03, 0x52,
-	0x03, 0x65, 0x6e, 0x64, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, 0x74, 0x18, 0x04, 0x20, 0x01,
-	0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12, 0x16, 0x0a, 0x06, 0x74, 0x6f, 0x6b, 0x65,
-	0x6e, 0x73, 0x18, 0x05, 0x20, 0x03, 0x28, 0x05, 0x52, 0x06, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73,
-	0x22, 0xe4, 0x01, 0x0a, 0x14, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61,
-	0x67, 0x65, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x16, 0x0a, 0x06, 0x68, 0x65, 0x69,
-	0x67, 0x68, 0x74, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52, 0x06, 0x68, 0x65, 0x69, 0x67, 0x68,
-	0x74, 0x12, 0x14, 0x0a, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x18, 0x02, 0x20, 0x01, 0x28, 0x05,
-	0x52, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x12, 0x12, 0x0a, 0x04, 0x6d, 0x6f, 0x64, 0x65, 0x18,
-	0x03, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x6d, 0x6f, 0x64, 0x65, 0x12, 0x12, 0x0a, 0x04, 0x73,
-	0x74, 0x65, 0x70, 0x18, 0x04, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x73, 0x74, 0x65, 0x70, 0x12,
-	0x12, 0x0a, 0x04, 0x73, 0x65, 0x65, 0x64, 0x18, 0x05, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x73,
-	0x65, 0x65, 0x64, 0x12, 0x27, 0x0a, 0x0f, 0x70, 0x6f, 0x73, 0x69, 0x74, 0x69, 0x76, 0x65, 0x5f,
-	0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18, 0x06, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0e, 0x70, 0x6f,
-	0x73, 0x69, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x12, 0x27, 0x0a, 0x0f,
-	0x6e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x5f, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18,
-	0x07, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0e, 0x6e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x50,
-	0x72, 0x6f, 0x6d, 0x70, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x08, 0x20, 0x01,
-	0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x22, 0x48, 0x0a, 0x0a, 0x54, 0x54, 0x53, 0x52, 0x65,
-	0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, 0x74, 0x18, 0x01, 0x20,
-	0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12, 0x14, 0x0a, 0x05, 0x6d, 0x6f, 0x64,
-	0x65, 0x6c, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x05, 0x6d, 0x6f, 0x64, 0x65, 0x6c, 0x12,
-	0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x03, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73,
-	0x74, 0x32, 0xeb, 0x03, 0x0a, 0x07, 0x42, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x12, 0x32, 0x0a,
-	0x06, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x12, 0x16, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
-	0x64, 0x2e, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x1a,
-	0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22,
-	0x00, 0x12, 0x34, 0x0a, 0x07, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x12, 0x17, 0x2e, 0x62,
-	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70,
-	0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e,
-	0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x35, 0x0a, 0x09, 0x4c, 0x6f, 0x61, 0x64, 0x4d,
-	0x6f, 0x64, 0x65, 0x6c, 0x12, 0x15, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x4d,
-	0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0f, 0x2e, 0x62, 0x61,
-	0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x3c,
-	0x0a, 0x0d, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x53, 0x74, 0x72, 0x65, 0x61, 0x6d, 0x12,
-	0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63,
-	0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65,
-	0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x30, 0x01, 0x12, 0x40, 0x0a, 0x09,
-	0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b,
-	0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f,
-	0x6e, 0x73, 0x1a, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x45, 0x6d, 0x62,
-	0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x41,
-	0x0a, 0x0d, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x12,
-	0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61,
-	0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x0f,
-	0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22,
-	0x00, 0x12, 0x4d, 0x0a, 0x12, 0x41, 0x75, 0x64, 0x69, 0x6f, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63,
-	0x72, 0x69, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x12, 0x1a, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
-	0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x71, 0x75,
-	0x65, 0x73, 0x74, 0x1a, 0x19, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72,
-	0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00,
-	0x12, 0x2d, 0x0a, 0x03, 0x54, 0x54, 0x53, 0x12, 0x13, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
-	0x64, 0x2e, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x0f, 0x2e, 0x62,
-	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x42,
-	0x5a, 0x0a, 0x19, 0x69, 0x6f, 0x2e, 0x73, 0x6b, 0x79, 0x6e, 0x65, 0x74, 0x2e, 0x6c, 0x6f, 0x63,
-	0x61, 0x6c, 0x61, 0x69, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x42, 0x0e, 0x4c, 0x6f,
-	0x63, 0x61, 0x6c, 0x41, 0x49, 0x42, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x50, 0x01, 0x5a, 0x2b,
-	0x67, 0x69, 0x74, 0x68, 0x75, 0x62, 0x2e, 0x63, 0x6f, 0x6d, 0x2f, 0x67, 0x6f, 0x2d, 0x73, 0x6b,
-	0x79, 0x6e, 0x65, 0x74, 0x2f, 0x4c, 0x6f, 0x63, 0x61, 0x6c, 0x41, 0x49, 0x2f, 0x70, 0x6b, 0x67,
-	0x2f, 0x67, 0x72, 0x70, 0x63, 0x2f, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x62, 0x06, 0x70, 0x72, 0x6f,
-	0x74, 0x6f, 0x33,
+	0x46, 0x72, 0x65, 0x71, 0x53, 0x63, 0x61, 0x6c, 0x65, 0x12, 0x1e, 0x0a, 0x0a, 0x52, 0x4d, 0x53,
+	0x4e, 0x6f, 0x72, 0x6d, 0x45, 0x70, 0x73, 0x18, 0x13, 0x20, 0x01, 0x28, 0x02, 0x52, 0x0a, 0x52,
+	0x4d, 0x53, 0x4e, 0x6f, 0x72, 0x6d, 0x45, 0x70, 0x73, 0x12, 0x12, 0x0a, 0x04, 0x4e, 0x47, 0x51,
+	0x41, 0x18, 0x14, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x4e, 0x47, 0x51, 0x41, 0x22, 0x3c, 0x0a,
+	0x06, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12, 0x18, 0x0a, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61,
+	0x67, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x07, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+	0x65, 0x12, 0x18, 0x0a, 0x07, 0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x18, 0x02, 0x20, 0x01,
+	0x28, 0x08, 0x52, 0x07, 0x73, 0x75, 0x63, 0x63, 0x65, 0x73, 0x73, 0x22, 0x31, 0x0a, 0x0f, 0x45,
+	0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12, 0x1e,
+	0x0a, 0x0a, 0x65, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, 0x18, 0x01, 0x20, 0x03,
+	0x28, 0x02, 0x52, 0x0a, 0x65, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x73, 0x22, 0x5b,
+	0x0a, 0x11, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x71, 0x75,
+	0x65, 0x73, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09,
+	0x52, 0x03, 0x64, 0x73, 0x74, 0x12, 0x1a, 0x0a, 0x08, 0x6c, 0x61, 0x6e, 0x67, 0x75, 0x61, 0x67,
+	0x65, 0x18, 0x03, 0x20, 0x01, 0x28, 0x09, 0x52, 0x08, 0x6c, 0x61, 0x6e, 0x67, 0x75, 0x61, 0x67,
+	0x65, 0x12, 0x18, 0x0a, 0x07, 0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x18, 0x04, 0x20, 0x01,
+	0x28, 0x0d, 0x52, 0x07, 0x74, 0x68, 0x72, 0x65, 0x61, 0x64, 0x73, 0x22, 0x5e, 0x0a, 0x10, 0x54,
+	0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x12,
+	0x36, 0x0a, 0x08, 0x73, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x73, 0x18, 0x01, 0x20, 0x03, 0x28,
+	0x0b, 0x32, 0x1a, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e,
+	0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x52, 0x08, 0x73,
+	0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74, 0x73, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, 0x74, 0x18,
+	0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x22, 0x77, 0x0a, 0x11, 0x54,
+	0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x53, 0x65, 0x67, 0x6d, 0x65, 0x6e, 0x74,
+	0x12, 0x0e, 0x0a, 0x02, 0x69, 0x64, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52, 0x02, 0x69, 0x64,
+	0x12, 0x14, 0x0a, 0x05, 0x73, 0x74, 0x61, 0x72, 0x74, 0x18, 0x02, 0x20, 0x01, 0x28, 0x03, 0x52,
+	0x05, 0x73, 0x74, 0x61, 0x72, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x65, 0x6e, 0x64, 0x18, 0x03, 0x20,
+	0x01, 0x28, 0x03, 0x52, 0x03, 0x65, 0x6e, 0x64, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78, 0x74,
+	0x18, 0x04, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12, 0x16, 0x0a, 0x06,
+	0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x18, 0x05, 0x20, 0x03, 0x28, 0x05, 0x52, 0x06, 0x74, 0x6f,
+	0x6b, 0x65, 0x6e, 0x73, 0x22, 0xe4, 0x01, 0x0a, 0x14, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74,
+	0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x16, 0x0a,
+	0x06, 0x68, 0x65, 0x69, 0x67, 0x68, 0x74, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52, 0x06, 0x68,
+	0x65, 0x69, 0x67, 0x68, 0x74, 0x12, 0x14, 0x0a, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x18, 0x02,
+	0x20, 0x01, 0x28, 0x05, 0x52, 0x05, 0x77, 0x69, 0x64, 0x74, 0x68, 0x12, 0x12, 0x0a, 0x04, 0x6d,
+	0x6f, 0x64, 0x65, 0x18, 0x03, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x6d, 0x6f, 0x64, 0x65, 0x12,
+	0x12, 0x0a, 0x04, 0x73, 0x74, 0x65, 0x70, 0x18, 0x04, 0x20, 0x01, 0x28, 0x05, 0x52, 0x04, 0x73,
+	0x74, 0x65, 0x70, 0x12, 0x12, 0x0a, 0x04, 0x73, 0x65, 0x65, 0x64, 0x18, 0x05, 0x20, 0x01, 0x28,
+	0x05, 0x52, 0x04, 0x73, 0x65, 0x65, 0x64, 0x12, 0x27, 0x0a, 0x0f, 0x70, 0x6f, 0x73, 0x69, 0x74,
+	0x69, 0x76, 0x65, 0x5f, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x18, 0x06, 0x20, 0x01, 0x28, 0x09,
+	0x52, 0x0e, 0x70, 0x6f, 0x73, 0x69, 0x74, 0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74,
+	0x12, 0x27, 0x0a, 0x0f, 0x6e, 0x65, 0x67, 0x61, 0x74, 0x69, 0x76, 0x65, 0x5f, 0x70, 0x72, 0x6f,
+	0x6d, 0x70, 0x74, 0x18, 0x07, 0x20, 0x01, 0x28, 0x09, 0x52, 0x0e, 0x6e, 0x65, 0x67, 0x61, 0x74,
+	0x69, 0x76, 0x65, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74,
+	0x18, 0x08, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x22, 0x48, 0x0a, 0x0a, 0x54,
+	0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x12, 0x0a, 0x04, 0x74, 0x65, 0x78,
+	0x74, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65, 0x78, 0x74, 0x12, 0x14, 0x0a,
+	0x05, 0x6d, 0x6f, 0x64, 0x65, 0x6c, 0x18, 0x02, 0x20, 0x01, 0x28, 0x09, 0x52, 0x05, 0x6d, 0x6f,
+	0x64, 0x65, 0x6c, 0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18, 0x03, 0x20, 0x01, 0x28, 0x09,
+	0x52, 0x03, 0x64, 0x73, 0x74, 0x32, 0xeb, 0x03, 0x0a, 0x07, 0x42, 0x61, 0x63, 0x6b, 0x65, 0x6e,
+	0x64, 0x12, 0x32, 0x0a, 0x06, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x12, 0x16, 0x2e, 0x62, 0x61,
+	0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, 0x73, 0x73,
+	0x61, 0x67, 0x65, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65,
+	0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x34, 0x0a, 0x07, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74,
+	0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69,
+	0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b,
+	0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x35, 0x0a, 0x09, 0x4c,
+	0x6f, 0x61, 0x64, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x15, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65,
+	0x6e, 0x64, 0x2e, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a,
+	0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74,
+	0x22, 0x00, 0x12, 0x3c, 0x0a, 0x0d, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x53, 0x74, 0x72,
+	0x65, 0x61, 0x6d, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72,
+	0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62,
+	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x30, 0x01,
+	0x12, 0x40, 0x0a, 0x09, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x12, 0x17, 0x2e,
+	0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f,
+	0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64,
+	0x2e, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74,
+	0x22, 0x00, 0x12, 0x41, 0x0a, 0x0d, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d,
+	0x61, 0x67, 0x65, 0x12, 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x47, 0x65,
+	0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71, 0x75, 0x65,
+	0x73, 0x74, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73,
+	0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x4d, 0x0a, 0x12, 0x41, 0x75, 0x64, 0x69, 0x6f, 0x54, 0x72,
+	0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x12, 0x1a, 0x2e, 0x62, 0x61,
+	0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74,
+	0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x19, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
+	0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73, 0x75,
+	0x6c, 0x74, 0x22, 0x00, 0x12, 0x2d, 0x0a, 0x03, 0x54, 0x54, 0x53, 0x12, 0x13, 0x2e, 0x62, 0x61,
+	0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74,
+	0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c,
+	0x74, 0x22, 0x00, 0x42, 0x5a, 0x0a, 0x19, 0x69, 0x6f, 0x2e, 0x73, 0x6b, 0x79, 0x6e, 0x65, 0x74,
+	0x2e, 0x6c, 0x6f, 0x63, 0x61, 0x6c, 0x61, 0x69, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64,
+	0x42, 0x0e, 0x4c, 0x6f, 0x63, 0x61, 0x6c, 0x41, 0x49, 0x42, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64,
+	0x50, 0x01, 0x5a, 0x2b, 0x67, 0x69, 0x74, 0x68, 0x75, 0x62, 0x2e, 0x63, 0x6f, 0x6d, 0x2f, 0x67,
+	0x6f, 0x2d, 0x73, 0x6b, 0x79, 0x6e, 0x65, 0x74, 0x2f, 0x4c, 0x6f, 0x63, 0x61, 0x6c, 0x41, 0x49,
+	0x2f, 0x70, 0x6b, 0x67, 0x2f, 0x67, 0x72, 0x70, 0x63, 0x2f, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x62,
+	0x06, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x33,
 }
 
 var (

pkg/grpc/proto/backend.proto

@@ -87,6 +87,8 @@ message ModelOptions {
   string LibrarySearchPath = 16;
   float RopeFreqBase = 17;
   float RopeFreqScale = 18;
+  float RMSNormEps = 19;
+  int32 NGQA = 20;
 }
 
 message Result {

pkg/model/loader.go

@@ -20,10 +20,12 @@ import (
 // These are the definitions of all possible variables LocalAI will currently populate for use in a prompt template file
 // Please note: Not all of these are populated on every endpoint - your template should either be tested for each endpoint you map it to, or tolerant of zero values.
 type PromptTemplateData struct {
-	Input        string
-	Instruction  string
-	Functions    []grammar.Function
-	MessageIndex int
+	SystemPrompt         string
+	SuppressSystemPrompt bool // used by chat specifically to indicate that SystemPrompt above should be _ignored_
+	Input                string
+	Instruction          string
+	Functions            []grammar.Function
+	MessageIndex         int
 }
 
 // TODO: Ask mudler about FunctionCall stuff being useful at the message level?
@@ -85,8 +87,8 @@ func (ml *ModelLoader) ListModels() ([]string, error) {
 
 	models := []string{}
 	for _, file := range files {
-		// Skip templates, YAML and .keep files
-		if strings.HasSuffix(file.Name(), ".tmpl") || strings.HasSuffix(file.Name(), ".keep") || strings.HasSuffix(file.Name(), ".yaml") || strings.HasSuffix(file.Name(), ".yml") {
+		// Skip templates, YAML, .keep, .json, and .DS_Store files - TODO: as this list grows, is there a more efficient method?
+		if strings.HasSuffix(file.Name(), ".tmpl") || strings.HasSuffix(file.Name(), ".keep") || strings.HasSuffix(file.Name(), ".yaml") || strings.HasSuffix(file.Name(), ".yml") || strings.HasSuffix(file.Name(), ".json") || strings.HasSuffix(file.Name(), ".DS_Store") {
 			continue
 		}
 

pkg/utils/json.go

@@ -0,0 +1,13 @@
+package utils
+
+import "regexp"
+
+var matchNewlines = regexp.MustCompile(`[\r\n]`)
+
+const doubleQuote = `"[^"\\]*(?:\\[\s\S][^"\\]*)*"`
+
+func EscapeNewLines(s string) string {
+	return regexp.MustCompile(doubleQuote).ReplaceAllStringFunc(s, func(s string) string {
+		return matchNewlines.ReplaceAllString(s, "\\n")
+	})
+}

pkg/utils/path.go

@@ -18,5 +18,5 @@ func inTrustedRoot(path string, trustedRoot string) error {
 // VerifyPath verifies that path is based in basePath.
 func VerifyPath(path, basePath string) error {
 	c := filepath.Clean(filepath.Join(basePath, path))
-	return inTrustedRoot(c, basePath)
+	return inTrustedRoot(c, filepath.Clean(basePath))
 }

pkg/utils/uri.go

@@ -2,8 +2,9 @@ package utils
 
 import (
 	"fmt"
-	"io/ioutil"
+	"io"
 	"net/http"
+	"os"
 	"strings"
 )
 
@@ -32,7 +33,7 @@ func GetURI(url string, f func(url string, i []byte) error) error {
 	if strings.HasPrefix(url, "file://") {
 		rawURL := strings.TrimPrefix(url, "file://")
 		// Read the response body
-		body, err := ioutil.ReadFile(rawURL)
+		body, err := os.ReadFile(rawURL)
 		if err != nil {
 			return err
 		}
@@ -49,7 +50,7 @@ func GetURI(url string, f func(url string, i []byte) error) error {
 	defer response.Body.Close()
 
 	// Read the response body
-	body, err := ioutil.ReadAll(response.Body)
+	body, err := io.ReadAll(response.Body)
 	if err != nil {
 		return err
 	}