Trigger CI

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

.air.toml

@@ -1,8 +0,0 @@
-# .air.toml
-[build]
-cmd = "make build"
-bin = "./local-ai"
-args_bin = [ "--debug" ]
-include_ext = ["go", "html", "yaml", "toml", "json", "txt", "md"]
-exclude_dir = ["pkg/grpc/proto"]
-delay = 1000

.dockerignore

@@ -6,10 +6,6 @@ models
 backends
 examples/chatbot-ui/models
 backend/go/image/stablediffusion-ggml/build/
-backend/go/*/build
-backend/go/*/.cache
-backend/go/*/sources
-backend/go/*/package
 examples/rwkv/models
 examples/**/models
 Dockerfile*

.github/gallery-agent/agent.go

@@ -1,288 +0,0 @@
-package main
-
-import (
-	"context"
-	"fmt"
-	"os"
-	"slices"
-	"strings"
-
-	hfapi "github.com/mudler/LocalAI/pkg/huggingface-api"
-	cogito "github.com/mudler/cogito"
-
-	"github.com/mudler/cogito/structures"
-	"github.com/sashabaranov/go-openai/jsonschema"
-)
-
-var (
-	openAIModel      = os.Getenv("OPENAI_MODEL")
-	openAIKey        = os.Getenv("OPENAI_KEY")
-	openAIBaseURL    = os.Getenv("OPENAI_BASE_URL")
-	galleryIndexPath = os.Getenv("GALLERY_INDEX_PATH")
-	//defaultclient
-	llm = cogito.NewOpenAILLM(openAIModel, openAIKey, openAIBaseURL)
-)
-
-// cleanTextContent removes trailing spaces, tabs, and normalizes line endings
-// to prevent YAML linting issues like trailing spaces and multiple empty lines
-func cleanTextContent(text string) string {
-	lines := strings.Split(text, "\n")
-	var cleanedLines []string
-	var prevEmpty bool
-	for _, line := range lines {
-		// Remove all trailing whitespace (spaces, tabs, etc.)
-		trimmed := strings.TrimRight(line, " \t\r")
-		// Avoid multiple consecutive empty lines
-		if trimmed == "" {
-			if !prevEmpty {
-				cleanedLines = append(cleanedLines, "")
-			}
-			prevEmpty = true
-		} else {
-			cleanedLines = append(cleanedLines, trimmed)
-			prevEmpty = false
-		}
-	}
-	// Remove trailing empty lines from the result
-	result := strings.Join(cleanedLines, "\n")
-	return strings.TrimRight(result, "\n")
-}
-
-// isModelExisting checks if a specific model ID exists in the gallery using text search
-func isModelExisting(modelID string) (bool, error) {
-	indexPath := getGalleryIndexPath()
-	content, err := os.ReadFile(indexPath)
-	if err != nil {
-		return false, fmt.Errorf("failed to read %s: %w", indexPath, err)
-	}
-
-	contentStr := string(content)
-	// Simple text search - if the model ID appears anywhere in the file, it exists
-	return strings.Contains(contentStr, modelID), nil
-}
-
-// filterExistingModels removes models that already exist in the gallery
-func filterExistingModels(models []ProcessedModel) ([]ProcessedModel, error) {
-	var filteredModels []ProcessedModel
-	for _, model := range models {
-		exists, err := isModelExisting(model.ModelID)
-		if err != nil {
-			fmt.Printf("Error checking if model %s exists: %v, skipping\n", model.ModelID, err)
-			continue
-		}
-
-		if !exists {
-			filteredModels = append(filteredModels, model)
-		} else {
-			fmt.Printf("Skipping existing model: %s\n", model.ModelID)
-		}
-	}
-
-	fmt.Printf("Filtered out %d existing models, %d new models remaining\n",
-		len(models)-len(filteredModels), len(filteredModels))
-
-	return filteredModels, nil
-}
-
-// getGalleryIndexPath returns the gallery index file path, with a default fallback
-func getGalleryIndexPath() string {
-	if galleryIndexPath != "" {
-		return galleryIndexPath
-	}
-	return "gallery/index.yaml"
-}
-
-func getRealReadme(ctx context.Context, repository string) (string, error) {
-	// Create a conversation fragment
-	fragment := cogito.NewEmptyFragment().
-		AddMessage("user",
-			`Your task is to get a clear description of a large language model from huggingface by using the provided tool. I will share with you a repository that might be quantized, and as such probably not by the original model author. We need to get the real  description of the model, and not the one that might be quantized. You will have to call the tool to get the readme more than once by figuring out from the quantized readme which is the base model readme. This is the repository: `+repository)
-
-	// Execute with tools
-	result, err := cogito.ExecuteTools(llm, fragment,
-		cogito.WithIterations(3),
-		cogito.WithMaxAttempts(3),
-		cogito.WithTools(&HFReadmeTool{client: hfapi.NewClient()}))
-	if err != nil {
-		return "", err
-	}
-
-	result = result.AddMessage("user", "Describe the model in a clear and concise way that can be shared in a model gallery.")
-
-	// Get a response
-	newFragment, err := llm.Ask(ctx, result)
-	if err != nil {
-		return "", err
-	}
-
-	content := newFragment.LastMessage().Content
-	return cleanTextContent(content), nil
-}
-
-func selectMostInterestingModels(ctx context.Context, searchResult *SearchResult) ([]ProcessedModel, error) {
-	// Create a conversation fragment
-	fragment := cogito.NewEmptyFragment().
-		AddMessage("user",
-			`Your task is to analyze a list of AI models and select the most interesting ones for a model gallery. You will be given detailed information about multiple models including their metadata, file information, and README content.
-
-Consider the following criteria when selecting models:
-1. Model popularity (download count)
-2. Model recency (last modified date)
-3. Model completeness (has preferred model file, README, etc.)
-4. Model uniqueness (not duplicates or very similar models)
-5. Model quality (based on README content and description)
-6. Model utility (practical applications)
-
-You should select models that would be most valuable for users browsing a model gallery. Prioritize models that are:
-- Well-documented with clear READMEs
-- Recently updated
-- Popular (high download count)
-- Have the preferred quantization format available
-- Offer unique capabilities or are from reputable authors
-
-Return your analysis and selection reasoning.`)
-
-	// Add the search results as context
-	modelsInfo := fmt.Sprintf("Found %d models matching '%s' with quantization preference '%s':\n\n",
-		searchResult.TotalModelsFound, searchResult.SearchTerm, searchResult.Quantization)
-
-	for i, model := range searchResult.Models {
-		modelsInfo += fmt.Sprintf("Model %d:\n", i+1)
-		modelsInfo += fmt.Sprintf("  ID: %s\n", model.ModelID)
-		modelsInfo += fmt.Sprintf("  Author: %s\n", model.Author)
-		modelsInfo += fmt.Sprintf("  Downloads: %d\n", model.Downloads)
-		modelsInfo += fmt.Sprintf("  Last Modified: %s\n", model.LastModified)
-		modelsInfo += fmt.Sprintf("  Files: %d files\n", len(model.Files))
-
-		if model.PreferredModelFile != nil {
-			modelsInfo += fmt.Sprintf("  Preferred Model File: %s (%d bytes)\n",
-				model.PreferredModelFile.Path, model.PreferredModelFile.Size)
-		} else {
-			modelsInfo += "  No preferred model file found\n"
-		}
-
-		if model.ReadmeContent != "" {
-			modelsInfo += fmt.Sprintf("  README: %s\n", model.ReadmeContent)
-		}
-
-		if model.ProcessingError != "" {
-			modelsInfo += fmt.Sprintf("  Processing Error: %s\n", model.ProcessingError)
-		}
-
-		modelsInfo += "\n"
-	}
-
-	fragment = fragment.AddMessage("user", modelsInfo)
-
-	fragment = fragment.AddMessage("user", "Based on your analysis, select the top 5 most interesting models and provide a brief explanation for each selection. Also, create a filtered SearchResult with only the selected models. Return just a list of repositories IDs, you will later be asked to output it as a JSON array with the json tool.")
-
-	// Get a response
-	newFragment, err := llm.Ask(ctx, fragment)
-	if err != nil {
-		return nil, err
-	}
-
-	fmt.Println(newFragment.LastMessage().Content)
-	repositories := struct {
-		Repositories []string `json:"repositories"`
-	}{}
-
-	s := structures.Structure{
-		Schema: jsonschema.Definition{
-			Type:                 jsonschema.Object,
-			AdditionalProperties: false,
-			Properties: map[string]jsonschema.Definition{
-				"repositories": {
-					Type:        jsonschema.Array,
-					Items:       &jsonschema.Definition{Type: jsonschema.String},
-					Description: "The trending repositories IDs",
-				},
-			},
-			Required: []string{"repositories"},
-		},
-		Object: &repositories,
-	}
-
-	err = newFragment.ExtractStructure(ctx, llm, s)
-	if err != nil {
-		return nil, err
-	}
-
-	filteredModels := []ProcessedModel{}
-	for _, m := range searchResult.Models {
-		if slices.Contains(repositories.Repositories, m.ModelID) {
-			filteredModels = append(filteredModels, m)
-		}
-	}
-
-	return filteredModels, nil
-}
-
-// ModelFamily represents a YAML anchor/family
-type ModelFamily struct {
-	Anchor string `json:"anchor"`
-	Name   string `json:"name"`
-}
-
-// selectModelFamily selects the appropriate model family/anchor for a given model
-func selectModelFamily(ctx context.Context, model ProcessedModel, availableFamilies []ModelFamily) (string, error) {
-	// Create a conversation fragment
-	fragment := cogito.NewEmptyFragment().
-		AddMessage("user",
-			`Your task is to select the most appropriate model family/anchor for a given AI model. You will be provided with:
-1. Information about the model (name, description, etc.)
-2. A list of available model families/anchors
-
-You need to select the family that best matches the model's architecture, capabilities, or characteristics. Consider:
-- Model architecture (e.g., Llama, Qwen, Mistral, etc.)
-- Model capabilities (e.g., vision, coding, chat, etc.)
-- Model size/type (e.g., small, medium, large)
-- Model purpose (e.g., general purpose, specialized, etc.)
-
-Return the anchor name that best fits the model.`)
-
-	// Add model information
-	modelInfo := "Model Information:\n"
-	modelInfo += fmt.Sprintf("  ID: %s\n", model.ModelID)
-	modelInfo += fmt.Sprintf("  Author: %s\n", model.Author)
-	modelInfo += fmt.Sprintf("  Downloads: %d\n", model.Downloads)
-	modelInfo += fmt.Sprintf("  Description: %s\n", model.ReadmeContentPreview)
-
-	fragment = fragment.AddMessage("user", modelInfo)
-
-	// Add available families
-	familiesInfo := "Available Model Families:\n"
-	for _, family := range availableFamilies {
-		familiesInfo += fmt.Sprintf("  - %s (%s)\n", family.Anchor, family.Name)
-	}
-
-	fragment = fragment.AddMessage("user", familiesInfo)
-	fragment = fragment.AddMessage("user", "Select the most appropriate family anchor for this model. Return just the anchor name.")
-
-	// Get a response
-	newFragment, err := llm.Ask(ctx, fragment)
-	if err != nil {
-		return "", err
-	}
-
-	// Extract the selected family
-	selectedFamily := strings.TrimSpace(newFragment.LastMessage().Content)
-
-	// Validate that the selected family exists in our list
-	for _, family := range availableFamilies {
-		if family.Anchor == selectedFamily {
-			return selectedFamily, nil
-		}
-	}
-
-	// If no exact match, try to find a close match
-	for _, family := range availableFamilies {
-		if strings.Contains(strings.ToLower(family.Anchor), strings.ToLower(selectedFamily)) ||
-			strings.Contains(strings.ToLower(selectedFamily), strings.ToLower(family.Anchor)) {
-			return family.Anchor, nil
-		}
-	}
-
-	// Default fallback
-	return "llama3", nil
-}

.github/gallery-agent/gallery.go

@@ -1,203 +0,0 @@
-package main
-
-import (
-	"context"
-	"fmt"
-	"os"
-	"strings"
-)
-
-// generateYAMLEntry generates a YAML entry for a model using the specified anchor
-func generateYAMLEntry(model ProcessedModel, familyAnchor string) string {
-	// Extract model name from ModelID
-	parts := strings.Split(model.ModelID, "/")
-	modelName := model.ModelID
-	if len(parts) > 0 {
-		modelName = strings.ToLower(parts[len(parts)-1])
-	}
-	// Remove common suffixes
-	modelName = strings.ReplaceAll(modelName, "-gguf", "")
-	modelName = strings.ReplaceAll(modelName, "-q4_k_m", "")
-	modelName = strings.ReplaceAll(modelName, "-q4_k_s", "")
-	modelName = strings.ReplaceAll(modelName, "-q3_k_m", "")
-	modelName = strings.ReplaceAll(modelName, "-q2_k", "")
-
-	fileName := ""
-	checksum := ""
-	if model.PreferredModelFile != nil {
-		fileParts := strings.Split(model.PreferredModelFile.Path, "/")
-		if len(fileParts) > 0 {
-			fileName = fileParts[len(fileParts)-1]
-		}
-		checksum = model.PreferredModelFile.SHA256
-	} else {
-		fileName = model.ModelID
-	}
-
-	description := model.ReadmeContent
-	if description == "" {
-		description = fmt.Sprintf("AI model: %s", modelName)
-	}
-
-	// Clean up description to prevent YAML linting issues
-	description = cleanTextContent(description)
-
-	// Format description for YAML (indent each line and ensure no trailing spaces)
-	lines := strings.Split(description, "\n")
-	var formattedLines []string
-	for _, line := range lines {
-		if strings.TrimSpace(line) == "" {
-			// Keep empty lines as empty (no indentation)
-			formattedLines = append(formattedLines, "")
-		} else {
-			// Add indentation to non-empty lines
-			formattedLines = append(formattedLines, "    "+line)
-		}
-	}
-	formattedDescription := strings.Join(formattedLines, "\n")
-	// Remove any trailing spaces from the formatted description
-	formattedDescription = strings.TrimRight(formattedDescription, " \t")
-	yamlTemplate := ""
-	if checksum != "" {
-		yamlTemplate = `- !!merge <<: *%s
-  name: "%s"
-  urls:
-    - https://huggingface.co/%s
-  description: |
-%s
-  overrides:
-    parameters:
-      model: %s
-  files:
-    - filename: %s
-      sha256: %s
-      uri: huggingface://%s/%s`
-		return fmt.Sprintf(yamlTemplate,
-			familyAnchor,
-			modelName,
-			model.ModelID,
-			formattedDescription,
-			fileName,
-			fileName,
-			checksum,
-			model.ModelID,
-			fileName,
-		)
-	} else {
-		yamlTemplate = `- !!merge <<: *%s
-  name: "%s"
-  urls:
-    - https://huggingface.co/%s
-  description: |
-%s
-  overrides:
-    parameters:
-      model: %s`
-		return fmt.Sprintf(yamlTemplate,
-			familyAnchor,
-			modelName,
-			model.ModelID,
-			formattedDescription,
-			fileName,
-		)
-	}
-}
-
-// extractModelFamilies extracts all YAML anchors from the gallery index.yaml file
-func extractModelFamilies() ([]ModelFamily, error) {
-	// Read the index.yaml file
-	indexPath := getGalleryIndexPath()
-	content, err := os.ReadFile(indexPath)
-	if err != nil {
-		return nil, fmt.Errorf("failed to read %s: %w", indexPath, err)
-	}
-
-	lines := strings.Split(string(content), "\n")
-	var families []ModelFamily
-
-	for _, line := range lines {
-		line = strings.TrimSpace(line)
-		// Look for YAML anchors (lines starting with "- &")
-		if strings.HasPrefix(line, "- &") {
-			// Extract the anchor name (everything after "- &")
-			anchor := strings.TrimPrefix(line, "- &")
-			// Remove any trailing colon or other characters
-			anchor = strings.Split(anchor, ":")[0]
-			anchor = strings.Split(anchor, " ")[0]
-
-			if anchor != "" {
-				families = append(families, ModelFamily{
-					Anchor: anchor,
-					Name:   anchor, // Use anchor as name for now
-				})
-			}
-		}
-	}
-
-	return families, nil
-}
-
-// generateYAMLForModels generates YAML entries for selected models and appends to index.yaml
-func generateYAMLForModels(ctx context.Context, models []ProcessedModel) error {
-	// Extract available model families
-	families, err := extractModelFamilies()
-	if err != nil {
-		return fmt.Errorf("failed to extract model families: %w", err)
-	}
-
-	fmt.Printf("Found %d model families: %v\n", len(families),
-		func() []string {
-			var names []string
-			for _, f := range families {
-				names = append(names, f.Anchor)
-			}
-			return names
-		}())
-
-	// Generate YAML entries for each model
-	var yamlEntries []string
-	for _, model := range models {
-		fmt.Printf("Selecting family for model: %s\n", model.ModelID)
-
-		// Select appropriate family for this model
-		familyAnchor, err := selectModelFamily(ctx, model, families)
-		if err != nil {
-			fmt.Printf("Error selecting family for %s: %v, using default\n", model.ModelID, err)
-			familyAnchor = "llama3" // Default fallback
-		}
-
-		fmt.Printf("Selected family '%s' for model %s\n", familyAnchor, model.ModelID)
-
-		// Generate YAML entry
-		yamlEntry := generateYAMLEntry(model, familyAnchor)
-		yamlEntries = append(yamlEntries, yamlEntry)
-	}
-
-	// Append to index.yaml
-	if len(yamlEntries) > 0 {
-		indexPath := getGalleryIndexPath()
-		fmt.Printf("Appending YAML entries to %s...\n", indexPath)
-
-		// Read current content
-		content, err := os.ReadFile(indexPath)
-		if err != nil {
-			return fmt.Errorf("failed to read %s: %w", indexPath, err)
-		}
-
-		// Append new entries
-		// Remove trailing whitespace from existing content and join entries without extra newlines
-		existingContent := strings.TrimRight(string(content), " \t\n\r")
-		yamlBlock := strings.Join(yamlEntries, "\n")
-		newContent := existingContent + "\n" + yamlBlock + "\n"
-
-		// Write back to file
-		err = os.WriteFile(indexPath, []byte(newContent), 0644)
-		if err != nil {
-			return fmt.Errorf("failed to write %s: %w", indexPath, err)
-		}
-
-		fmt.Printf("Successfully added %d models to %s\n", len(yamlEntries), indexPath)
-	}
-
-	return nil
-}

.github/gallery-agent/main.go

@@ -1,351 +0,0 @@
-package main
-
-import (
-	"context"
-	"encoding/json"
-	"fmt"
-	"os"
-	"strconv"
-	"strings"
-	"time"
-
-	hfapi "github.com/mudler/LocalAI/pkg/huggingface-api"
-)
-
-// ProcessedModelFile represents a processed model file with additional metadata
-type ProcessedModelFile struct {
-	Path     string `json:"path"`
-	Size     int64  `json:"size"`
-	SHA256   string `json:"sha256"`
-	IsReadme bool   `json:"is_readme"`
-	FileType string `json:"file_type"` // "model", "readme", "other"
-}
-
-// ProcessedModel represents a processed model with all gathered metadata
-type ProcessedModel struct {
-	ModelID                 string               `json:"model_id"`
-	Author                  string               `json:"author"`
-	Downloads               int                  `json:"downloads"`
-	LastModified            string               `json:"last_modified"`
-	Files                   []ProcessedModelFile `json:"files"`
-	PreferredModelFile      *ProcessedModelFile  `json:"preferred_model_file,omitempty"`
-	ReadmeFile              *ProcessedModelFile  `json:"readme_file,omitempty"`
-	ReadmeContent           string               `json:"readme_content,omitempty"`
-	ReadmeContentPreview    string               `json:"readme_content_preview,omitempty"`
-	QuantizationPreferences []string             `json:"quantization_preferences"`
-	ProcessingError         string               `json:"processing_error,omitempty"`
-}
-
-// SearchResult represents the complete result of searching and processing models
-type SearchResult struct {
-	SearchTerm       string           `json:"search_term"`
-	Limit            int              `json:"limit"`
-	Quantization     string           `json:"quantization"`
-	TotalModelsFound int              `json:"total_models_found"`
-	Models           []ProcessedModel `json:"models"`
-	FormattedOutput  string           `json:"formatted_output"`
-}
-
-// AddedModelSummary represents a summary of models added to the gallery
-type AddedModelSummary struct {
-	SearchTerm     string   `json:"search_term"`
-	TotalFound     int      `json:"total_found"`
-	ModelsAdded    int      `json:"models_added"`
-	AddedModelIDs  []string `json:"added_model_ids"`
-	AddedModelURLs []string `json:"added_model_urls"`
-	Quantization   string   `json:"quantization"`
-	ProcessingTime string   `json:"processing_time"`
-}
-
-func main() {
-	startTime := time.Now()
-
-	// Check for synthetic mode
-	syntheticMode := os.Getenv("SYNTHETIC_MODE")
-	if syntheticMode == "true" || syntheticMode == "1" {
-		fmt.Println("Running in SYNTHETIC MODE - generating random test data")
-		err := runSyntheticMode()
-		if err != nil {
-			fmt.Fprintf(os.Stderr, "Error in synthetic mode: %v\n", err)
-			os.Exit(1)
-		}
-		return
-	}
-
-	// Get configuration from environment variables
-	searchTerm := os.Getenv("SEARCH_TERM")
-	if searchTerm == "" {
-		searchTerm = "GGUF"
-	}
-
-	limitStr := os.Getenv("LIMIT")
-	if limitStr == "" {
-		limitStr = "5"
-	}
-	limit, err := strconv.Atoi(limitStr)
-	if err != nil {
-		fmt.Fprintf(os.Stderr, "Error parsing LIMIT: %v\n", err)
-		os.Exit(1)
-	}
-
-	quantization := os.Getenv("QUANTIZATION")
-
-	maxModels := os.Getenv("MAX_MODELS")
-	if maxModels == "" {
-		maxModels = "1"
-	}
-	maxModelsInt, err := strconv.Atoi(maxModels)
-	if err != nil {
-		fmt.Fprintf(os.Stderr, "Error parsing MAX_MODELS: %v\n", err)
-		os.Exit(1)
-	}
-
-	// Print configuration
-	fmt.Printf("Gallery Agent Configuration:\n")
-	fmt.Printf("  Search Term: %s\n", searchTerm)
-	fmt.Printf("  Limit: %d\n", limit)
-	fmt.Printf("  Quantization: %s\n", quantization)
-	fmt.Printf("  Max Models to Add: %d\n", maxModelsInt)
-	fmt.Printf("  Gallery Index Path: %s\n", os.Getenv("GALLERY_INDEX_PATH"))
-	fmt.Println()
-
-	result, err := searchAndProcessModels(searchTerm, limit, quantization)
-	if err != nil {
-		fmt.Fprintf(os.Stderr, "Error: %v\n", err)
-		os.Exit(1)
-	}
-
-	fmt.Println(result.FormattedOutput)
-
-	// Use AI agent to select the most interesting models
-	fmt.Println("Using AI agent to select the most interesting models...")
-	models, err := selectMostInterestingModels(context.Background(), result)
-	if err != nil {
-		fmt.Fprintf(os.Stderr, "Error in model selection: %v\n", err)
-		// Continue with original result if selection fails
-		models = result.Models
-	}
-
-	fmt.Print(models)
-
-	// Filter out models that already exist in the gallery
-	fmt.Println("Filtering out existing models...")
-	models, err = filterExistingModels(models)
-	if err != nil {
-		fmt.Fprintf(os.Stderr, "Error filtering existing models: %v\n", err)
-		os.Exit(1)
-	}
-
-	// Limit to maxModelsInt after filtering
-	if len(models) > maxModelsInt {
-		models = models[:maxModelsInt]
-	}
-
-	// Track added models for summary
-	var addedModelIDs []string
-	var addedModelURLs []string
-
-	// Generate YAML entries and append to gallery/index.yaml
-	if len(models) > 0 {
-		for _, model := range models {
-			addedModelIDs = append(addedModelIDs, model.ModelID)
-			// Generate Hugging Face URL for the model
-			modelURL := fmt.Sprintf("https://huggingface.co/%s", model.ModelID)
-			addedModelURLs = append(addedModelURLs, modelURL)
-		}
-		fmt.Println("Generating YAML entries for selected models...")
-		err = generateYAMLForModels(context.Background(), models)
-		if err != nil {
-			fmt.Fprintf(os.Stderr, "Error generating YAML entries: %v\n", err)
-			os.Exit(1)
-		}
-	} else {
-		fmt.Println("No new models to add to the gallery.")
-	}
-
-	// Create and write summary
-	processingTime := time.Since(startTime).String()
-	summary := AddedModelSummary{
-		SearchTerm:     searchTerm,
-		TotalFound:     result.TotalModelsFound,
-		ModelsAdded:    len(addedModelIDs),
-		AddedModelIDs:  addedModelIDs,
-		AddedModelURLs: addedModelURLs,
-		Quantization:   quantization,
-		ProcessingTime: processingTime,
-	}
-
-	// Write summary to file
-	summaryData, err := json.MarshalIndent(summary, "", "  ")
-	if err != nil {
-		fmt.Fprintf(os.Stderr, "Error marshaling summary: %v\n", err)
-	} else {
-		err = os.WriteFile("gallery-agent-summary.json", summaryData, 0644)
-		if err != nil {
-			fmt.Fprintf(os.Stderr, "Error writing summary file: %v\n", err)
-		} else {
-			fmt.Printf("Summary written to gallery-agent-summary.json\n")
-		}
-	}
-}
-
-func searchAndProcessModels(searchTerm string, limit int, quantization string) (*SearchResult, error) {
-	client := hfapi.NewClient()
-	var outputBuilder strings.Builder
-
-	fmt.Println("Searching for models...")
-	// Initialize the result struct
-	result := &SearchResult{
-		SearchTerm:   searchTerm,
-		Limit:        limit,
-		Quantization: quantization,
-		Models:       []ProcessedModel{},
-	}
-
-	models, err := client.GetLatest(searchTerm, limit)
-	if err != nil {
-		return nil, fmt.Errorf("failed to fetch models: %w", err)
-	}
-
-	fmt.Println("Models found:", len(models))
-	result.TotalModelsFound = len(models)
-
-	if len(models) == 0 {
-		outputBuilder.WriteString("No models found.\n")
-		result.FormattedOutput = outputBuilder.String()
-		return result, nil
-	}
-
-	outputBuilder.WriteString(fmt.Sprintf("Found %d models matching '%s':\n\n", len(models), searchTerm))
-
-	// Process each model
-	for i, model := range models {
-		outputBuilder.WriteString(fmt.Sprintf("%d. Processing Model: %s\n", i+1, model.ModelID))
-		outputBuilder.WriteString(fmt.Sprintf("   Author: %s\n", model.Author))
-		outputBuilder.WriteString(fmt.Sprintf("   Downloads: %d\n", model.Downloads))
-		outputBuilder.WriteString(fmt.Sprintf("   Last Modified: %s\n", model.LastModified))
-
-		// Initialize processed model struct
-		processedModel := ProcessedModel{
-			ModelID:                 model.ModelID,
-			Author:                  model.Author,
-			Downloads:               model.Downloads,
-			LastModified:            model.LastModified,
-			QuantizationPreferences: []string{quantization, "Q4_K_M", "Q4_K_S", "Q3_K_M", "Q2_K"},
-		}
-
-		// Get detailed model information
-		details, err := client.GetModelDetails(model.ModelID)
-		if err != nil {
-			errorMsg := fmt.Sprintf("   Error getting model details: %v\n", err)
-			outputBuilder.WriteString(errorMsg)
-			processedModel.ProcessingError = err.Error()
-			result.Models = append(result.Models, processedModel)
-			continue
-		}
-
-		// Define quantization preferences (in order of preference)
-		quantizationPreferences := []string{quantization, "Q4_K_M", "Q4_K_S", "Q3_K_M", "Q2_K"}
-
-		// Find preferred model file
-		preferredModelFile := hfapi.FindPreferredModelFile(details.Files, quantizationPreferences)
-
-		// Process files
-		processedFiles := make([]ProcessedModelFile, len(details.Files))
-		for j, file := range details.Files {
-			fileType := "other"
-			if file.IsReadme {
-				fileType = "readme"
-			} else if preferredModelFile != nil && file.Path == preferredModelFile.Path {
-				fileType = "model"
-			}
-
-			processedFiles[j] = ProcessedModelFile{
-				Path:     file.Path,
-				Size:     file.Size,
-				SHA256:   file.SHA256,
-				IsReadme: file.IsReadme,
-				FileType: fileType,
-			}
-		}
-
-		processedModel.Files = processedFiles
-
-		// Set preferred model file
-		if preferredModelFile != nil {
-			for _, file := range processedFiles {
-				if file.Path == preferredModelFile.Path {
-					processedModel.PreferredModelFile = &file
-					break
-				}
-			}
-		}
-
-		// Print file information
-		outputBuilder.WriteString(fmt.Sprintf("   Files found: %d\n", len(details.Files)))
-
-		if preferredModelFile != nil {
-			outputBuilder.WriteString(fmt.Sprintf("   Preferred Model File: %s (SHA256: %s)\n",
-				preferredModelFile.Path,
-				preferredModelFile.SHA256))
-		} else {
-			outputBuilder.WriteString(fmt.Sprintf("   No model file found with quantization preferences: %v\n", quantizationPreferences))
-		}
-
-		if details.ReadmeFile != nil {
-			outputBuilder.WriteString(fmt.Sprintf("   README File: %s\n", details.ReadmeFile.Path))
-
-			// Find and set readme file
-			for _, file := range processedFiles {
-				if file.IsReadme {
-					processedModel.ReadmeFile = &file
-					break
-				}
-			}
-
-			fmt.Println("Getting real readme for", model.ModelID, "waiting...")
-			// Use agent to get the real readme and prepare the model description
-			readmeContent, err := getRealReadme(context.Background(), model.ModelID)
-			if err == nil {
-				processedModel.ReadmeContent = readmeContent
-				processedModel.ReadmeContentPreview = truncateString(readmeContent, 200)
-				outputBuilder.WriteString(fmt.Sprintf("   README Content Preview: %s\n",
-					processedModel.ReadmeContentPreview))
-			} else {
-				continue
-			}
-			fmt.Println("Real readme got", readmeContent)
-			// Get README content
-			// readmeContent, err := client.GetReadmeContent(model.ModelID, details.ReadmeFile.Path)
-			// if err == nil {
-			// 	processedModel.ReadmeContent = readmeContent
-			// 	processedModel.ReadmeContentPreview = truncateString(readmeContent, 200)
-			// 	outputBuilder.WriteString(fmt.Sprintf("   README Content Preview: %s\n",
-			// 		processedModel.ReadmeContentPreview))
-			// }
-		}
-
-		// Print all files with their checksums
-		outputBuilder.WriteString("   All Files:\n")
-		for _, file := range processedFiles {
-			outputBuilder.WriteString(fmt.Sprintf("     - %s (%s, %d bytes", file.Path, file.FileType, file.Size))
-			if file.SHA256 != "" {
-				outputBuilder.WriteString(fmt.Sprintf(", SHA256: %s", file.SHA256))
-			}
-			outputBuilder.WriteString(")\n")
-		}
-
-		outputBuilder.WriteString("\n")
-		result.Models = append(result.Models, processedModel)
-	}
-
-	result.FormattedOutput = outputBuilder.String()
-	return result, nil
-}
-
-func truncateString(s string, maxLen int) string {
-	if len(s) <= maxLen {
-		return s
-	}
-	return s[:maxLen] + "..."
-}

.github/gallery-agent/testing.go

@@ -1,190 +0,0 @@
-package main
-
-import (
-	"context"
-	"fmt"
-	"math/rand"
-	"strings"
-	"time"
-)
-
-// runSyntheticMode generates synthetic test data and appends it to the gallery
-func runSyntheticMode() error {
-	generator := NewSyntheticDataGenerator()
-
-	// Generate a random number of synthetic models (1-3)
-	numModels := generator.rand.Intn(3) + 1
-	fmt.Printf("Generating %d synthetic models for testing...\n", numModels)
-
-	var models []ProcessedModel
-	for i := 0; i < numModels; i++ {
-		model := generator.GenerateProcessedModel()
-		models = append(models, model)
-		fmt.Printf("Generated synthetic model: %s\n", model.ModelID)
-	}
-
-	// Generate YAML entries and append to gallery/index.yaml
-	fmt.Println("Generating YAML entries for synthetic models...")
-	err := generateYAMLForModels(context.Background(), models)
-	if err != nil {
-		return fmt.Errorf("error generating YAML entries: %w", err)
-	}
-
-	fmt.Printf("Successfully added %d synthetic models to the gallery for testing!\n", len(models))
-	return nil
-}
-
-// SyntheticDataGenerator provides methods to generate synthetic test data
-type SyntheticDataGenerator struct {
-	rand *rand.Rand
-}
-
-// NewSyntheticDataGenerator creates a new synthetic data generator
-func NewSyntheticDataGenerator() *SyntheticDataGenerator {
-	return &SyntheticDataGenerator{
-		rand: rand.New(rand.NewSource(time.Now().UnixNano())),
-	}
-}
-
-// GenerateProcessedModelFile creates a synthetic ProcessedModelFile
-func (g *SyntheticDataGenerator) GenerateProcessedModelFile() ProcessedModelFile {
-	fileTypes := []string{"model", "readme", "other"}
-	fileType := fileTypes[g.rand.Intn(len(fileTypes))]
-
-	var path string
-	var isReadme bool
-
-	switch fileType {
-	case "model":
-		path = fmt.Sprintf("model-%s.gguf", g.randomString(8))
-		isReadme = false
-	case "readme":
-		path = "README.md"
-		isReadme = true
-	default:
-		path = fmt.Sprintf("file-%s.txt", g.randomString(6))
-		isReadme = false
-	}
-
-	return ProcessedModelFile{
-		Path:     path,
-		Size:     int64(g.rand.Intn(1000000000) + 1000000), // 1MB to 1GB
-		SHA256:   g.randomSHA256(),
-		IsReadme: isReadme,
-		FileType: fileType,
-	}
-}
-
-// GenerateProcessedModel creates a synthetic ProcessedModel
-func (g *SyntheticDataGenerator) GenerateProcessedModel() ProcessedModel {
-	authors := []string{"microsoft", "meta", "google", "openai", "anthropic", "mistralai", "huggingface"}
-	modelNames := []string{"llama", "gpt", "claude", "mistral", "gemma", "phi", "qwen", "codellama"}
-
-	author := authors[g.rand.Intn(len(authors))]
-	modelName := modelNames[g.rand.Intn(len(modelNames))]
-	modelID := fmt.Sprintf("%s/%s-%s", author, modelName, g.randomString(6))
-
-	// Generate files
-	numFiles := g.rand.Intn(5) + 2 // 2-6 files
-	files := make([]ProcessedModelFile, numFiles)
-
-	// Ensure at least one model file and one readme
-	hasModelFile := false
-	hasReadme := false
-
-	for i := 0; i < numFiles; i++ {
-		files[i] = g.GenerateProcessedModelFile()
-		if files[i].FileType == "model" {
-			hasModelFile = true
-		}
-		if files[i].FileType == "readme" {
-			hasReadme = true
-		}
-	}
-
-	// Add required files if missing
-	if !hasModelFile {
-		modelFile := g.GenerateProcessedModelFile()
-		modelFile.FileType = "model"
-		modelFile.Path = fmt.Sprintf("%s-Q4_K_M.gguf", modelName)
-		files = append(files, modelFile)
-	}
-
-	if !hasReadme {
-		readmeFile := g.GenerateProcessedModelFile()
-		readmeFile.FileType = "readme"
-		readmeFile.Path = "README.md"
-		readmeFile.IsReadme = true
-		files = append(files, readmeFile)
-	}
-
-	// Find preferred model file
-	var preferredModelFile *ProcessedModelFile
-	for i := range files {
-		if files[i].FileType == "model" {
-			preferredModelFile = &files[i]
-			break
-		}
-	}
-
-	// Find readme file
-	var readmeFile *ProcessedModelFile
-	for i := range files {
-		if files[i].FileType == "readme" {
-			readmeFile = &files[i]
-			break
-		}
-	}
-
-	readmeContent := g.generateReadmeContent(modelName, author)
-
-	return ProcessedModel{
-		ModelID:                 modelID,
-		Author:                  author,
-		Downloads:               g.rand.Intn(1000000) + 1000,
-		LastModified:            g.randomDate(),
-		Files:                   files,
-		PreferredModelFile:      preferredModelFile,
-		ReadmeFile:              readmeFile,
-		ReadmeContent:           readmeContent,
-		ReadmeContentPreview:    truncateString(readmeContent, 200),
-		QuantizationPreferences: []string{"Q4_K_M", "Q4_K_S", "Q3_K_M", "Q2_K"},
-		ProcessingError:         "",
-	}
-}
-
-// Helper methods for synthetic data generation
-func (g *SyntheticDataGenerator) randomString(length int) string {
-	const charset = "abcdefghijklmnopqrstuvwxyz0123456789"
-	b := make([]byte, length)
-	for i := range b {
-		b[i] = charset[g.rand.Intn(len(charset))]
-	}
-	return string(b)
-}
-
-func (g *SyntheticDataGenerator) randomSHA256() string {
-	const charset = "0123456789abcdef"
-	b := make([]byte, 64)
-	for i := range b {
-		b[i] = charset[g.rand.Intn(len(charset))]
-	}
-	return string(b)
-}
-
-func (g *SyntheticDataGenerator) randomDate() string {
-	now := time.Now()
-	daysAgo := g.rand.Intn(365) // Random date within last year
-	pastDate := now.AddDate(0, 0, -daysAgo)
-	return pastDate.Format("2006-01-02T15:04:05.000Z")
-}
-
-func (g *SyntheticDataGenerator) generateReadmeContent(modelName, author string) string {
-	templates := []string{
-		fmt.Sprintf("# %s Model\n\nThis is a %s model developed by %s. It's designed for various natural language processing tasks including text generation, question answering, and conversation.\n\n## Features\n\n- High-quality text generation\n- Efficient inference\n- Multiple quantization options\n- Easy to use with LocalAI\n\n## Usage\n\nUse this model with LocalAI for various AI tasks.", strings.Title(modelName), modelName, author),
-		fmt.Sprintf("# %s\n\nA powerful language model from %s. This model excels at understanding and generating human-like text across multiple domains.\n\n## Capabilities\n\n- Text completion\n- Code generation\n- Creative writing\n- Technical documentation\n\n## Model Details\n\n- Architecture: Transformer-based\n- Training: Large-scale supervised learning\n- Quantization: Available in multiple formats", strings.Title(modelName), author),
-		fmt.Sprintf("# %s Language Model\n\nDeveloped by %s, this model represents state-of-the-art performance in natural language understanding and generation.\n\n## Key Features\n\n- Multilingual support\n- Context-aware responses\n- Efficient memory usage\n- Fast inference speed\n\n## Applications\n\n- Chatbots and virtual assistants\n- Content generation\n- Code completion\n- Educational tools", strings.Title(modelName), author),
-	}
-
-	return templates[g.rand.Intn(len(templates))]
-}

.github/gallery-agent/tools.go

@@ -1,46 +0,0 @@
-package main
-
-import (
-	"fmt"
-
-	hfapi "github.com/mudler/LocalAI/pkg/huggingface-api"
-	openai "github.com/sashabaranov/go-openai"
-	jsonschema "github.com/sashabaranov/go-openai/jsonschema"
-)
-
-// Get repository README from HF
-type HFReadmeTool struct {
-	client *hfapi.Client
-}
-
-func (s *HFReadmeTool) Execute(args map[string]any) (string, error) {
-	q, ok := args["repository"].(string)
-	if !ok {
-		return "", fmt.Errorf("no query")
-	}
-	readme, err := s.client.GetReadmeContent(q, "README.md")
-	if err != nil {
-		return "", err
-	}
-	return readme, nil
-}
-
-func (s *HFReadmeTool) Tool() openai.Tool {
-	return openai.Tool{
-		Type: openai.ToolTypeFunction,
-		Function: &openai.FunctionDefinition{
-			Name:        "hf_readme",
-			Description: "A tool to get the README content of a huggingface repository",
-			Parameters: jsonschema.Definition{
-				Type: jsonschema.Object,
-				Properties: map[string]jsonschema.Definition{
-					"repository": {
-						Type:        jsonschema.String,
-						Description: "The huggingface repository to get the README content of",
-					},
-				},
-				Required: []string{"repository"},
-			},
-		},
-	}
-}

.github/workflows/backend.yml

@@ -2,6 +2,7 @@
 name: 'build backend container images'
 
 on:
+  pull_request:
   push:
     branches:
       - master
@@ -63,6 +64,18 @@ jobs:
             backend: "llama-cpp"
             dockerfile: "./backend/Dockerfile.llama-cpp"
             context: "./"
+          - build-type: ''
+            cuda-major-version: ""
+            cuda-minor-version: ""
+            platforms: 'linux/amd64'
+            tag-latest: 'auto'
+            tag-suffix: '-cpu-transformers'
+            runs-on: 'ubuntu-latest'
+            base-image: "ubuntu:22.04"
+            skip-drivers: 'true'
+            backend: "transformers"
+            dockerfile: "./backend/Dockerfile.python"
+            context: "./backend"
           - build-type: 'cublas'
             cuda-major-version: "11"
             cuda-minor-version: "7"
@@ -111,18 +124,6 @@ jobs:
             backend: "diffusers"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          - build-type: ''
-            cuda-major-version: ""
-            cuda-minor-version: ""
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-cpu-chatterbox'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            skip-drivers: 'true'
-            backend: "chatterbox"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
           # CUDA 11 additional backends
           - build-type: 'cublas'
             cuda-major-version: "11"
@@ -242,7 +243,7 @@ jobs:
             runs-on: 'ubuntu-latest'
             base-image: "ubuntu:22.04"
             skip-drivers: 'false'
-            backend: "diffusers"
+            backend: "diffusers"            
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
           # CUDA 12 additional backends
@@ -489,18 +490,6 @@ jobs:
             backend: "diffusers"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          - build-type: 'l4t'
-            cuda-major-version: "12"
-            cuda-minor-version: "0"
-            platforms: 'linux/arm64'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-nvidia-l4t-kokoro'
-            runs-on: 'ubuntu-24.04-arm'
-            base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
-            skip-drivers: 'true'
-            backend: "kokoro"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
           # SYCL additional backends
           - build-type: 'intel'
             cuda-major-version: ""
@@ -787,7 +776,7 @@ jobs:
             cuda-minor-version: ""
             platforms: 'linux/amd64'
             tag-latest: 'auto'
-            tag-suffix: '-gpu-rocm-hipblas-whisper'
+            tag-suffix: '-gpu-hipblas-whisper'
             base-image: "rocm/dev-ubuntu-22.04:6.4.3"
             runs-on: 'ubuntu-latest'
             skip-drivers: 'false'
@@ -882,7 +871,7 @@ jobs:
             backend: "rfdetr"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          - build-type: 'l4t'
+          - build-type: 'cublas'
             cuda-major-version: "12"
             cuda-minor-version: "0"
             platforms: 'linux/arm64'
@@ -955,18 +944,6 @@ jobs:
             backend: "exllama2"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          - build-type: 'l4t'
-            cuda-major-version: "12"
-            cuda-minor-version: "0"
-            platforms: 'linux/arm64'
-            skip-drivers: 'true'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-nvidia-l4t-arm64-chatterbox'
-            base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
-            runs-on: 'ubuntu-24.04-arm'
-            backend: "chatterbox"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
           # runs out of space on the runner
           # - build-type: 'hipblas'
           #   cuda-major-version: ""
@@ -993,90 +970,54 @@ jobs:
             backend: "kitten-tts"
             dockerfile: "./backend/Dockerfile.python"
             context: "./backend"
-          # neutts
-          - build-type: ''
-            cuda-major-version: ""
-            cuda-minor-version: ""
-            platforms: 'linux/amd64,linux/arm64'
-            tag-latest: 'auto'
-            tag-suffix: '-cpu-neutts'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            skip-drivers: 'false'
-            backend: "neutts"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'cublas'
-            cuda-major-version: "12"
-            cuda-minor-version: "0"
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-nvidia-cuda-12-neutts'
-            runs-on: 'ubuntu-latest'
-            base-image: "ubuntu:22.04"
-            skip-drivers: 'false'
-            backend: "neutts"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'hipblas'
-            cuda-major-version: ""
-            cuda-minor-version: ""
-            platforms: 'linux/amd64'
-            tag-latest: 'auto'
-            tag-suffix: '-gpu-rocm-hipblas-neutts'
-            runs-on: 'arc-runner-set'
-            base-image: "rocm/dev-ubuntu-22.04:6.4.3"
-            skip-drivers: 'false'
-            backend: "neutts"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-          - build-type: 'l4t'
-            cuda-major-version: "12"
-            cuda-minor-version: "0"
-            platforms: 'linux/arm64'
-            skip-drivers: 'true'
-            tag-latest: 'auto'
-            tag-suffix: '-nvidia-l4t-arm64-neutts'
-            base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
-            runs-on: 'ubuntu-24.04-arm'
-            backend: "neutts"
-            dockerfile: "./backend/Dockerfile.python"
-            context: "./backend"
-  backend-jobs-darwin:
+  transformers-darwin:
     uses: ./.github/workflows/backend_build_darwin.yml
-    strategy:
-      matrix:
-        include:
-          - backend: "diffusers"
-            tag-suffix: "-metal-darwin-arm64-diffusers"
-            build-type: "mps"
-          - backend: "mlx"
-            tag-suffix: "-metal-darwin-arm64-mlx"
-            build-type: "mps"
-          - backend: "chatterbox"
-            tag-suffix: "-metal-darwin-arm64-chatterbox"
-            build-type: "mps"
-          - backend: "mlx-vlm"
-            tag-suffix: "-metal-darwin-arm64-mlx-vlm"
-            build-type: "mps"
-          - backend: "mlx-audio"
-            tag-suffix: "-metal-darwin-arm64-mlx-audio"
-            build-type: "mps"
-          - backend: "stablediffusion-ggml"
-            tag-suffix: "-metal-darwin-arm64-stablediffusion-ggml"
-            build-type: "metal"
-            lang: "go"
-          - backend: "whisper"
-            tag-suffix: "-metal-darwin-arm64-whisper"
-            build-type: "metal"
-            lang: "go"
     with:
-      backend: ${{ matrix.backend }}
-      build-type: ${{ matrix.build-type }}
+      backend: "transformers"
+      build-type: "mps"
       go-version: "1.24.x"
-      tag-suffix: ${{ matrix.tag-suffix }}
-      lang: ${{ matrix.lang || 'python' }}
-      use-pip: ${{ matrix.backend == 'diffusers' }}
+      tag-suffix: "-metal-darwin-arm64-transformers"
+      use-pip: true
+      runs-on: "macOS-14"
+    secrets:
+      dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+      dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+      quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+      quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+  diffusers-darwin:
+    uses: ./.github/workflows/backend_build_darwin.yml
+    with:
+      backend: "diffusers"
+      build-type: "mps"
+      go-version: "1.24.x"
+      tag-suffix: "-metal-darwin-arm64-diffusers"
+      use-pip: true
+      runs-on: "macOS-14"
+    secrets:
+      dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+      dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+      quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+      quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+  mlx-darwin:
+    uses: ./.github/workflows/backend_build_darwin.yml
+    with:
+      backend: "mlx"
+      build-type: "mps"
+      go-version: "1.24.x"
+      tag-suffix: "-metal-darwin-arm64-mlx"
+      runs-on: "macOS-14"
+    secrets:
+      dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
+      dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
+      quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
+      quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
+  mlx-vlm-darwin:
+    uses: ./.github/workflows/backend_build_darwin.yml
+    with:
+      backend: "mlx-vlm"
+      build-type: "mps"
+      go-version: "1.24.x"
+      tag-suffix: "-metal-darwin-arm64-mlx-vlm"
       runs-on: "macOS-14"
     secrets:
       dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
@@ -1109,7 +1050,7 @@ jobs:
           make protogen-go
           make backends/llama-cpp-darwin
       - name: Upload llama-cpp.tar
-        uses: actions/upload-artifact@v5
+        uses: actions/upload-artifact@v4
         with:
           name: llama-cpp-tar
           path: backend-images/llama-cpp.tar
@@ -1119,7 +1060,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Download llama-cpp.tar
-        uses: actions/download-artifact@v6
+        uses: actions/download-artifact@v5
         with:
           name: llama-cpp-tar
           path: .
@@ -1197,7 +1138,7 @@ jobs:
           export PLATFORMARCH=darwin/amd64
           make backends/llama-cpp-darwin
       - name: Upload llama-cpp.tar
-        uses: actions/upload-artifact@v5
+        uses: actions/upload-artifact@v4
         with:
           name: llama-cpp-tar-x86
           path: backend-images/llama-cpp.tar
@@ -1207,7 +1148,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Download llama-cpp.tar
-        uses: actions/download-artifact@v6
+        uses: actions/download-artifact@v5
         with:
           name: llama-cpp-tar-x86
           path: .

.github/workflows/backend_build_darwin.yml

@@ -16,10 +16,6 @@ on:
         description: 'Use pip to install dependencies'
         default: false
         type: boolean
-      lang:
-        description: 'Programming language (e.g. go)'
-        default: 'python'
-        type: string
       go-version:
         description: 'Go version to use'
         default: '1.24.x'
@@ -53,28 +49,28 @@ jobs:
         uses: actions/checkout@v5
         with:
           submodules: true
-
+          
       - name: Setup Go ${{ matrix.go-version }}
         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 make protoc-gen-go protoc-gen-go-grpc libomp llvm
-
+          
       - name: Build ${{ inputs.backend }}-darwin
         run: |
           make protogen-go
-          BACKEND=${{ inputs.backend }} BUILD_TYPE=${{ inputs.build-type }} USE_PIP=${{ inputs.use-pip }} make build-darwin-${{ inputs.lang }}-backend
-
+          BACKEND=${{ inputs.backend }} BUILD_TYPE=${{ inputs.build-type }} USE_PIP=${{ inputs.use-pip }} make build-darwin-python-backend
+          
       - name: Upload ${{ inputs.backend }}.tar
-        uses: actions/upload-artifact@v5
+        uses: actions/upload-artifact@v4
         with:
           name: ${{ inputs.backend }}-tar
           path: backend-images/${{ inputs.backend }}.tar
@@ -85,24 +81,24 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Download ${{ inputs.backend }}.tar
-        uses: actions/download-artifact@v6
+        uses: actions/download-artifact@v5
         with:
           name: ${{ inputs.backend }}-tar
           path: .
-
+          
       - name: Install crane
         run: |
           curl -L https://github.com/google/go-containerregistry/releases/latest/download/go-containerregistry_Linux_x86_64.tar.gz | tar -xz
           sudo mv crane /usr/local/bin/
-
+          
       - name: Log in to DockerHub
         run: |
           echo "${{ secrets.dockerPassword }}" | crane auth login docker.io -u "${{ secrets.dockerUsername }}" --password-stdin
-
+          
       - name: Log in to quay.io
         run: |
           echo "${{ secrets.quayPassword }}" | crane auth login quay.io -u "${{ secrets.quayUsername }}" --password-stdin
-
+          
       - name: Docker meta
         id: meta
         uses: docker/metadata-action@v5
@@ -116,7 +112,7 @@ jobs:
           flavor: |
             latest=auto
             suffix=${{ inputs.tag-suffix }},onlatest=true
-
+            
       - name: Docker meta
         id: quaymeta
         uses: docker/metadata-action@v5
@@ -130,13 +126,13 @@ jobs:
           flavor: |
             latest=auto
             suffix=${{ inputs.tag-suffix }},onlatest=true
-
+            
       - name: Push Docker image (DockerHub)
         run: |
           for tag in $(echo "${{ steps.meta.outputs.tags }}" | tr ',' '\n'); do
             crane push ${{ inputs.backend }}.tar $tag
           done
-
+          
       - name: Push Docker image (Quay)
         run: |
           for tag in $(echo "${{ steps.quaymeta.outputs.tags }}" | tr ',' '\n'); do

.github/workflows/backend_pr.yml

@@ -12,9 +12,7 @@ jobs:
     runs-on: ubuntu-latest
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
-      matrix-darwin: ${{ steps.set-matrix.outputs.matrix-darwin }}
       has-backends: ${{ steps.set-matrix.outputs.has-backends }}
-      has-backends-darwin: ${{ steps.set-matrix.outputs.has-backends-darwin }}
     steps:
       - name: Checkout repository
         uses: actions/checkout@v5
@@ -58,21 +56,3 @@ jobs:
     strategy:
       fail-fast: true
       matrix: ${{ fromJson(needs.generate-matrix.outputs.matrix) }}
-  backend-jobs-darwin:
-    needs: generate-matrix
-    uses: ./.github/workflows/backend_build_darwin.yml
-    if: needs.generate-matrix.outputs.has-backends-darwin == 'true'
-    with:
-      backend: ${{ matrix.backend }}
-      build-type: ${{ matrix.build-type }}
-      go-version: "1.24.x"
-      tag-suffix: ${{ matrix.tag-suffix }}
-      lang: ${{ matrix.lang || 'python' }}
-      use-pip: ${{ matrix.backend == 'diffusers' }}
-      runs-on: "macOS-14"
-    secrets:
-      quayUsername: ${{ secrets.LOCALAI_REGISTRY_USERNAME }}
-      quayPassword: ${{ secrets.LOCALAI_REGISTRY_PASSWORD }}
-    strategy:
-      fail-fast: true
-      matrix: ${{ fromJson(needs.generate-matrix.outputs.matrix-darwin) }}

.github/workflows/build-test.yaml

@@ -17,51 +17,7 @@ jobs:
       - name: Set up Go
         uses: actions/setup-go@v5
         with:
-          go-version: 1.25
+          go-version: 1.23
       - name: Run GoReleaser
         run: |
           make dev-dist
-  launcher-build-darwin:
-    runs-on: macos-latest
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v5
-        with:
-          fetch-depth: 0
-      - name: Set up Go
-        uses: actions/setup-go@v5
-        with:
-          go-version: 1.25
-      - name: Build launcher for macOS ARM64
-        run: |
-          make build-launcher-darwin
-          ls -liah dist
-      - name: Upload macOS launcher artifacts
-        uses: actions/upload-artifact@v5
-        with:
-          name: launcher-macos
-          path: dist/
-          retention-days: 30
-      
-  launcher-build-linux:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v5
-        with:
-          fetch-depth: 0
-      - name: Set up Go
-        uses: actions/setup-go@v5
-        with:
-          go-version: 1.25
-      - name: Build launcher for Linux
-        run: |
-          sudo apt-get update
-          sudo apt-get install golang gcc libgl1-mesa-dev xorg-dev libxkbcommon-dev
-          make build-launcher-linux
-      - name: Upload Linux launcher artifacts
-        uses: actions/upload-artifact@v5
-        with:
-          name: launcher-linux
-          path: local-ai-launcher-linux.tar.xz
-          retention-days: 30

.github/workflows/bump_deps.yaml

@@ -1,10 +1,10 @@
-name: Bump Backend dependencies
+name: Bump dependencies
 on:
   schedule:
     - cron: 0 20 * * *
   workflow_dispatch:
 jobs:
-  bump-backends:
+  bump:
     strategy:
       fail-fast: false
       matrix:

.github/workflows/bump_docs.yaml

@@ -1,10 +1,10 @@
-name: Bump Documentation
+name: Bump dependencies
 on:
   schedule:
     - cron: 0 20 * * *
   workflow_dispatch:
 jobs:
-  bump-docs:
+  bump:
     strategy:
       fail-fast: false
       matrix:

.github/workflows/deploy-explorer.yaml

@@ -33,7 +33,7 @@ jobs:
         run: |
           CGO_ENABLED=0 make build
       - name: rm
-        uses: appleboy/ssh-action@v1.2.3
+        uses: appleboy/ssh-action@v1.2.2
         with:
             host: ${{ secrets.EXPLORER_SSH_HOST }}
             username: ${{ secrets.EXPLORER_SSH_USERNAME }}
@@ -53,7 +53,7 @@ jobs:
             rm: true
             target: ./local-ai
       - name: restarting
-        uses: appleboy/ssh-action@v1.2.3
+        uses: appleboy/ssh-action@v1.2.2
         with:
             host: ${{ secrets.EXPLORER_SSH_HOST }}
             username: ${{ secrets.EXPLORER_SSH_USERNAME }}

.github/workflows/gallery-agent.yaml

@@ -1,119 +0,0 @@
-name: Gallery Agent
-on:
-
-  schedule:
-    - cron: '0 */3 * * *'  # Run every 4 hours
-  workflow_dispatch:
-    inputs:
-      search_term:
-        description: 'Search term for models'
-        required: false
-        default: 'GGUF'
-        type: string
-      limit:
-        description: 'Maximum number of models to process'
-        required: false
-        default: '15'
-        type: string
-      quantization:
-        description: 'Preferred quantization format'
-        required: false
-        default: 'Q4_K_M'
-        type: string
-      max_models:
-        description: 'Maximum number of models to add to the gallery'
-        required: false
-        default: '1'
-        type: string
-jobs:
-  gallery-agent:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v5
-        with:
-          token: ${{ secrets.GITHUB_TOKEN }}
-
-      - name: Set up Go
-        uses: actions/setup-go@v5
-        with:
-          go-version: '1.21'
-
-
-      - name: Run gallery agent
-        env:
-          OPENAI_MODEL: ${{ secrets.OPENAI_MODEL }}
-          OPENAI_KEY: ${{ secrets.OPENAI_KEY }}
-          OPENAI_BASE_URL: ${{ secrets.OPENAI_BASE_URL }}
-          SEARCH_TERM: ${{ github.event.inputs.search_term || 'GGUF' }}
-          LIMIT: ${{ github.event.inputs.limit || '15' }}
-          QUANTIZATION: ${{ github.event.inputs.quantization || 'Q4_K_M' }}
-          MAX_MODELS: ${{ github.event.inputs.max_models || '1' }}
-        run: |
-          export GALLERY_INDEX_PATH=$PWD/gallery/index.yaml
-          go run .github/gallery-agent
-
-      - name: Check for changes
-        id: check_changes
-        run: |
-          if git diff --quiet gallery/index.yaml; then
-            echo "changes=false" >> $GITHUB_OUTPUT
-            echo "No changes detected in gallery/index.yaml"
-          else
-            echo "changes=true" >> $GITHUB_OUTPUT
-            echo "Changes detected in gallery/index.yaml"
-            git diff gallery/index.yaml
-          fi
-
-      - name: Read gallery agent summary
-        id: read_summary
-        if: steps.check_changes.outputs.changes == 'true'
-        run: |
-          if [ -f ".github/gallery-agent/gallery-agent-summary.json" ]; then
-            echo "summary_exists=true" >> $GITHUB_OUTPUT
-            # Extract summary data using jq
-            echo "search_term=$(jq -r '.search_term' .github/gallery-agent/gallery-agent-summary.json)" >> $GITHUB_OUTPUT
-            echo "total_found=$(jq -r '.total_found' .github/gallery-agent/gallery-agent-summary.json)" >> $GITHUB_OUTPUT
-            echo "models_added=$(jq -r '.models_added' .github/gallery-agent/gallery-agent-summary.json)" >> $GITHUB_OUTPUT
-            echo "quantization=$(jq -r '.quantization' .github/gallery-agent/gallery-agent-summary.json)" >> $GITHUB_OUTPUT
-            echo "processing_time=$(jq -r '.processing_time' .github/gallery-agent/gallery-agent-summary.json)" >> $GITHUB_OUTPUT
-            
-            # Create a formatted list of added models with URLs
-            added_models=$(jq -r 'range(0; .added_model_ids | length) as $i | "- [\(.added_model_ids[$i])](\(.added_model_urls[$i]))"' .github/gallery-agent/gallery-agent-summary.json | tr '\n' '\n')
-            echo "added_models<<EOF" >> $GITHUB_OUTPUT
-            echo "$added_models" >> $GITHUB_OUTPUT
-            echo "EOF" >> $GITHUB_OUTPUT
-            rm -f .github/gallery-agent/gallery-agent-summary.json
-          else
-            echo "summary_exists=false" >> $GITHUB_OUTPUT
-          fi
-
-      - name: Create Pull Request
-        if: steps.check_changes.outputs.changes == 'true'
-        uses: peter-evans/create-pull-request@v7
-        with:
-          token: ${{ secrets.UPDATE_BOT_TOKEN }}
-          push-to-fork: ci-forks/LocalAI
-          commit-message: 'chore(model gallery): :robot: add new models via gallery agent'
-          title: 'chore(model gallery): :robot: add ${{ steps.read_summary.outputs.models_added || 0 }} new models via gallery agent'
-          # Branch has to be unique so PRs are not overriding each other
-          branch-suffix: timestamp
-          body: |
-            This PR was automatically created by the gallery agent workflow.
-            
-            **Summary:**
-            - **Search Term:** ${{ steps.read_summary.outputs.search_term || github.event.inputs.search_term || 'GGUF' }}
-            - **Models Found:** ${{ steps.read_summary.outputs.total_found || 'N/A' }}
-            - **Models Added:** ${{ steps.read_summary.outputs.models_added || '0' }}
-            - **Quantization:** ${{ steps.read_summary.outputs.quantization || github.event.inputs.quantization || 'Q4_K_M' }}
-            - **Processing Time:** ${{ steps.read_summary.outputs.processing_time || 'N/A' }}
-            
-            **Added Models:**
-            ${{ steps.read_summary.outputs.added_models || '- No models added' }}
-            
-            **Workflow Details:**
-            - Triggered by: `${{ github.event_name }}`
-            - Run ID: `${{ github.run_id }}`
-            - Commit: `${{ github.sha }}`
-          signoff: true
-          delete-branch: true

.github/workflows/labeler.yml

@@ -9,4 +9,4 @@ jobs:
       pull-requests: write
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/labeler@v6
+    - uses: actions/labeler@v5

.github/workflows/localaibot_automerge.yml

@@ -6,12 +6,11 @@ permissions:
   contents: write
   pull-requests: write
   packages: read
-  issues: write # for Homebrew/actions/post-comment
-  actions: write # to dispatch publish workflow
+
 jobs:
   dependabot:
     runs-on: ubuntu-latest
-    if: ${{ github.actor == 'localai-bot' && !contains(github.event.pull_request.title, 'chore(model gallery):') }}
+    if: ${{ github.actor == 'localai-bot' }}
     steps:
       - name: Checkout repository
         uses: actions/checkout@v5

.github/workflows/notify-models.yaml

@@ -1,27 +1,22 @@
 name: Notifications for new models
 on:
-  pull_request_target:
+  pull_request:
      types:
        - closed
 
-permissions:
-  contents: read
-  pull-requests: read
-
 jobs:
   notify-discord:
     if: ${{ (github.event.pull_request.merged == true) && (contains(github.event.pull_request.labels.*.name, 'area/ai-model')) }}
     env:
-        MODEL_NAME: gemma-3-12b-it-qat
+        MODEL_NAME: gemma-3-12b-it
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v5
       with:
         fetch-depth: 0 # needed to checkout all branches for this Action to work
-        ref: ${{ github.event.pull_request.head.sha }} # Checkout the PR head to get the actual changes
     - uses: mudler/localai-github-action@v1
       with:
-        model: 'gemma-3-12b-it-qat' # Any from models.localai.io, or from huggingface.com with: "huggingface://<repository>/file"
+        model: 'gemma-3-12b-it' # Any from models.localai.io, or from huggingface.com with: "huggingface://<repository>/file"
         # Check the PR diff using the current branch and the base branch of the PR
     - uses: GrantBirki/git-diff-action@v2.8.1
       id: git-diff-action
@@ -84,7 +79,7 @@ jobs:
         args: ${{ steps.summarize.outputs.message }}
     - name: Setup tmate session if fails
       if: ${{ failure() }}
-      uses: mxschmitt/action-tmate@v3.23
+      uses: mxschmitt/action-tmate@v3.22
       with:
         detached: true
         connect-timeout-seconds: 180
@@ -92,13 +87,12 @@ jobs:
   notify-twitter:
     if: ${{ (github.event.pull_request.merged == true) && (contains(github.event.pull_request.labels.*.name, 'area/ai-model')) }}
     env:
-        MODEL_NAME: gemma-3-12b-it-qat
+        MODEL_NAME: gemma-3-12b-it
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v5
       with:
         fetch-depth: 0 # needed to checkout all branches for this Action to work
-        ref: ${{ github.event.pull_request.head.sha }} # Checkout the PR head to get the actual changes
     - name: Start LocalAI
       run: |
         echo "Starting LocalAI..."
@@ -167,7 +161,7 @@ jobs:
         TWITTER_ACCESS_TOKEN_SECRET: ${{ secrets.TWITTER_ACCESS_TOKEN_SECRET }}
     - name: Setup tmate session if fails
       if: ${{ failure() }}
-      uses: mxschmitt/action-tmate@v3.23
+      uses: mxschmitt/action-tmate@v3.22
       with:
         detached: true
         connect-timeout-seconds: 180

.github/workflows/notify-releases.yaml

@@ -11,11 +11,10 @@ jobs:
         RELEASE_BODY: ${{ github.event.release.body }}
         RELEASE_TITLE: ${{ github.event.release.name }}
         RELEASE_TAG_NAME: ${{ github.event.release.tag_name }}
-        MODEL_NAME: gemma-3-12b-it-qat
     steps:
     - uses: mudler/localai-github-action@v1
       with:
-        model: 'gemma-3-12b-it-qat' # Any from models.localai.io, or from huggingface.com with: "huggingface://<repository>/file"
+        model: 'gemma-3-12b-it' # Any from models.localai.io, or from huggingface.com with: "huggingface://<repository>/file"
     - name: Summarize
       id: summarize
       run: |

.github/workflows/release.yaml

@@ -23,42 +23,4 @@ jobs:
           version: v2.11.0
           args: release --clean
         env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-  launcher-build-darwin:
-    runs-on: macos-latest
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v5
-        with:
-          fetch-depth: 0
-      - name: Set up Go
-        uses: actions/setup-go@v5
-        with:
-          go-version: 1.23
-      - name: Build launcher for macOS ARM64
-        run: |
-          make build-launcher-darwin
-      - name: Upload DMG to Release
-        uses: softprops/action-gh-release@v2
-        with:
-          files: ./dist/LocalAI.dmg
-  launcher-build-linux:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v5
-        with:
-          fetch-depth: 0
-      - name: Set up Go
-        uses: actions/setup-go@v5
-        with:
-          go-version: 1.23
-      - name: Build launcher for Linux
-        run: |
-          sudo apt-get update
-          sudo apt-get install golang gcc libgl1-mesa-dev xorg-dev libxkbcommon-dev
-          make build-launcher-linux
-      - name: Upload Linux launcher artifacts
-        uses: softprops/action-gh-release@v2
-        with:
-          files: ./local-ai-launcher-linux.tar.xz
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} 

.github/workflows/secscan.yaml

@@ -18,13 +18,13 @@ jobs:
         if: ${{ github.actor != 'dependabot[bot]' }}
       - name: Run Gosec Security Scanner
         if: ${{ github.actor != 'dependabot[bot]' }}
-        uses: securego/gosec@v2.22.9
+        uses: securego/gosec@v2.22.8
         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@v4
+        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/stalebot.yml

@@ -10,7 +10,7 @@ jobs:
   stale:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/stale@5f858e3efba33a5ca4407a664cc011ad407f2008 # v9
+      - uses: actions/stale@5bef64f19d7facfb25b37b414482c7164d639639 # v9
         with:
           stale-issue-message: 'This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 5 days.'
           stale-pr-message: 'This PR is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 10 days.'

.github/workflows/test.yml

@@ -21,7 +21,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        go-version: ['1.25.x']
+        go-version: ['1.21.x']
     steps:
       - name: Free Disk Space (Ubuntu)
         uses: jlumbroso/free-disk-space@main
@@ -124,7 +124,7 @@ jobs:
           PATH="$PATH:/root/go/bin" GO_TAGS="tts" make --jobs 5 --output-sync=target test
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.23
+        uses: mxschmitt/action-tmate@v3.22
         with:
           detached: true
           connect-timeout-seconds: 180
@@ -183,7 +183,7 @@ jobs:
             PATH="$PATH:$HOME/go/bin" make backends/local-store backends/silero-vad backends/llama-cpp backends/whisper backends/piper backends/stablediffusion-ggml docker-build-aio e2e-aio
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.23
+        uses: mxschmitt/action-tmate@v3.22
         with:
           detached: true
           connect-timeout-seconds: 180
@@ -193,7 +193,7 @@ jobs:
     runs-on: macOS-14
     strategy:
       matrix:
-        go-version: ['1.25.x']
+        go-version: ['1.21.x']
     steps:
       - name: Clone
         uses: actions/checkout@v5
@@ -226,7 +226,7 @@ jobs:
           PATH="$PATH:$HOME/go/bin" BUILD_TYPE="GITHUB_CI_HAS_BROKEN_METAL" CMAKE_ARGS="-DGGML_F16C=OFF -DGGML_AVX512=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF" make --jobs 4 --output-sync=target test
       - name: Setup tmate session if tests fail
         if: ${{ failure() }}
-        uses: mxschmitt/action-tmate@v3.23
+        uses: mxschmitt/action-tmate@v3.22
         with:
           detached: true
           connect-timeout-seconds: 180

.gitignore

@@ -24,7 +24,7 @@ go-bert
 
 # LocalAI build binary
 LocalAI
-/local-ai
+local-ai
 # prevent above rules from omitting the helm chart
 !charts/*
 # prevent above rules from omitting the api/localai folder

.gitmodules

@@ -1,3 +1,6 @@
 [submodule "docs/themes/hugo-theme-relearn"]
 	path = docs/themes/hugo-theme-relearn
 	url = https://github.com/McShelby/hugo-theme-relearn.git
+[submodule "docs/themes/lotusdocs"]
+	path = docs/themes/lotusdocs
+	url = https://github.com/colinwilson/lotusdocs

.goreleaser.yaml

@@ -8,7 +8,7 @@ source:
   enabled: true
   name_template: '{{ .ProjectName }}-{{ .Tag }}-source'
 builds:
-  - main: ./cmd/local-ai
+  -
     env:
       - CGO_ENABLED=0
     ldflags:

CONTRIBUTING.md

@@ -30,7 +30,6 @@ Thank you for your interest in contributing to LocalAI! We appreciate your time
 3. Install the required dependencies ( see https://localai.io/basics/build/#build-localai-locally )
 4. Build LocalAI: `make build`
 5. Run LocalAI: `./local-ai`
-6. To Build and live reload: `make build-dev`
 
 ## Contributing
 
@@ -77,7 +76,7 @@ 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 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.

Dockerfile

@@ -78,16 +78,6 @@ RUN <<EOT bash
     fi
 EOT
 
-# https://github.com/NVIDIA/Isaac-GR00T/issues/343
-RUN <<EOT bash
-    if [ "${BUILD_TYPE}" = "cublas" ] && [ "${TARGETARCH}" = "arm64" ]; then
-        wget https://developer.download.nvidia.com/compute/cudss/0.6.0/local_installers/cudss-local-tegra-repo-ubuntu2204-0.6.0_0.6.0-1_arm64.deb && \
-        dpkg -i cudss-local-tegra-repo-ubuntu2204-0.6.0_0.6.0-1_arm64.deb && \
-        cp /var/cudss-local-tegra-repo-ubuntu2204-0.6.0/cudss-*-keyring.gpg /usr/share/keyrings/ && \
-        apt-get update && apt-get -y install cudss
-    fi
-EOT
-
 # If we are building with clblas support, we need the libraries for the builds
 RUN if [ "${BUILD_TYPE}" = "clblas" ] && [ "${SKIP_DRIVERS}" = "false" ]; then \
         apt-get update && \
@@ -110,10 +100,6 @@ RUN if [ "${BUILD_TYPE}" = "hipblas" ] && [ "${SKIP_DRIVERS}" = "false" ]; then
         ldconfig \
     ; fi
 
-RUN if [ "${BUILD_TYPE}" = "hipblas" ]; then \
-    ln -s /opt/rocm-**/lib/llvm/lib/libomp.so /usr/lib/libomp.so \
-    ; fi
-
 RUN expr "${BUILD_TYPE}" = intel && echo "intel" > /run/localai/capability || echo "not intel"
 
 # Cuda

Makefile

@@ -2,7 +2,6 @@ GOCMD=go
 GOTEST=$(GOCMD) test
 GOVET=$(GOCMD) vet
 BINARY_NAME=local-ai
-LAUNCHER_BINARY_NAME=local-ai-launcher
 
 GORELEASER?=
 
@@ -91,21 +90,7 @@ build: protogen-go install-go-tools ## Build the project
 	$(info ${GREEN}I LD_FLAGS: ${YELLOW}$(LD_FLAGS)${RESET})
 	$(info ${GREEN}I UPX: ${YELLOW}$(UPX)${RESET})
 	rm -rf $(BINARY_NAME) || true
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o $(BINARY_NAME) ./cmd/local-ai
-
-build-launcher: ## Build the launcher application
-	$(info ${GREEN}I local-ai launcher build info:${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})
-	rm -rf $(LAUNCHER_BINARY_NAME) || true
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o $(LAUNCHER_BINARY_NAME) ./cmd/launcher
-
-build-all: build build-launcher ## Build both server and launcher
-
-build-dev: ## Run LocalAI in dev mode with live reload
-	@command -v air >/dev/null 2>&1 || go install github.com/air-verse/air@latest
-	air -c .air.toml
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o $(BINARY_NAME) ./
 
 dev-dist:
 	$(GORELEASER) build --snapshot --clean
@@ -121,8 +106,8 @@ run: ## run local-ai
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GOCMD) run ./
 
 test-models/testmodel.ggml:
-	mkdir -p test-models
-	mkdir -p test-dir
+	mkdir test-models
+	mkdir test-dir
 	wget -q https://huggingface.co/mradermacher/gpt2-alpaca-gpt4-GGUF/resolve/main/gpt2-alpaca-gpt4.Q4_K_M.gguf -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
@@ -373,22 +358,13 @@ backends/kitten-tts: docker-build-kitten-tts docker-save-kitten-tts build
 backends/kokoro: docker-build-kokoro docker-save-kokoro build
 	./local-ai backends install "ocifile://$(abspath ./backend-images/kokoro.tar)"
 
-backends/chatterbox: docker-build-chatterbox docker-save-chatterbox build
-	./local-ai backends install "ocifile://$(abspath ./backend-images/chatterbox.tar)"
-
 backends/llama-cpp-darwin: build
 	bash ./scripts/build/llama-cpp-darwin.sh
 	./local-ai backends install "ocifile://$(abspath ./backend-images/llama-cpp.tar)"
 
-backends/neutts: docker-build-neutts docker-save-neutts build
-	./local-ai backends install "ocifile://$(abspath ./backend-images/neutts.tar)"
-
 build-darwin-python-backend: build
 	bash ./scripts/build/python-darwin.sh
 
-build-darwin-go-backend: build
-	bash ./scripts/build/golang-darwin.sh
-
 backends/mlx:
 	BACKEND=mlx $(MAKE) build-darwin-python-backend
 	./local-ai backends install "ocifile://$(abspath ./backend-images/mlx.tar)"
@@ -401,14 +377,6 @@ backends/mlx-vlm:
 	BACKEND=mlx-vlm $(MAKE) build-darwin-python-backend
 	./local-ai backends install "ocifile://$(abspath ./backend-images/mlx-vlm.tar)"
 
-backends/mlx-audio:
-	BACKEND=mlx-audio $(MAKE) build-darwin-python-backend
-	./local-ai backends install "ocifile://$(abspath ./backend-images/mlx-audio.tar)"
-
-backends/stablediffusion-ggml-darwin:
-	BACKEND=stablediffusion-ggml BUILD_TYPE=metal $(MAKE) build-darwin-go-backend
-	./local-ai backends install "ocifile://$(abspath ./backend-images/stablediffusion-ggml.tar)"
-
 backend-images:
 	mkdir -p backend-images
 
@@ -436,15 +404,6 @@ docker-build-kitten-tts:
 docker-save-kitten-tts: backend-images
 	docker save local-ai-backend:kitten-tts -o backend-images/kitten-tts.tar
 
-docker-save-chatterbox: backend-images
-	docker save local-ai-backend:chatterbox -o backend-images/chatterbox.tar
-
-docker-build-neutts:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:neutts -f backend/Dockerfile.python --build-arg BACKEND=neutts ./backend
-
-docker-save-neutts: backend-images
-	docker save local-ai-backend:neutts -o backend-images/neutts.tar
-
 docker-build-kokoro:
 	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:kokoro -f backend/Dockerfile.python --build-arg BACKEND=kokoro ./backend
 
@@ -512,7 +471,7 @@ docker-build-bark:
 	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:bark -f backend/Dockerfile.python --build-arg BACKEND=bark .
 
 docker-build-chatterbox:
-	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:chatterbox -f backend/Dockerfile.python --build-arg BACKEND=chatterbox ./backend
+	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:chatterbox -f backend/Dockerfile.python --build-arg BACKEND=chatterbox .
 
 docker-build-exllama2:
 	docker build --build-arg BUILD_TYPE=$(BUILD_TYPE) --build-arg BASE_IMAGE=$(BASE_IMAGE) -t local-ai-backend:exllama2 -f backend/Dockerfile.python --build-arg BACKEND=exllama2 .
@@ -548,19 +507,3 @@ docs-clean:
 .PHONY: docs
 docs: docs/static/gallery.html
 	cd docs && hugo serve
-
-########################################################
-## Platform-specific builds
-########################################################
-
-## fyne cross-platform build
-build-launcher-darwin: build-launcher
-	go run github.com/tiagomelo/macos-dmg-creator/cmd/createdmg@latest \
-	--appName "LocalAI" \
-	--appBinaryPath "$(LAUNCHER_BINARY_NAME)" \
-	--bundleIdentifier "com.localai.launcher" \
-	--iconPath "core/http/static/logo.png" \
-	--outputDir "dist/"
-
-build-launcher-linux:
-	cd cmd/launcher && go run fyne.io/tools/cmd/fyne@latest package -os linux -icon ../../core/http/static/logo.png --executable $(LAUNCHER_BINARY_NAME)-linux && mv launcher.tar.xz ../../$(LAUNCHER_BINARY_NAME)-linux.tar.xz

README.md

@@ -43,7 +43,7 @@
 
 > :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/) [🖼️ Models](https://models.localai.io/) [🚀 Roadmap](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap) [🛫 Examples](https://github.com/mudler/LocalAI-examples) Try on 
+> [💻 Quickstart](https://localai.io/basics/getting_started/) [🖼️ Models](https://models.localai.io/) [🚀 Roadmap](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap) [🥽 Demo](https://demo.localai.io) [🌍 Explorer](https://explorer.localai.io) [🛫 Examples](https://github.com/mudler/LocalAI-examples) Try on 
 [![Telegram](https://img.shields.io/badge/Telegram-2CA5E0?style=for-the-badge&logo=telegram&logoColor=white)](https://t.me/localaiofficial_bot)
 
 [![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)
@@ -110,23 +110,8 @@ curl https://localai.io/install.sh | sh
 
 For more installation options, see [Installer Options](https://localai.io/docs/advanced/installer/).
 
-### macOS Download:
-
-<a href="https://github.com/mudler/LocalAI/releases/latest/download/LocalAI.dmg">
-  <img src="https://img.shields.io/badge/Download-macOS-blue?style=for-the-badge&logo=apple&logoColor=white" alt="Download LocalAI for macOS"/>
-</a>
-
-> Note: the DMGs are not signed by Apple as quarantined. See https://github.com/mudler/LocalAI/issues/6268 for a workaround, fix is tracked here: https://github.com/mudler/LocalAI/issues/6244
-
 Or run with docker:
 
-> **💡 Docker Run vs Docker Start**
-> 
-> - `docker run` creates and starts a new container. If a container with the same name already exists, this command will fail.
-> - `docker start` starts an existing container that was previously created with `docker run`.
-> 
-> If you've already run LocalAI before and want to start it again, use: `docker start -i local-ai`
-
 ### CPU only image:
 
 ```bash
@@ -202,12 +187,10 @@ local-ai run oci://localai/phi-2:latest
 
 > ⚡ **Automatic Backend Detection**: When you install models from the gallery or YAML files, LocalAI automatically detects your system's GPU capabilities (NVIDIA, AMD, Intel) and downloads the appropriate backend. For advanced configuration options, see [GPU Acceleration](https://localai.io/features/gpu-acceleration/#automatic-backend-detection).
 
-For more information, see [💻 Getting started](https://localai.io/basics/getting_started/index.html), if you are interested in our roadmap items and future enhancements, you can see the [Issues labeled as Roadmap here](https://github.com/mudler/LocalAI/issues?q=is%3Aissue+is%3Aopen+label%3Aroadmap)
+For more information, see [💻 Getting started](https://localai.io/basics/getting_started/index.html)
 
 ## 📰 Latest project news
 
-- October 2025: 🔌 [Model Context Protocol (MCP)](https://localai.io/docs/features/mcp/) support added for agentic capabilities with external tools
-- September 2025: New Launcher application for MacOS and Linux, extended support to many backends for Mac and Nvidia L4T devices. Models: Added MLX-Audio, WAN 2.2. WebUI improvements and Python-based backends now ships portable python environments.
 - August 2025: MLX, MLX-VLM, Diffusers and llama.cpp are now supported on Mac M1/M2/M3+ chips ( with `development` suffix in the gallery ): https://github.com/mudler/LocalAI/pull/6049 https://github.com/mudler/LocalAI/pull/6119 https://github.com/mudler/LocalAI/pull/6121 https://github.com/mudler/LocalAI/pull/6060
 - July/August 2025: 🔍 [Object Detection](https://localai.io/features/object-detection/) added to the API featuring [rf-detr](https://github.com/roboflow/rf-detr)
 - July 2025: All backends migrated outside of the main binary. LocalAI is now more lightweight, small, and automatically downloads the required backend to run the model. [Read the release notes](https://github.com/mudler/LocalAI/releases/tag/v3.2.0)
@@ -246,65 +229,10 @@ Roadmap items: [List of issues](https://github.com/mudler/LocalAI/issues?q=is%3A
 - 🔍 [Object Detection](https://localai.io/features/object-detection/)
 - 📈 [Reranker API](https://localai.io/features/reranker/)
 - 🆕🖧 [P2P Inferencing](https://localai.io/features/distribute/)
-- 🆕🔌 [Model Context Protocol (MCP)](https://localai.io/docs/features/mcp/) - Agentic capabilities with external tools and [LocalAGI's Agentic capabilities](https://github.com/mudler/LocalAGI)
+- [Agentic capabilities](https://github.com/mudler/LocalAGI)
 - 🔊 Voice activity detection (Silero-VAD support)
 - 🌍 Integrated WebUI!
 
-## 🧩 Supported Backends & Acceleration
-
-LocalAI supports a comprehensive range of AI backends with multiple acceleration options:
-
-### Text Generation & Language Models
-| Backend | Description | Acceleration Support |
-|---------|-------------|---------------------|
-| **llama.cpp** | LLM inference in C/C++ | CUDA 11/12, ROCm, Intel SYCL, Vulkan, Metal, CPU |
-| **vLLM** | Fast LLM inference with PagedAttention | CUDA 12, ROCm, Intel |
-| **transformers** | HuggingFace transformers framework | CUDA 11/12, ROCm, Intel, CPU |
-| **exllama2** | GPTQ inference library | CUDA 12 |
-| **MLX** | Apple Silicon LLM inference | Metal (M1/M2/M3+) |
-| **MLX-VLM** | Apple Silicon Vision-Language Models | Metal (M1/M2/M3+) |
-
-### Audio & Speech Processing
-| Backend | Description | Acceleration Support |
-|---------|-------------|---------------------|
-| **whisper.cpp** | OpenAI Whisper in C/C++ | CUDA 12, ROCm, Intel SYCL, Vulkan, CPU |
-| **faster-whisper** | Fast Whisper with CTranslate2 | CUDA 12, ROCm, Intel, CPU |
-| **bark** | Text-to-audio generation | CUDA 12, ROCm, Intel |
-| **bark-cpp** | C++ implementation of Bark | CUDA, Metal, CPU |
-| **coqui** | Advanced TTS with 1100+ languages | CUDA 12, ROCm, Intel, CPU |
-| **kokoro** | Lightweight TTS model | CUDA 12, ROCm, Intel, CPU |
-| **chatterbox** | Production-grade TTS | CUDA 11/12, CPU |
-| **piper** | Fast neural TTS system | CPU |
-| **kitten-tts** | Kitten TTS models | CPU |
-| **silero-vad** | Voice Activity Detection | CPU |
-| **neutts** | Text-to-speech with voice cloning | CUDA 12, ROCm, CPU |
-
-### Image & Video Generation
-| Backend | Description | Acceleration Support |
-|---------|-------------|---------------------|
-| **stablediffusion.cpp** | Stable Diffusion in C/C++ | CUDA 12, Intel SYCL, Vulkan, CPU |
-| **diffusers** | HuggingFace diffusion models | CUDA 11/12, ROCm, Intel, Metal, CPU |
-
-### Specialized AI Tasks
-| Backend | Description | Acceleration Support |
-|---------|-------------|---------------------|
-| **rfdetr** | Real-time object detection | CUDA 12, Intel, CPU |
-| **rerankers** | Document reranking API | CUDA 11/12, ROCm, Intel, CPU |
-| **local-store** | Vector database | CPU |
-| **huggingface** | HuggingFace API integration | API-based |
-
-### Hardware Acceleration Matrix
-
-| Acceleration Type | Supported Backends | Hardware Support |
-|-------------------|-------------------|------------------|
-| **NVIDIA CUDA 11** | llama.cpp, whisper, stablediffusion, diffusers, rerankers, bark, chatterbox | Nvidia hardware |
-| **NVIDIA CUDA 12** | All CUDA-compatible backends | Nvidia hardware |
-| **AMD ROCm** | llama.cpp, whisper, vllm, transformers, diffusers, rerankers, coqui, kokoro, bark, neutts | AMD Graphics |
-| **Intel oneAPI** | llama.cpp, whisper, stablediffusion, vllm, transformers, diffusers, rfdetr, rerankers, exllama2, coqui, kokoro, bark | Intel Arc, Intel iGPUs |
-| **Apple Metal** | llama.cpp, whisper, diffusers, MLX, MLX-VLM, bark-cpp | Apple M1/M2/M3+ |
-| **Vulkan** | llama.cpp, whisper, stablediffusion | Cross-platform GPUs |
-| **NVIDIA Jetson** | llama.cpp, whisper, stablediffusion, diffusers, rfdetr | ARM64 embedded AI |
-| **CPU Optimized** | All backends | AVX/AVX2/AVX512, quantization support |
 
 ### 🔗 Community and integrations
 
@@ -316,18 +244,9 @@ WebUIs:
 - 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
 
-Agentic Libraries:
-- https://github.com/mudler/cogito
-
-MCPs:
-- https://github.com/mudler/MCPs
-
 Model galleries
 - https://github.com/go-skynet/model-gallery
 
-Voice:
-- https://github.com/richiejp/VoxInput
-
 Other:
 - Helm chart https://github.com/go-skynet/helm-charts
 - VSCode extension https://github.com/badgooooor/localai-vscode-plugin

aio/cpu/vision.yaml

@@ -2,10 +2,10 @@ context_size: 4096
 f16: true
 backend: llama-cpp
 mmap: true
-mmproj: minicpm-v-4_5-mmproj-f16.gguf
+mmproj: minicpm-v-2_6-mmproj-f16.gguf
 name: gpt-4o
 parameters:
-  model: minicpm-v-4_5-Q4_K_M.gguf
+  model: minicpm-v-2_6-Q4_K_M.gguf
 stopwords:
 - <|im_end|>
 - <dummy32000>
@@ -42,9 +42,9 @@ template:
     <|im_start|>assistant
 
 download_files:
-- filename: minicpm-v-4_5-Q4_K_M.gguf
-  sha256: c1c3c33100b15b4caf7319acce4e23c0eb0ce1cbd12f70e8d24f05aa67b7512f
-  uri: huggingface://openbmb/MiniCPM-V-4_5-gguf/ggml-model-Q4_K_M.gguf
-- filename: minicpm-v-4_5-mmproj-f16.gguf
-  uri: huggingface://openbmb/MiniCPM-V-4_5-gguf/mmproj-model-f16.gguf
-  sha256: 7a7225a32e8d453aaa3d22d8c579b5bf833c253f784cdb05c99c9a76fd616df8
+- filename: minicpm-v-2_6-Q4_K_M.gguf
+  sha256: 3a4078d53b46f22989adbf998ce5a3fd090b6541f112d7e936eb4204a04100b1
+  uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/ggml-model-Q4_K_M.gguf
+- filename: minicpm-v-2_6-mmproj-f16.gguf
+  uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/mmproj-model-f16.gguf
+  sha256: 4485f68a0f1aa404c391e788ea88ea653c100d8e98fe572698f701e5809711fd

aio/gpu-8g/vision.yaml

@@ -2,10 +2,10 @@ context_size: 4096
 backend: llama-cpp
 f16: true
 mmap: true
-mmproj: minicpm-v-4_5-mmproj-f16.gguf
+mmproj: minicpm-v-2_6-mmproj-f16.gguf
 name: gpt-4o
 parameters:
-  model: minicpm-v-4_5-Q4_K_M.gguf
+  model: minicpm-v-2_6-Q4_K_M.gguf
 stopwords:
 - <|im_end|>
 - <dummy32000>
@@ -42,9 +42,9 @@ template:
     <|im_start|>assistant
 
 download_files:
-- filename: minicpm-v-4_5-Q4_K_M.gguf
-  sha256: c1c3c33100b15b4caf7319acce4e23c0eb0ce1cbd12f70e8d24f05aa67b7512f
-  uri: huggingface://openbmb/MiniCPM-V-4_5-gguf/ggml-model-Q4_K_M.gguf
-- filename: minicpm-v-4_5-mmproj-f16.gguf
-  uri: huggingface://openbmb/MiniCPM-V-4_5-gguf/mmproj-model-f16.gguf
-  sha256: 7a7225a32e8d453aaa3d22d8c579b5bf833c253f784cdb05c99c9a76fd616df8
+- filename: minicpm-v-2_6-Q4_K_M.gguf
+  sha256: 3a4078d53b46f22989adbf998ce5a3fd090b6541f112d7e936eb4204a04100b1
+  uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/ggml-model-Q4_K_M.gguf
+- filename: minicpm-v-2_6-mmproj-f16.gguf
+  uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/mmproj-model-f16.gguf
+  sha256: 4485f68a0f1aa404c391e788ea88ea653c100d8e98fe572698f701e5809711fd

aio/intel/vision.yaml

@@ -2,10 +2,10 @@ context_size: 4096
 backend: llama-cpp
 f16: true
 mmap: true
-mmproj: minicpm-v-4_5-mmproj-f16.gguf
+mmproj: minicpm-v-2_6-mmproj-f16.gguf
 name: gpt-4o
 parameters:
-  model: minicpm-v-4_5-Q4_K_M.gguf
+  model: minicpm-v-2_6-Q4_K_M.gguf
 stopwords:
 - <|im_end|>
 - <dummy32000>
@@ -43,9 +43,9 @@ template:
 
 
 download_files:
-- filename: minicpm-v-4_5-Q4_K_M.gguf
-  sha256: c1c3c33100b15b4caf7319acce4e23c0eb0ce1cbd12f70e8d24f05aa67b7512f
-  uri: huggingface://openbmb/MiniCPM-V-4_5-gguf/ggml-model-Q4_K_M.gguf
-- filename: minicpm-v-4_5-mmproj-f16.gguf
-  uri: huggingface://openbmb/MiniCPM-V-4_5-gguf/mmproj-model-f16.gguf
-  sha256: 7a7225a32e8d453aaa3d22d8c579b5bf833c253f784cdb05c99c9a76fd616df8
+- filename: minicpm-v-2_6-Q4_K_M.gguf
+  sha256: 3a4078d53b46f22989adbf998ce5a3fd090b6541f112d7e936eb4204a04100b1
+  uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/ggml-model-Q4_K_M.gguf
+- filename: minicpm-v-2_6-mmproj-f16.gguf
+  uri: huggingface://openbmb/MiniCPM-V-2_6-gguf/mmproj-model-f16.gguf
+  sha256: 4485f68a0f1aa404c391e788ea88ea653c100d8e98fe572698f701e5809711fd

backend/Dockerfile.llama-cpp

@@ -197,7 +197,7 @@ EOT
 
 
 # Copy libraries using a script to handle architecture differences
-RUN make -BC /LocalAI/backend/cpp/llama-cpp package
+RUN make -C /LocalAI/backend/cpp/llama-cpp package
 
 
 FROM scratch

backend/Dockerfile.python

@@ -28,7 +28,7 @@ RUN apt-get update && \
         curl python3-pip \
         python-is-python3 \
         python3-dev llvm \
-        python3-venv make cmake && \
+        python3-venv make && \
     apt-get clean && \
     rm -rf /var/lib/apt/lists/* && \
     pip install --upgrade pip

backend/README.md

@@ -1,213 +0,0 @@
-# LocalAI Backend Architecture
-
-This directory contains the core backend infrastructure for LocalAI, including the gRPC protocol definition, multi-language Dockerfiles, and language-specific backend implementations.
-
-## Overview
-
-LocalAI uses a unified gRPC-based architecture that allows different programming languages to implement AI backends while maintaining consistent interfaces and capabilities. The backend system supports multiple hardware acceleration targets and provides a standardized way to integrate various AI models and frameworks.
-
-## Architecture Components
-
-### 1. Protocol Definition (`backend.proto`)
-
-The `backend.proto` file defines the gRPC service interface that all backends must implement. This ensures consistency across different language implementations and provides a contract for communication between LocalAI core and backend services.
-
-#### Core Services
-
-- **Text Generation**: `Predict`, `PredictStream` for LLM inference
-- **Embeddings**: `Embedding` for text vectorization
-- **Image Generation**: `GenerateImage` for stable diffusion and image models
-- **Audio Processing**: `AudioTranscription`, `TTS`, `SoundGeneration`
-- **Video Generation**: `GenerateVideo` for video synthesis
-- **Object Detection**: `Detect` for computer vision tasks
-- **Vector Storage**: `StoresSet`, `StoresGet`, `StoresFind` for RAG operations
-- **Reranking**: `Rerank` for document relevance scoring
-- **Voice Activity Detection**: `VAD` for audio segmentation
-
-#### Key Message Types
-
-- **`PredictOptions`**: Comprehensive configuration for text generation
-- **`ModelOptions`**: Model loading and configuration parameters
-- **`Result`**: Standardized response format
-- **`StatusResponse`**: Backend health and memory usage information
-
-### 2. Multi-Language Dockerfiles
-
-The backend system provides language-specific Dockerfiles that handle the build environment and dependencies for different programming languages:
-
-- `Dockerfile.python`
-- `Dockerfile.golang`
-- `Dockerfile.llama-cpp`
-
-### 3. Language-Specific Implementations
-
-#### Python Backends (`python/`)
-- **transformers**: Hugging Face Transformers framework
-- **vllm**: High-performance LLM inference
-- **mlx**: Apple Silicon optimization
-- **diffusers**: Stable Diffusion models
-- **Audio**: bark, coqui, faster-whisper, kitten-tts
-- **Vision**: mlx-vlm, rfdetr
-- **Specialized**: rerankers, chatterbox, kokoro
-
-#### Go Backends (`go/`)
-- **whisper**: OpenAI Whisper speech recognition in Go with GGML cpp backend (whisper.cpp)
-- **stablediffusion-ggml**: Stable Diffusion in Go with GGML Cpp backend
-- **huggingface**: Hugging Face model integration
-- **piper**: Text-to-speech synthesis Golang with C bindings using rhaspy/piper
-- **bark-cpp**: Bark TTS models Golang with Cpp bindings
-- **local-store**: Vector storage backend
-
-#### C++ Backends (`cpp/`)
-- **llama-cpp**: Llama.cpp integration
-- **grpc**: GRPC utilities and helpers
-
-## Hardware Acceleration Support
-
-### CUDA (NVIDIA)
-- **Versions**: CUDA 11.x, 12.x
-- **Features**: cuBLAS, cuDNN, TensorRT optimization
-- **Targets**: x86_64, ARM64 (Jetson)
-
-### ROCm (AMD)
-- **Features**: HIP, rocBLAS, MIOpen
-- **Targets**: AMD GPUs with ROCm support
-
-### Intel
-- **Features**: oneAPI, Intel Extension for PyTorch
-- **Targets**: Intel GPUs, XPUs, CPUs
-
-### Vulkan
-- **Features**: Cross-platform GPU acceleration
-- **Targets**: Windows, Linux, Android, macOS
-
-### Apple Silicon
-- **Features**: MLX framework, Metal Performance Shaders
-- **Targets**: M1/M2/M3 Macs
-
-## Backend Registry (`index.yaml`)
-
-The `index.yaml` file serves as a central registry for all available backends, providing:
-
-- **Metadata**: Name, description, license, icons
-- **Capabilities**: Hardware targets and optimization profiles
-- **Tags**: Categorization for discovery
-- **URLs**: Source code and documentation links
-
-## Building Backends
-
-### Prerequisites
-- Docker with multi-architecture support
-- Appropriate hardware drivers (CUDA, ROCm, etc.)
-- Build tools (make, cmake, compilers)
-
-### Build Commands
-
-Example of build commands with Docker
-
-```bash
-# Build Python backend
-docker build -f backend/Dockerfile.python \
-  --build-arg BACKEND=transformers \
-  --build-arg BUILD_TYPE=cublas12 \
-  --build-arg CUDA_MAJOR_VERSION=12 \
-  --build-arg CUDA_MINOR_VERSION=0 \
-  -t localai-backend-transformers .
-
-# Build Go backend
-docker build -f backend/Dockerfile.golang \
-  --build-arg BACKEND=whisper \
-  --build-arg BUILD_TYPE=cpu \
-  -t localai-backend-whisper .
-
-# Build C++ backend
-docker build -f backend/Dockerfile.llama-cpp \
-  --build-arg BACKEND=llama-cpp \
-  --build-arg BUILD_TYPE=cublas12 \
-  -t localai-backend-llama-cpp .
-```
-
-For ARM64/Mac builds, docker can't be used, and the makefile in the respective backend has to be used.
-
-### Build Types
-
-- **`cpu`**: CPU-only optimization
-- **`cublas11`**: CUDA 11.x with cuBLAS
-- **`cublas12`**: CUDA 12.x with cuBLAS
-- **`hipblas`**: ROCm with rocBLAS
-- **`intel`**: Intel oneAPI optimization
-- **`vulkan`**: Vulkan-based acceleration
-- **`metal`**: Apple Metal optimization
-
-## Backend Development
-
-### Creating a New Backend
-
-1. **Choose Language**: Select Python, Go, or C++ based on requirements
-2. **Implement Interface**: Implement the gRPC service defined in `backend.proto`
-3. **Add Dependencies**: Create appropriate requirements files
-4. **Configure Build**: Set up Dockerfile and build scripts
-5. **Register Backend**: Add entry to `index.yaml`
-6. **Test Integration**: Verify gRPC communication and functionality
-
-### Backend Structure
-
-```
-backend-name/
-├── backend.py/go/cpp    # Main implementation
-├── requirements.txt      # Dependencies
-├── Dockerfile           # Build configuration
-├── install.sh           # Installation script
-├── run.sh              # Execution script
-├── test.sh             # Test script
-└── README.md           # Backend documentation
-```
-
-### Required gRPC Methods
-
-At minimum, backends must implement:
-- `Health()` - Service health check
-- `LoadModel()` - Model loading and initialization
-- `Predict()` - Main inference endpoint
-- `Status()` - Backend status and metrics
-
-## Integration with LocalAI Core
-
-Backends communicate with LocalAI core through gRPC:
-
-1. **Service Discovery**: Core discovers available backends
-2. **Model Loading**: Core requests model loading via `LoadModel`
-3. **Inference**: Core sends requests via `Predict` or specialized endpoints
-4. **Streaming**: Core handles streaming responses for real-time generation
-5. **Monitoring**: Core tracks backend health and performance
-
-## Performance Optimization
-
-### Memory Management
-- **Model Caching**: Efficient model loading and caching
-- **Batch Processing**: Optimize for multiple concurrent requests
-- **Memory Pinning**: GPU memory optimization for CUDA/ROCm
-
-### Hardware Utilization
-- **Multi-GPU**: Support for tensor parallelism
-- **Mixed Precision**: FP16/BF16 for memory efficiency
-- **Kernel Fusion**: Optimized CUDA/ROCm kernels
-
-## Troubleshooting
-
-### Common Issues
-
-1. **GRPC Connection**: Verify backend service is running and accessible
-2. **Model Loading**: Check model paths and dependencies
-3. **Hardware Detection**: Ensure appropriate drivers and libraries
-4. **Memory Issues**: Monitor GPU memory usage and model sizes
-
-## Contributing
-
-When contributing to the backend system:
-
-1. **Follow Protocol**: Implement the exact gRPC interface
-2. **Add Tests**: Include comprehensive test coverage
-3. **Document**: Provide clear usage examples
-4. **Optimize**: Consider performance and resource usage
-5. **Validate**: Test across different hardware targets

backend/backend.proto

@@ -154,10 +154,6 @@ message PredictOptions {
   repeated string Videos = 45;
   repeated string Audios = 46;
   string CorrelationId = 47;
-  string Tools = 48;  // JSON array of available tools/functions for tool calling
-  string ToolChoice = 49;  // JSON string or object specifying tool choice behavior
-  int32 Logprobs = 50;  // Number of top logprobs to return (maps to OpenAI logprobs parameter)
-  int32 TopLogprobs = 51;  // Number of top logprobs to return per token (maps to OpenAI top_logprobs parameter)
 }
 
 // The response message containing the result
@@ -168,7 +164,6 @@ message Reply {
   double timing_prompt_processing = 4;
   double timing_token_generation = 5;
   bytes audio = 6;
-  bytes logprobs = 7;  // JSON-encoded logprobs data matching OpenAI format
 }
 
 message GrammarTrigger {
@@ -247,7 +242,7 @@ message ModelOptions {
 
   string Type = 49;
 
-  string FlashAttention = 56;
+  bool FlashAttention = 56;
   bool NoKVOffload = 57;
 
   string ModelPath = 59;
@@ -281,7 +276,6 @@ message TranscriptRequest {
   string language = 3;
   uint32 threads = 4;
   bool translate = 5;
-  bool diarize = 6;
 }
 
 message TranscriptResult {
@@ -311,24 +305,22 @@ message GenerateImageRequest {
   // Diffusers
   string EnableParameters = 10;
   int32 CLIPSkip = 11;
-
+  
   // Reference images for models that support them (e.g., Flux Kontext)
   repeated string ref_images = 12;
 }
 
 message GenerateVideoRequest {
   string prompt = 1;
-  string negative_prompt = 2;  // Negative prompt for video generation
-  string start_image = 3;  // Path or base64 encoded image for the start frame
-  string end_image = 4;    // Path or base64 encoded image for the end frame
-  int32 width = 5;
-  int32 height = 6;
-  int32 num_frames = 7;    // Number of frames to generate
-  int32 fps = 8;          // Frames per second
-  int32 seed = 9;
-  float cfg_scale = 10;    // Classifier-free guidance scale
-  int32 step = 11;         // Number of inference steps
-  string dst = 12;        // Output path for the generated video
+  string start_image = 2;  // Path or base64 encoded image for the start frame
+  string end_image = 3;    // Path or base64 encoded image for the end frame
+  int32 width = 4;
+  int32 height = 5;
+  int32 num_frames = 6;    // Number of frames to generate
+  int32 fps = 7;          // Frames per second
+  int32 seed = 8;
+  float cfg_scale = 9;    // Classifier-free guidance scale
+  string dst = 10;        // Output path for the generated video
 }
 
 message TTSRequest {
@@ -387,11 +379,6 @@ message StatusResponse {
 message Message {
   string role = 1;
   string content = 2;
-  // Optional fields for OpenAI-compatible message format
-  string name = 3;                    // Tool name (for tool messages)
-  string tool_call_id = 4;            // Tool call ID (for tool messages)
-  string reasoning_content = 5;       // Reasoning content (for thinking models)
-  string tool_calls = 6;              // Tool calls as JSON string (for assistant messages with tool calls)
 }
 
 message DetectOptions {

backend/cpp/llama-cpp/Makefile

@@ -1,5 +1,5 @@
 
-LLAMA_VERSION?=10e9780154365b191fb43ca4830659ef12def80f
+LLAMA_VERSION?=710dfc465a68f7443b87d9f792cffba00ed739fe
 LLAMA_REPO?=https://github.com/ggerganov/llama.cpp
 
 CMAKE_ARGS?=
@@ -14,7 +14,7 @@ CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF -DLLAMA_CURL=OFF
 
 CURRENT_MAKEFILE_DIR := $(dir $(abspath $(lastword $(MAKEFILE_LIST))))
 ifeq ($(NATIVE),false)
-	CMAKE_ARGS+=-DGGML_NATIVE=OFF -DLLAMA_OPENSSL=OFF
+	CMAKE_ARGS+=-DGGML_NATIVE=OFF
 endif
 # If build type is cublas, then we set -DGGML_CUDA=ON to CMAKE_ARGS automatically
 ifeq ($(BUILD_TYPE),cublas)

backend/cpp/llama-cpp/prepare.sh

@@ -14,8 +14,6 @@ cp -r grpc-server.cpp llama.cpp/tools/grpc-server/
 cp -rfv llama.cpp/vendor/nlohmann/json.hpp llama.cpp/tools/grpc-server/
 cp -rfv llama.cpp/tools/server/utils.hpp llama.cpp/tools/grpc-server/
 cp -rfv llama.cpp/vendor/cpp-httplib/httplib.h llama.cpp/tools/grpc-server/
-cp -rfv llama.cpp/tools/server/server-http.cpp llama.cpp/tools/grpc-server/
-cp -rfv llama.cpp/tools/server/server-http.h llama.cpp/tools/grpc-server/
 
 set +e
 if grep -q "grpc-server" llama.cpp/tools/CMakeLists.txt; then

backend/go/stablediffusion-ggml/.gitignore

@@ -1,6 +0,0 @@
-package/
-sources/
-.cache/
-build/
-libgosd.so
-stablediffusion-ggml

backend/go/stablediffusion-ggml/CMakeLists.txt

@@ -5,11 +5,7 @@ set(CMAKE_POSITION_INDEPENDENT_CODE ON)
 add_subdirectory(./sources/stablediffusion-ggml.cpp)
 
 add_library(gosd MODULE gosd.cpp)
-target_link_libraries(gosd PRIVATE stable-diffusion ggml)
-
-if(CMAKE_CXX_COMPILER_ID MATCHES "GNU" AND CMAKE_CXX_COMPILER_VERSION VERSION_LESS 9.0)
-    target_link_libraries(gosd PRIVATE stdc++fs)
-endif()
+target_link_libraries(gosd PRIVATE stable-diffusion ggml stdc++fs)
 
 target_include_directories(gosd PUBLIC
     stable-diffusion.cpp

backend/go/stablediffusion-ggml/Makefile

@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)
 
 # stablediffusion.cpp (ggml)
 STABLEDIFFUSION_GGML_REPO?=https://github.com/leejet/stable-diffusion.cpp
-STABLEDIFFUSION_GGML_VERSION?=0ebe6fe118f125665939b27c89f34ed38716bff8
+STABLEDIFFUSION_GGML_VERSION?=5900ef6605c6fbf7934239f795c13c97bc993853
 
 CMAKE_ARGS+=-DGGML_MAX_NAME=128
 
@@ -29,6 +29,8 @@ 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+=-DSD_HIPBLAS=ON -DGGML_HIPBLAS=ON
+# If it's OSX, DO NOT embed the metal library - -DGGML_METAL_EMBED_LIBRARY=ON requires further investigation
+# But if it's OSX without metal, disable it here
 else ifeq ($(BUILD_TYPE),vulkan)
 	CMAKE_ARGS+=-DSD_VULKAN=ON -DGGML_VULKAN=ON
 else ifeq ($(OS),Darwin)
@@ -72,10 +74,10 @@ libgosd.so: sources/stablediffusion-ggml.cpp CMakeLists.txt gosd.cpp gosd.h
 stablediffusion-ggml: main.go gosd.go libgosd.so
 	CGO_ENABLED=0 $(GOCMD) build -tags "$(GO_TAGS)" -o stablediffusion-ggml ./
 
-package: stablediffusion-ggml
+package:
 	bash package.sh
 
-build: package
+build: stablediffusion-ggml package
 
 clean:
-	rm -rf libgosd.so build stablediffusion-ggml package sources
+	rm -rf libgosd.o build stablediffusion-ggml

backend/go/stablediffusion-ggml/gosd.cpp

@@ -4,11 +4,17 @@
 #include <stdio.h>
 #include <string.h>
 #include <time.h>
+#include <iostream>
+#include <random>
 #include <string>
 #include <vector>
 #include <filesystem>
 #include "gosd.h"
 
+// #include "preprocessing.hpp"
+#include "flux.hpp"
+#include "stable-diffusion.h"
+
 #define STB_IMAGE_IMPLEMENTATION
 #define STB_IMAGE_STATIC
 #include "stb_image.h"
@@ -23,7 +29,7 @@
 
 // Names of the sampler method, same order as enum sample_method in stable-diffusion.h
 const char* sample_method_str[] = {
-    "default",
+    "euler_a",
     "euler",
     "heun",
     "dpm2",
@@ -35,27 +41,19 @@ const char* sample_method_str[] = {
     "lcm",
     "ddim_trailing",
     "tcd",
-    "euler_a",
 };
 
-static_assert(std::size(sample_method_str) == SAMPLE_METHOD_COUNT, "sample method mismatch");
-
 // Names of the sigma schedule overrides, same order as sample_schedule in stable-diffusion.h
-const char* schedulers[] = {
+const char* schedule_str[] = {
     "default",
     "discrete",
     "karras",
     "exponential",
     "ays",
     "gits",
-    "smoothstep",
 };
 
-static_assert(std::size(schedulers) == SCHEDULE_COUNT, "schedulers mismatch");
-
 sd_ctx_t* sd_c;
-// Moved from the context (load time) to generation time params
-scheduler_t scheduler = scheduler_t::DEFAULT;
 
 sample_method_t sample_method;
 
@@ -107,7 +105,7 @@ int load_model(const char *model, char *model_path, char* options[], int threads
     const char *clip_g_path  = "";
     const char *t5xxl_path  = "";
     const char *vae_path  = "";
-    const char *scheduler_str = "";
+    const char *scheduler = "";
     const char *sampler = "";
     char *lora_dir = model_path;
     bool lora_dir_allocated = false;
@@ -135,7 +133,7 @@ int load_model(const char *model, char *model_path, char* options[], int threads
             vae_path = optval;
         }
         if (!strcmp(optname, "scheduler")) {
-            scheduler_str = optval;
+            scheduler = optval;
         }
         if (!strcmp(optname, "sampler")) {
             sampler = optval;
@@ -168,17 +166,26 @@ int load_model(const char *model, char *model_path, char* options[], int threads
     }
     if (sample_method_found == -1) {
         fprintf(stderr, "Invalid sample method, default to EULER_A!\n");
-        sample_method_found = sample_method_t::SAMPLE_METHOD_DEFAULT;
+        sample_method_found = EULER_A;
     }
     sample_method = (sample_method_t)sample_method_found;
 
+    int schedule_found            = -1;
     for (int d = 0; d < SCHEDULE_COUNT; d++) {
-        if (!strcmp(scheduler_str, schedulers[d])) {
-            scheduler = (scheduler_t)d;
-            fprintf (stderr, "Found scheduler: %s\n", scheduler_str);
+        if (!strcmp(scheduler, schedule_str[d])) {
+            schedule_found = d;
+                fprintf (stderr, "Found scheduler: %s\n", scheduler);
+
         }
     }
 
+    if (schedule_found == -1) {
+        fprintf (stderr, "Invalid scheduler! using DEFAULT\n");
+        schedule_found = DEFAULT;
+    }
+
+    schedule_t schedule = (schedule_t)schedule_found;
+
     fprintf (stderr, "Creating context\n");
     sd_ctx_params_t ctx_params;
     sd_ctx_params_init(&ctx_params);
@@ -192,10 +199,13 @@ int load_model(const char *model, char *model_path, char* options[], int threads
     ctx_params.control_net_path = "";
     ctx_params.lora_model_dir = lora_dir;
     ctx_params.embedding_dir = "";
+    ctx_params.stacked_id_embed_dir = "";
     ctx_params.vae_decode_only = false;
+    ctx_params.vae_tiling = false;
     ctx_params.free_params_immediately = false;
     ctx_params.n_threads = threads;
     ctx_params.rng_type = STD_DEFAULT_RNG;
+    ctx_params.schedule = schedule;
     sd_ctx_t* sd_ctx = new_sd_ctx(&ctx_params);
 
     if (sd_ctx == NULL) {
@@ -218,49 +228,7 @@ int load_model(const char *model, char *model_path, char* options[], int threads
     return 0;
 }
 
-void sd_tiling_params_set_enabled(sd_tiling_params_t *params, bool enabled) {
-    params->enabled = enabled;
-}
-
-void sd_tiling_params_set_tile_sizes(sd_tiling_params_t *params, int tile_size_x, int tile_size_y) {
-    params->tile_size_x = tile_size_x;
-    params->tile_size_y = tile_size_y;
-}
-
-void sd_tiling_params_set_rel_sizes(sd_tiling_params_t *params, float rel_size_x, float rel_size_y) {
-    params->rel_size_x = rel_size_x;
-    params->rel_size_y = rel_size_y;
-}
-
-void sd_tiling_params_set_target_overlap(sd_tiling_params_t *params, float target_overlap) {
-    params->target_overlap = target_overlap;
-}
-
-sd_tiling_params_t* sd_img_gen_params_get_vae_tiling_params(sd_img_gen_params_t *params) {
-    return &params->vae_tiling_params;
-}
-
-sd_img_gen_params_t* sd_img_gen_params_new(void) {
-    sd_img_gen_params_t *params = (sd_img_gen_params_t *)std::malloc(sizeof(sd_img_gen_params_t));
-    sd_img_gen_params_init(params);
-    return params;
-}
-
-void sd_img_gen_params_set_prompts(sd_img_gen_params_t *params, const char *prompt, const char *negative_prompt) {
-    params->prompt = prompt;
-    params->negative_prompt = negative_prompt;
-}
-
-void sd_img_gen_params_set_dimensions(sd_img_gen_params_t *params, int width, int height) {
-    params->width = width;
-    params->height = height;
-}
-
-void sd_img_gen_params_set_seed(sd_img_gen_params_t *params, int64_t seed) {
-    params->seed = seed;
-}
-
-int gen_image(sd_img_gen_params_t *p, int steps, char *dst, float cfg_scale, char *src_image, float strength, char *mask_image, char **ref_images, int ref_images_count) {
+int gen_image(char *text, char *negativeText, int width, int height, int steps, int64_t seed, char *dst, float cfg_scale, char *src_image, float strength, char *mask_image, char **ref_images, int ref_images_count) {
 
     sd_image_t* results;
 
@@ -268,15 +236,20 @@ int gen_image(sd_img_gen_params_t *p, int steps, char *dst, float cfg_scale, cha
 
     fprintf (stderr, "Generating image\n");
 
-    p->sample_params.guidance.txt_cfg = cfg_scale;
-    p->sample_params.guidance.slg.layers = skip_layers.data();
-    p->sample_params.guidance.slg.layer_count = skip_layers.size();
-    p->sample_params.sample_method = sample_method;
-    p->sample_params.sample_steps = steps;
-    p->sample_params.scheduler = scheduler;
+    sd_img_gen_params_t p;
+    sd_img_gen_params_init(&p);
 
-    int width = p->width;
-    int height = p->height;
+    p.prompt = text;
+    p.negative_prompt = negativeText;
+    p.guidance.txt_cfg = cfg_scale;
+    p.guidance.slg.layers = skip_layers.data();
+    p.guidance.slg.layer_count = skip_layers.size();
+    p.width = width;
+    p.height = height;
+    p.sample_method = sample_method;
+    p.sample_steps = steps;
+    p.seed = seed;
+    p.input_id_images_path = "";
 
     // Handle input image for img2img
     bool has_input_image = (src_image != NULL && strlen(src_image) > 0);
@@ -325,13 +298,13 @@ int gen_image(sd_img_gen_params_t *p, int steps, char *dst, float cfg_scale, cha
             input_image_buffer = resized_image_buffer;
         }
 
-        p->init_image = {(uint32_t)width, (uint32_t)height, 3, input_image_buffer};
-        p->strength = strength;
+        p.init_image = {(uint32_t)width, (uint32_t)height, 3, input_image_buffer};
+        p.strength = strength;
         fprintf(stderr, "Using img2img with strength: %.2f\n", strength);
     } else {
         // No input image, use empty image for text-to-image
-        p->init_image = {(uint32_t)width, (uint32_t)height, 3, NULL};
-        p->strength = 0.0f;
+        p.init_image = {(uint32_t)width, (uint32_t)height, 3, NULL};
+        p.strength = 0.0f;
     }
 
     // Handle mask image for inpainting
@@ -371,12 +344,12 @@ int gen_image(sd_img_gen_params_t *p, int steps, char *dst, float cfg_scale, cha
             mask_image_buffer = resized_mask_buffer;
         }
 
-        p->mask_image = {(uint32_t)width, (uint32_t)height, 1, mask_image_buffer};
+        p.mask_image = {(uint32_t)width, (uint32_t)height, 1, mask_image_buffer};
         fprintf(stderr, "Using inpainting with mask\n");
     } else {
         // No mask image, create default full mask
         default_mask_image_vec.resize(width * height, 255);
-        p->mask_image = {(uint32_t)width, (uint32_t)height, 1, default_mask_image_vec.data()};
+        p.mask_image = {(uint32_t)width, (uint32_t)height, 1, default_mask_image_vec.data()};
     }
 
     // Handle reference images
@@ -434,15 +407,13 @@ int gen_image(sd_img_gen_params_t *p, int steps, char *dst, float cfg_scale, cha
         }
 
         if (!ref_images_vec.empty()) {
-            p->ref_images = ref_images_vec.data();
-            p->ref_images_count = ref_images_vec.size();
+            p.ref_images = ref_images_vec.data();
+            p.ref_images_count = ref_images_vec.size();
             fprintf(stderr, "Using %zu reference images\n", ref_images_vec.size());
         }
     }
 
-    results = generate_image(sd_c, p);
-
-    std::free(p);
+    results = generate_image(sd_c, &p);
 
     if (results == NULL) {
         fprintf (stderr, "NO results\n");

backend/go/stablediffusion-ggml/gosd.go

@@ -22,18 +22,7 @@ type SDGGML struct {
 
 var (
 	LoadModel func(model, model_apth string, options []uintptr, threads int32, diff int) int
-	GenImage  func(params uintptr, steps int, dst string, cfgScale float32, srcImage string, strength float32, maskImage string, refImages []string, refImagesCount int) int
-
-	TilingParamsSetEnabled       func(params uintptr, enabled bool)
-	TilingParamsSetTileSizes     func(params uintptr, tileSizeX int, tileSizeY int)
-	TilingParamsSetRelSizes      func(params uintptr, relSizeX float32, relSizeY float32)
-	TilingParamsSetTargetOverlap func(params uintptr, targetOverlap float32)
-
-	ImgGenParamsNew                func() uintptr
-	ImgGenParamsSetPrompts         func(params uintptr, prompt string, negativePrompt string)
-	ImgGenParamsSetDimensions      func(params uintptr, width int, height int)
-	ImgGenParamsSetSeed            func(params uintptr, seed int64)
-	ImgGenParamsGetVaeTilingParams func(params uintptr) uintptr
+	GenImage  func(text, negativeText string, width, height, steps int, seed int64, dst string, cfgScale float32, srcImage string, strength float32, maskImage string, refImages []string, refImagesCount int) int
 )
 
 // Copied from Purego internal/strings
@@ -131,15 +120,7 @@ func (sd *SDGGML) GenerateImage(opts *pb.GenerateImageRequest) error {
 	// Default strength for img2img (0.75 is a good default)
 	strength := float32(0.75)
 
-	// free'd by GenImage
-	p := ImgGenParamsNew()
-	ImgGenParamsSetPrompts(p, t, negative)
-	ImgGenParamsSetDimensions(p, int(opts.Width), int(opts.Height))
-	ImgGenParamsSetSeed(p, int64(opts.Seed))
-	vaep := ImgGenParamsGetVaeTilingParams(p)
-	TilingParamsSetEnabled(vaep, false)
-
-	ret := GenImage(p, int(opts.Step), dst, sd.cfgScale, srcImage, strength, maskImage, refImages, refImagesCount)
+	ret := GenImage(t, negative, int(opts.Width), int(opts.Height), int(opts.Step), int64(opts.Seed), dst, sd.cfgScale, srcImage, strength, maskImage, refImages, refImagesCount)
 	if ret != 0 {
 		return fmt.Errorf("inference failed")
 	}

backend/go/stablediffusion-ggml/gosd.h

@@ -1,23 +1,8 @@
-#include <cstdint>
-#include "stable-diffusion.h"
-
 #ifdef __cplusplus
 extern "C" {
 #endif
-
-void sd_tiling_params_set_enabled(sd_tiling_params_t *params, bool enabled);
-void sd_tiling_params_set_tile_sizes(sd_tiling_params_t *params, int tile_size_x, int tile_size_y);
-void sd_tiling_params_set_rel_sizes(sd_tiling_params_t *params, float rel_size_x, float rel_size_y);
-void sd_tiling_params_set_target_overlap(sd_tiling_params_t *params, float target_overlap);
-sd_tiling_params_t* sd_img_gen_params_get_vae_tiling_params(sd_img_gen_params_t *params);
-
-sd_img_gen_params_t* sd_img_gen_params_new(void);
-void sd_img_gen_params_set_prompts(sd_img_gen_params_t *params, const char *prompt, const char *negative_prompt);
-void sd_img_gen_params_set_dimensions(sd_img_gen_params_t *params, int width, int height);
-void sd_img_gen_params_set_seed(sd_img_gen_params_t *params, int64_t seed);
-
 int load_model(const char *model, char *model_path, char* options[], int threads, int diffusionModel);
-int gen_image(sd_img_gen_params_t *p, int steps, char *dst, float cfg_scale, char *src_image, float strength, char *mask_image, char **ref_images, int ref_images_count);
+int gen_image(char *text, char *negativeText, int width, int height, int steps, int64_t seed, char *dst, float cfg_scale, char *src_image, float strength, char *mask_image, char **ref_images, int ref_images_count);
 #ifdef __cplusplus
 }
 #endif

backend/go/stablediffusion-ggml/main.go

@@ -11,35 +11,14 @@ var (
 	addr = flag.String("addr", "localhost:50051", "the address to connect to")
 )
 
-type LibFuncs struct {
-	FuncPtr any
-	Name    string
-}
-
 func main() {
 	gosd, err := purego.Dlopen("./libgosd.so", purego.RTLD_NOW|purego.RTLD_GLOBAL)
 	if err != nil {
 		panic(err)
 	}
 
-	libFuncs := []LibFuncs{
-		{&LoadModel, "load_model"},
-		{&GenImage, "gen_image"},
-		{&TilingParamsSetEnabled, "sd_tiling_params_set_enabled"},
-		{&TilingParamsSetTileSizes, "sd_tiling_params_set_tile_sizes"},
-		{&TilingParamsSetRelSizes, "sd_tiling_params_set_rel_sizes"},
-		{&TilingParamsSetTargetOverlap, "sd_tiling_params_set_target_overlap"},
-
-		{&ImgGenParamsNew, "sd_img_gen_params_new"},
-		{&ImgGenParamsSetPrompts, "sd_img_gen_params_set_prompts"},
-		{&ImgGenParamsSetDimensions, "sd_img_gen_params_set_dimensions"},
-		{&ImgGenParamsSetSeed, "sd_img_gen_params_set_seed"},
-		{&ImgGenParamsGetVaeTilingParams, "sd_img_gen_params_get_vae_tiling_params"},
-	}
-
-	for _, lf := range libFuncs {
-		purego.RegisterLibFunc(lf.FuncPtr, gosd, lf.Name)
-	}
+	purego.RegisterLibFunc(&LoadModel, gosd, "load_model")
+	purego.RegisterLibFunc(&GenImage, gosd, "gen_image")
 
 	flag.Parse()
 

backend/go/stablediffusion-ggml/package.sh

@@ -10,9 +10,9 @@ CURDIR=$(dirname "$(realpath $0)")
 # Create lib directory
 mkdir -p $CURDIR/package/lib
 
-cp -avf $CURDIR/libgosd.so $CURDIR/package/
-cp -avf $CURDIR/stablediffusion-ggml $CURDIR/package/
-cp -fv $CURDIR/run.sh $CURDIR/package/
+cp -avrf $CURDIR/libgosd.so $CURDIR/package/
+cp -avrf $CURDIR/stablediffusion-ggml $CURDIR/package/
+cp -rfv $CURDIR/run.sh $CURDIR/package/
 
 # Detect architecture and copy appropriate libraries
 if [ -f "/lib64/ld-linux-x86-64.so.2" ]; then
@@ -43,8 +43,6 @@ elif [ -f "/lib/ld-linux-aarch64.so.1" ]; then
     cp -arfLv /lib/aarch64-linux-gnu/libdl.so.2 $CURDIR/package/lib/libdl.so.2
     cp -arfLv /lib/aarch64-linux-gnu/librt.so.1 $CURDIR/package/lib/librt.so.1
     cp -arfLv /lib/aarch64-linux-gnu/libpthread.so.0 $CURDIR/package/lib/libpthread.so.0
-elif [ $(uname -s) = "Darwin" ]; then
-    echo "Detected Darwin"
 else
     echo "Error: Could not detect architecture"
     exit 1

backend/go/whisper/.gitignore

@@ -1,7 +0,0 @@
-.cache/
-sources/
-build/
-package/
-whisper
-libgowhisper.so
-

backend/go/whisper/CMakeLists.txt

@@ -1,16 +0,0 @@
-cmake_minimum_required(VERSION 3.12)
-project(gowhisper LANGUAGES C CXX)
-set(CMAKE_POSITION_INDEPENDENT_CODE ON)
-set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
-
-add_subdirectory(./sources/whisper.cpp)
-
-add_library(gowhisper MODULE gowhisper.cpp)
-target_link_libraries(gowhisper PRIVATE whisper ggml)
-
-if(CMAKE_CXX_COMPILER_ID MATCHES "GNU" AND CMAKE_CXX_COMPILER_VERSION VERSION_LESS 9.0)
-    target_link_libraries(gosd PRIVATE stdc++fs)
-endif()
-
-set_property(TARGET gowhisper PROPERTY CXX_STANDARD 17)
-set_target_properties(gowhisper PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR})

backend/go/whisper/Makefile

@@ -1,54 +1,110 @@
-CMAKE_ARGS?=
-BUILD_TYPE?=
+GOCMD=go
 NATIVE?=false
 
-GOCMD?=go
-GO_TAGS?=
-JOBS?=$(shell nproc --ignore=1)
+BUILD_TYPE?=
+CMAKE_ARGS?=
 
 # whisper.cpp version
 WHISPER_REPO?=https://github.com/ggml-org/whisper.cpp
-WHISPER_CPP_VERSION?=b12abefa9be2abae39a73fa903322af135024a36
-SO_TARGET?=libgowhisper.so
+WHISPER_CPP_VERSION?=fc45bb86251f774ef817e89878bb4c2636c8a58f
 
-CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF
+export WHISPER_CMAKE_ARGS?=-DBUILD_SHARED_LIBS=OFF
+export WHISPER_DIR=$(abspath ./sources/whisper.cpp)
+export WHISPER_INCLUDE_PATH=$(WHISPER_DIR)/include:$(WHISPER_DIR)/ggml/include
+export WHISPER_LIBRARY_PATH=$(WHISPER_DIR)/build/src/:$(WHISPER_DIR)/build/ggml/src
 
+CGO_LDFLAGS_WHISPER?=
+CGO_LDFLAGS_WHISPER+=-lggml
+CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF -DLLAMA_CURL=OFF
+CUDA_LIBPATH?=/usr/local/cuda/lib64/
+
+ONEAPI_VERSION?=2025.2
+
+# IF native is false, we add -DGGML_NATIVE=OFF to CMAKE_ARGS
+ifeq ($(NATIVE),false)
+	CMAKE_ARGS+=-DGGML_NATIVE=OFF
+	WHISPER_CMAKE_ARGS+=-DGGML_NATIVE=OFF
+endif
+CURRENT_MAKEFILE_DIR := $(dir $(abspath $(lastword $(MAKEFILE_LIST))))
 ifeq ($(NATIVE),false)
 	CMAKE_ARGS+=-DGGML_NATIVE=OFF
 endif
-
+# If build type is cublas, then we set -DGGML_CUDA=ON to CMAKE_ARGS automatically
 ifeq ($(BUILD_TYPE),cublas)
+	CGO_LDFLAGS+=-lcublas -lcudart -L$(CUDA_LIBPATH) -L$(CUDA_LIBPATH)/stubs/ -lcuda
 	CMAKE_ARGS+=-DGGML_CUDA=ON
+	CGO_LDFLAGS_WHISPER+=-lcufft -lggml-cuda
+	export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-cuda/
+# If build type is openblas then we set -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
+# to CMAKE_ARGS automatically
 else ifeq ($(BUILD_TYPE),openblas)
 	CMAKE_ARGS+=-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS
+# If build type is clblas (openCL) we set -DGGML_CLBLAST=ON -DCLBlast_DIR=/some/path
 else ifeq ($(BUILD_TYPE),clblas)
 	CMAKE_ARGS+=-DGGML_CLBLAST=ON -DCLBlast_DIR=/some/path
+# If it's hipblas we do have also to set CC=/opt/rocm/llvm/bin/clang CXX=/opt/rocm/llvm/bin/clang++
 else ifeq ($(BUILD_TYPE),hipblas)
-	CMAKE_ARGS+=-DGGML_HIPBLAS=ON
+	ROCM_HOME ?= /opt/rocm
+	ROCM_PATH ?= /opt/rocm
+	LD_LIBRARY_PATH ?= /opt/rocm/lib:/opt/rocm/llvm/lib
+	export STABLE_BUILD_TYPE=
+	export CXX=$(ROCM_HOME)/llvm/bin/clang++
+	export CC=$(ROCM_HOME)/llvm/bin/clang
+#	GPU_TARGETS ?= gfx803,gfx900,gfx906,gfx908,gfx90a,gfx942,gfx1010,gfx1030,gfx1032,gfx1100,gfx1101,gfx1102
+#	AMDGPU_TARGETS ?= "$(GPU_TARGETS)"
+	CMAKE_ARGS+=-DGGML_HIP=ON
+	CGO_LDFLAGS += -O3 --rtlib=compiler-rt -unwindlib=libgcc -lhipblas -lrocblas --hip-link -L${ROCM_HOME}/lib/llvm/lib -L$(CURRENT_MAKEFILE_DIR)/sources/whisper.cpp/build/ggml/src/ggml-hip/ -lggml-hip
+#	CMAKE_ARGS+=-DGGML_HIP=ON -DAMDGPU_TARGETS="$(AMDGPU_TARGETS)" -DGPU_TARGETS="$(GPU_TARGETS)"
 else ifeq ($(BUILD_TYPE),vulkan)
-	CMAKE_ARGS+=-DGGML_VULKAN=ON
+	CMAKE_ARGS+=-DGGML_VULKAN=1
+	CGO_LDFLAGS_WHISPER+=-lggml-vulkan -lvulkan
+	export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-vulkan/
 else ifeq ($(OS),Darwin)
+	ifeq ($(BUILD_TYPE),)
+		BUILD_TYPE=metal
+	endif
 	ifneq ($(BUILD_TYPE),metal)
 		CMAKE_ARGS+=-DGGML_METAL=OFF
+		CGO_LDFLAGS_WHISPER+=-lggml-blas
+		export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-blas
 	else
 		CMAKE_ARGS+=-DGGML_METAL=ON
 		CMAKE_ARGS+=-DGGML_METAL_EMBED_LIBRARY=ON
+		CMAKE_ARGS+=-DGGML_METAL_USE_BF16=ON
+		CMAKE_ARGS+=-DGGML_OPENMP=OFF
+		CMAKE_ARGS+=-DWHISPER_BUILD_EXAMPLES=OFF
+		CMAKE_ARGS+=-DWHISPER_BUILD_TESTS=OFF
+		CMAKE_ARGS+=-DWHISPER_BUILD_SERVER=OFF
+		CGO_LDFLAGS += -framework Accelerate
+		CGO_LDFLAGS_WHISPER+=-lggml-metal -lggml-blas
+		export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-metal/:$(WHISPER_DIR)/build/ggml/src/ggml-blas
 	endif
+	TARGET+=--target ggml-metal
 endif
 
-ifeq ($(BUILD_TYPE),sycl_f16)
+ifneq (,$(findstring sycl,$(BUILD_TYPE)))
+	export CC=icx
+	export CXX=icpx
+	CGO_LDFLAGS_WHISPER += -fsycl -L${DNNLROOT}/lib -rpath ${ONEAPI_ROOT}/${ONEAPI_VERSION}/lib -ldnnl ${MKLROOT}/lib/intel64/libmkl_sycl.a -fiopenmp -fopenmp-targets=spir64 -lOpenCL -lggml-sycl
+	CGO_LDFLAGS_WHISPER += $(shell pkg-config --libs mkl-static-lp64-gomp)
+	CGO_CXXFLAGS_WHISPER += -fiopenmp -fopenmp-targets=spir64
+	CGO_CXXFLAGS_WHISPER += $(shell pkg-config --cflags mkl-static-lp64-gomp )
+	export WHISPER_LIBRARY_PATH:=$(WHISPER_LIBRARY_PATH):$(WHISPER_DIR)/build/ggml/src/ggml-sycl/
 	CMAKE_ARGS+=-DGGML_SYCL=ON \
 		-DCMAKE_C_COMPILER=icx \
 		-DCMAKE_CXX_COMPILER=icpx \
-		-DGGML_SYCL_F16=ON
+		-DCMAKE_CXX_FLAGS="-fsycl"
 endif
 
-ifeq ($(BUILD_TYPE),sycl_f32)
-	CMAKE_ARGS+=-DGGML_SYCL=ON \
-		-DCMAKE_C_COMPILER=icx \
-		-DCMAKE_CXX_COMPILER=icpx
+ifeq ($(BUILD_TYPE),sycl_f16)
+	CMAKE_ARGS+=-DGGML_SYCL_F16=ON
 endif
 
+ifneq ($(OS),Darwin)
+	CGO_LDFLAGS_WHISPER+=-lgomp
+endif
+
+## whisper
 sources/whisper.cpp:
 	mkdir -p sources/whisper.cpp
 	cd sources/whisper.cpp && \
@@ -58,65 +114,18 @@ sources/whisper.cpp:
 	git checkout $(WHISPER_CPP_VERSION) && \
 	git submodule update --init --recursive --depth 1 --single-branch
 
-# Detect OS
-UNAME_S := $(shell uname -s)
+sources/whisper.cpp/build/src/libwhisper.a: sources/whisper.cpp
+	cd sources/whisper.cpp && cmake $(CMAKE_ARGS) $(WHISPER_CMAKE_ARGS) . -B ./build
+	cd sources/whisper.cpp/build && cmake --build . --config Release
 
-# Only build CPU variants on Linux
-ifeq ($(UNAME_S),Linux)
-	VARIANT_TARGETS = libgowhisper-avx.so libgowhisper-avx2.so libgowhisper-avx512.so libgowhisper-fallback.so
-else
-	# On non-Linux (e.g., Darwin), build only fallback variant
-	VARIANT_TARGETS = libgowhisper-fallback.so
-endif
+whisper: sources/whisper.cpp sources/whisper.cpp/build/src/libwhisper.a
+	$(GOCMD) mod edit -replace github.com/ggerganov/whisper.cpp=$(CURDIR)/sources/whisper.cpp
+	$(GOCMD) mod edit -replace github.com/ggerganov/whisper.cpp/bindings/go=$(CURDIR)/sources/whisper.cpp/bindings/go
+	CGO_LDFLAGS="$(CGO_LDFLAGS) $(CGO_LDFLAGS_WHISPER)" C_INCLUDE_PATH="${WHISPER_INCLUDE_PATH}" LIBRARY_PATH="${WHISPER_LIBRARY_PATH}" LD_LIBRARY_PATH="${WHISPER_LIBRARY_PATH}" \
+	CGO_CXXFLAGS="$(CGO_CXXFLAGS_WHISPER)" \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o whisper ./
 
-whisper: main.go gowhisper.go $(VARIANT_TARGETS)
-	CGO_ENABLED=0 $(GOCMD) build -tags "$(GO_TAGS)" -o whisper ./
-
-package: whisper
+package:
 	bash package.sh
 
-build: package
-
-clean: purge
-	rm -rf libgowhisper*.so sources/whisper.cpp whisper
-
-purge:
-	rm -rf build*
-
-# Build all variants (Linux only)
-ifeq ($(UNAME_S),Linux)
-libgowhisper-avx.so: sources/whisper.cpp
-	$(MAKE) purge
-	$(info ${GREEN}I whisper build info:avx${RESET})
-	SO_TARGET=libgowhisper-avx.so CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=on -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off" $(MAKE) libgowhisper-custom
-	rm -rfv build*
-
-libgowhisper-avx2.so: sources/whisper.cpp
-	$(MAKE) purge
-	$(info ${GREEN}I whisper build info:avx2${RESET})
-	SO_TARGET=libgowhisper-avx2.so CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=on -DGGML_AVX2=on -DGGML_AVX512=off -DGGML_FMA=on -DGGML_F16C=on" $(MAKE) libgowhisper-custom
-	rm -rfv build*
-
-libgowhisper-avx512.so: sources/whisper.cpp
-	$(MAKE) purge
-	$(info ${GREEN}I whisper build info:avx512${RESET})
-	SO_TARGET=libgowhisper-avx512.so CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=on -DGGML_AVX2=off -DGGML_AVX512=on -DGGML_FMA=on -DGGML_F16C=on" $(MAKE) libgowhisper-custom
-	rm -rfv build*
-endif
-
-# Build fallback variant (all platforms)
-libgowhisper-fallback.so: sources/whisper.cpp
-	$(MAKE) purge
-	$(info ${GREEN}I whisper build info:fallback${RESET})
-	SO_TARGET=libgowhisper-fallback.so CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off" $(MAKE) libgowhisper-custom
-	rm -rfv build*
-
-libgowhisper-custom: CMakeLists.txt gowhisper.cpp gowhisper.h
-	mkdir -p build-$(SO_TARGET) && \
-	cd build-$(SO_TARGET) && \
-	cmake .. $(CMAKE_ARGS) && \
-	cmake --build . --config Release -j$(JOBS) && \
-	cd .. && \
-	mv build-$(SO_TARGET)/libgowhisper.so ./$(SO_TARGET)
-
-all: whisper package
+build: whisper package

backend/go/whisper/gowhisper.cpp

@@ -1,154 +0,0 @@
-#include "gowhisper.h"
-#include "ggml-backend.h"
-#include "whisper.h"
-#include <vector>
-
-static struct whisper_vad_context *vctx;
-static struct whisper_context *ctx;
-static std::vector<float> flat_segs;
-
-static void ggml_log_cb(enum ggml_log_level level, const char *log,
-                        void *data) {
-  const char *level_str;
-
-  if (!log) {
-    return;
-  }
-
-  switch (level) {
-  case GGML_LOG_LEVEL_DEBUG:
-    level_str = "DEBUG";
-    break;
-  case GGML_LOG_LEVEL_INFO:
-    level_str = "INFO";
-    break;
-  case GGML_LOG_LEVEL_WARN:
-    level_str = "WARN";
-    break;
-  case GGML_LOG_LEVEL_ERROR:
-    level_str = "ERROR";
-    break;
-  default: /* Potential future-proofing */
-    level_str = "?????";
-    break;
-  }
-
-  fprintf(stderr, "[%-5s] ", level_str);
-  fputs(log, stderr);
-  fflush(stderr);
-}
-
-int load_model(const char *const model_path) {
-  whisper_log_set(ggml_log_cb, nullptr);
-  ggml_backend_load_all();
-
-  struct whisper_context_params cparams = whisper_context_default_params();
-
-  ctx = whisper_init_from_file_with_params(model_path, cparams);
-  if (ctx == nullptr) {
-    fprintf(stderr, "error: Also failed to init model as transcriber\n");
-    return 1;
-  }
-
-  return 0;
-}
-
-int load_model_vad(const char *const model_path) {
-  whisper_log_set(ggml_log_cb, nullptr);
-  ggml_backend_load_all();
-
-  struct whisper_vad_context_params vcparams =
-      whisper_vad_default_context_params();
-
-  // XXX: Overridden to false in upstream due to performance?
-  // vcparams.use_gpu = true;
-
-  vctx = whisper_vad_init_from_file_with_params(model_path, vcparams);
-  if (vctx == nullptr) {
-    fprintf(stderr, "error: Failed to init model as VAD\n");
-    return 1;
-  }
-
-  return 0;
-}
-
-int vad(float pcmf32[], size_t pcmf32_len, float **segs_out,
-        size_t *segs_out_len) {
-  if (!whisper_vad_detect_speech(vctx, pcmf32, pcmf32_len)) {
-    fprintf(stderr, "error: failed to detect speech\n");
-    return 1;
-  }
-
-  struct whisper_vad_params params = whisper_vad_default_params();
-  struct whisper_vad_segments *segs =
-      whisper_vad_segments_from_probs(vctx, params);
-  size_t segn = whisper_vad_segments_n_segments(segs);
-
-  // fprintf(stderr, "Got segments %zd\n", segn);
-
-  flat_segs.clear();
-
-  for (int i = 0; i < segn; i++) {
-    flat_segs.push_back(whisper_vad_segments_get_segment_t0(segs, i));
-    flat_segs.push_back(whisper_vad_segments_get_segment_t1(segs, i));
-  }
-
-  // fprintf(stderr, "setting out variables: %p=%p -> %p, %p=%zx -> %zx\n",
-  //         segs_out, *segs_out, flat_segs.data(), segs_out_len, *segs_out_len,
-  //         flat_segs.size());
-  *segs_out = flat_segs.data();
-  *segs_out_len = flat_segs.size();
-
-  // fprintf(stderr, "freeing segs\n");
-  whisper_vad_free_segments(segs);
-
-  // fprintf(stderr, "returning\n");
-  return 0;
-}
-
-int transcribe(uint32_t threads, char *lang, bool translate, bool tdrz,
-               float pcmf32[], size_t pcmf32_len, size_t *segs_out_len) {
-  whisper_full_params wparams =
-      whisper_full_default_params(WHISPER_SAMPLING_GREEDY);
-
-  wparams.n_threads = threads;
-  if (*lang != '\0')
-    wparams.language = lang;
-  else {
-    wparams.language = nullptr;
-  }
-
-  wparams.translate = translate;
-  wparams.debug_mode = true;
-  wparams.print_progress = true;
-  wparams.tdrz_enable = tdrz;
-
-  fprintf(stderr, "info: Enable tdrz: %d\n", tdrz);
-
-  if (whisper_full(ctx, wparams, pcmf32, pcmf32_len)) {
-    fprintf(stderr, "error: transcription failed\n");
-    return 1;
-  }
-
-  *segs_out_len = whisper_full_n_segments(ctx);
-
-  return 0;
-}
-
-const char *get_segment_text(int i) {
-  return whisper_full_get_segment_text(ctx, i);
-}
-
-int64_t get_segment_t0(int i) { return whisper_full_get_segment_t0(ctx, i); }
-
-int64_t get_segment_t1(int i) { return whisper_full_get_segment_t1(ctx, i); }
-
-int n_tokens(int i) { return whisper_full_n_tokens(ctx, i); }
-
-int32_t get_token_id(int i, int j) {
-  return whisper_full_get_token_id(ctx, i, j);
-}
-
-bool get_segment_speaker_turn_next(int i) {
-  return whisper_full_get_segment_speaker_turn_next(ctx, i);
-}

backend/go/whisper/gowhisper.go

@@ -1,161 +0,0 @@
-package main
-
-import (
-	"fmt"
-	"os"
-	"path/filepath"
-	"strings"
-	"unsafe"
-
-	"github.com/go-audio/wav"
-	"github.com/mudler/LocalAI/pkg/grpc/base"
-	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
-	"github.com/mudler/LocalAI/pkg/utils"
-)
-
-var (
-	CppLoadModel                 func(modelPath string) int
-	CppLoadModelVAD              func(modelPath string) int
-	CppVAD                       func(pcmf32 []float32, pcmf32Size uintptr, segsOut unsafe.Pointer, segsOutLen unsafe.Pointer) int
-	CppTranscribe                func(threads uint32, lang string, translate bool, diarize bool, pcmf32 []float32, pcmf32Len uintptr, segsOutLen unsafe.Pointer) int
-	CppGetSegmentText            func(i int) string
-	CppGetSegmentStart           func(i int) int64
-	CppGetSegmentEnd             func(i int) int64
-	CppNTokens                   func(i int) int
-	CppGetTokenID                func(i int, j int) int
-	CppGetSegmentSpeakerTurnNext func(i int) bool
-)
-
-type Whisper struct {
-	base.SingleThread
-}
-
-func (w *Whisper) Load(opts *pb.ModelOptions) error {
-	vadOnly := false
-
-	for _, oo := range opts.Options {
-		if oo == "vad_only" {
-			vadOnly = true
-		} else {
-			fmt.Fprintf(os.Stderr, "Unrecognized option: %v\n", oo)
-		}
-	}
-
-	if vadOnly {
-		if ret := CppLoadModelVAD(opts.ModelFile); ret != 0 {
-			return fmt.Errorf("Failed to load Whisper VAD model")
-		}
-
-		return nil
-	}
-
-	if ret := CppLoadModel(opts.ModelFile); ret != 0 {
-		return fmt.Errorf("Failed to load Whisper transcription model")
-	}
-
-	return nil
-}
-
-func (w *Whisper) VAD(req *pb.VADRequest) (pb.VADResponse, error) {
-	audio := req.Audio
-	// We expect 0xdeadbeef to be overwritten and if we see it in a stack trace we know it wasn't
-	segsPtr, segsLen := uintptr(0xdeadbeef), uintptr(0xdeadbeef)
-	segsPtrPtr, segsLenPtr := unsafe.Pointer(&segsPtr), unsafe.Pointer(&segsLen)
-
-	if ret := CppVAD(audio, uintptr(len(audio)), segsPtrPtr, segsLenPtr); ret != 0 {
-		return pb.VADResponse{}, fmt.Errorf("Failed VAD")
-	}
-
-	// Happens when CPP vector has not had any elements pushed to it
-	if segsPtr == 0 {
-		return pb.VADResponse{
-			Segments: []*pb.VADSegment{},
-		}, nil
-	}
-
-	// unsafeptr warning is caused by segsPtr being on the stack and therefor being subject to stack copying AFAICT
-	// however the stack shouldn't have grown between setting segsPtr and now, also the memory pointed to is allocated by C++
-	segs := unsafe.Slice((*float32)(unsafe.Pointer(segsPtr)), segsLen)
-
-	vadSegments := []*pb.VADSegment{}
-	for i := range len(segs) >> 1 {
-		s := segs[2*i] / 100
-		t := segs[2*i+1] / 100
-		vadSegments = append(vadSegments, &pb.VADSegment{
-			Start: s,
-			End:   t,
-		})
-	}
-
-	return pb.VADResponse{
-		Segments: vadSegments,
-	}, nil
-}
-
-func (w *Whisper) AudioTranscription(opts *pb.TranscriptRequest) (pb.TranscriptResult, error) {
-	dir, err := os.MkdirTemp("", "whisper")
-	if err != nil {
-		return pb.TranscriptResult{}, err
-	}
-	defer os.RemoveAll(dir)
-
-	convertedPath := filepath.Join(dir, "converted.wav")
-
-	if err := utils.AudioToWav(opts.Dst, convertedPath); err != nil {
-		return pb.TranscriptResult{}, err
-	}
-
-	// Open samples
-	fh, err := os.Open(convertedPath)
-	if err != nil {
-		return pb.TranscriptResult{}, err
-	}
-	defer fh.Close()
-
-	// Read samples
-	d := wav.NewDecoder(fh)
-	buf, err := d.FullPCMBuffer()
-	if err != nil {
-		return pb.TranscriptResult{}, err
-	}
-
-	data := buf.AsFloat32Buffer().Data
-	segsLen := uintptr(0xdeadbeef)
-	segsLenPtr := unsafe.Pointer(&segsLen)
-
-	if ret := CppTranscribe(opts.Threads, opts.Language, opts.Translate, opts.Diarize, data, uintptr(len(data)), segsLenPtr); ret != 0 {
-		return pb.TranscriptResult{}, fmt.Errorf("Failed Transcribe")
-	}
-
-	segments := []*pb.TranscriptSegment{}
-	text := ""
-	for i := range int(segsLen) {
-		s := CppGetSegmentStart(i)
-		t := CppGetSegmentEnd(i)
-		txt := strings.Clone(CppGetSegmentText(i))
-		tokens := make([]int32, CppNTokens(i))
-
-		if opts.Diarize && CppGetSegmentSpeakerTurnNext(i) {
-			txt += " [SPEAKER_TURN]"
-		}
-
-		for j := range tokens {
-			tokens[j] = int32(CppGetTokenID(i, j))
-		}
-		segment := &pb.TranscriptSegment{
-			Id:    int32(i),
-			Text:  txt,
-			Start: s, End: t,
-			Tokens: tokens,
-		}
-
-		segments = append(segments, segment)
-
-		text += " " + strings.TrimSpace(txt)
-	}
-
-	return pb.TranscriptResult{
-		Segments: segments,
-		Text:     strings.TrimSpace(text),
-	}, nil
-}

backend/go/whisper/gowhisper.h

@@ -1,17 +0,0 @@
-#include <cstddef>
-#include <cstdint>
-
-extern "C" {
-int load_model(const char *const model_path);
-int load_model_vad(const char *const model_path);
-int vad(float pcmf32[], size_t pcmf32_size, float **segs_out,
-        size_t *segs_out_len);
-int transcribe(uint32_t threads, char *lang, bool translate, bool tdrz,
-               float pcmf32[], size_t pcmf32_len, size_t *segs_out_len);
-const char *get_segment_text(int i);
-int64_t get_segment_t0(int i);
-int64_t get_segment_t1(int i);
-int n_tokens(int i);
-int32_t get_token_id(int i, int j);
-bool get_segment_speaker_turn_next(int i);
-}

backend/go/whisper/main.go

@@ -1,11 +1,10 @@
 package main
 
 // Note: this is started internally by LocalAI and a server is allocated for each model
+
 import (
 	"flag"
-	"os"
 
-	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
 )
 
@@ -13,40 +12,7 @@ var (
 	addr = flag.String("addr", "localhost:50051", "the address to connect to")
 )
 
-type LibFuncs struct {
-	FuncPtr any
-	Name    string
-}
-
 func main() {
-	// Get library name from environment variable, default to fallback
-	libName := os.Getenv("WHISPER_LIBRARY")
-	if libName == "" {
-		libName = "./libgowhisper-fallback.so"
-	}
-
-	gosd, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)
-	if err != nil {
-		panic(err)
-	}
-
-	libFuncs := []LibFuncs{
-		{&CppLoadModel, "load_model"},
-		{&CppLoadModelVAD, "load_model_vad"},
-		{&CppVAD, "vad"},
-		{&CppTranscribe, "transcribe"},
-		{&CppGetSegmentText, "get_segment_text"},
-		{&CppGetSegmentStart, "get_segment_t0"},
-		{&CppGetSegmentEnd, "get_segment_t1"},
-		{&CppNTokens, "n_tokens"},
-		{&CppGetTokenID, "get_token_id"},
-		{&CppGetSegmentSpeakerTurnNext, "get_segment_speaker_turn_next"},
-	}
-
-	for _, lf := range libFuncs {
-		purego.RegisterLibFunc(lf.FuncPtr, gosd, lf.Name)
-	}
-
 	flag.Parse()
 
 	if err := grpc.StartServer(*addr, &Whisper{}); err != nil {

backend/go/whisper/package.sh

@@ -10,9 +10,8 @@ CURDIR=$(dirname "$(realpath $0)")
 # Create lib directory
 mkdir -p $CURDIR/package/lib
 
-cp -avf $CURDIR/whisper $CURDIR/package/
-cp -fv $CURDIR/libgowhisper-*.so $CURDIR/package/
-cp -fv $CURDIR/run.sh $CURDIR/package/
+cp -avrf $CURDIR/whisper $CURDIR/package/
+cp -rfv $CURDIR/run.sh $CURDIR/package/
 
 # Detect architecture and copy appropriate libraries
 if [ -f "/lib64/ld-linux-x86-64.so.2" ]; then
@@ -43,13 +42,11 @@ elif [ -f "/lib/ld-linux-aarch64.so.1" ]; then
     cp -arfLv /lib/aarch64-linux-gnu/libdl.so.2 $CURDIR/package/lib/libdl.so.2
     cp -arfLv /lib/aarch64-linux-gnu/librt.so.1 $CURDIR/package/lib/librt.so.1
     cp -arfLv /lib/aarch64-linux-gnu/libpthread.so.0 $CURDIR/package/lib/libpthread.so.0
-elif [ $(uname -s) = "Darwin" ]; then
-    echo "Detected Darwin"
 else
     echo "Error: Could not detect architecture"
     exit 1
 fi
 
-echo "Packaging completed successfully"
+echo "Packaging completed successfully" 
 ls -liah $CURDIR/package/
-ls -liah $CURDIR/package/lib/
+ls -liah $CURDIR/package/lib/

backend/go/whisper/run.sh

@@ -1,52 +1,14 @@
 #!/bin/bash
 set -ex
 
-# Get the absolute current dir where the script is located
 CURDIR=$(dirname "$(realpath $0)")
 
-cd /
-
-echo "CPU info:"
-if [ "$(uname)" != "Darwin" ]; then
-	grep -e "model\sname" /proc/cpuinfo | head -1
-	grep -e "flags" /proc/cpuinfo | head -1
-fi
-
-LIBRARY="$CURDIR/libgowhisper-fallback.so"
-
-if [ "$(uname)" != "Darwin" ]; then
-	if grep -q -e "\savx\s" /proc/cpuinfo ; then
-		echo "CPU:    AVX    found OK"
-		if [ -e $CURDIR/libgowhisper-avx.so ]; then
-			LIBRARY="$CURDIR/libgowhisper-avx.so"
-		fi
-	fi
-
-	if grep -q -e "\savx2\s" /proc/cpuinfo ; then
-		echo "CPU:    AVX2   found OK"
-		if [ -e $CURDIR/libgowhisper-avx2.so ]; then
-			LIBRARY="$CURDIR/libgowhisper-avx2.so"
-		fi
-	fi
-
-	# Check avx 512
-	if grep -q -e "\savx512f\s" /proc/cpuinfo ; then
-		echo "CPU:    AVX512F found OK"
-		if [ -e $CURDIR/libgowhisper-avx512.so ]; then
-			LIBRARY="$CURDIR/libgowhisper-avx512.so"
-		fi
-	fi
-fi
-
 export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
-export WHISPER_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it
 if [ -f $CURDIR/lib/ld.so ]; then
 	echo "Using lib/ld.so"
-	echo "Using library: $LIBRARY"
 	exec $CURDIR/lib/ld.so $CURDIR/whisper "$@"
 fi
 
-echo "Using library: $LIBRARY"
 exec $CURDIR/whisper "$@"

backend/go/whisper/whisper.go

@@ -0,0 +1,105 @@
+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 (
+	"os"
+	"path/filepath"
+
+	"github.com/ggerganov/whisper.cpp/bindings/go/pkg/whisper"
+	"github.com/go-audio/wav"
+	"github.com/mudler/LocalAI/pkg/grpc/base"
+	pb "github.com/mudler/LocalAI/pkg/grpc/proto"
+	"github.com/mudler/LocalAI/pkg/utils"
+)
+
+type Whisper struct {
+	base.SingleThread
+	whisper whisper.Model
+}
+
+func (sd *Whisper) Load(opts *pb.ModelOptions) error {
+	// Note: the Model here is a path to a directory containing the model files
+	w, err := whisper.New(opts.ModelFile)
+	sd.whisper = w
+	return err
+}
+
+func (sd *Whisper) AudioTranscription(opts *pb.TranscriptRequest) (pb.TranscriptResult, error) {
+
+	dir, err := os.MkdirTemp("", "whisper")
+	if err != nil {
+		return pb.TranscriptResult{}, err
+	}
+	defer os.RemoveAll(dir)
+
+	convertedPath := filepath.Join(dir, "converted.wav")
+
+	if err := utils.AudioToWav(opts.Dst, convertedPath); err != nil {
+		return pb.TranscriptResult{}, err
+	}
+
+	// Open samples
+	fh, err := os.Open(convertedPath)
+	if err != nil {
+		return pb.TranscriptResult{}, err
+	}
+	defer fh.Close()
+
+	// Read samples
+	d := wav.NewDecoder(fh)
+	buf, err := d.FullPCMBuffer()
+	if err != nil {
+		return pb.TranscriptResult{}, err
+	}
+
+	data := buf.AsFloat32Buffer().Data
+
+	// Process samples
+	context, err := sd.whisper.NewContext()
+	if err != nil {
+		return pb.TranscriptResult{}, err
+
+	}
+
+	context.SetThreads(uint(opts.Threads))
+
+	if opts.Language != "" {
+		context.SetLanguage(opts.Language)
+	} else {
+		context.SetLanguage("auto")
+	}
+
+	if opts.Translate {
+		context.SetTranslate(true)
+	}
+
+	if err := context.Process(data, nil, nil, nil); err != nil {
+		return pb.TranscriptResult{}, err
+	}
+
+	segments := []*pb.TranscriptSegment{}
+	text := ""
+	for {
+		s, err := context.NextSegment()
+		if err != nil {
+			break
+		}
+
+		var tokens []int32
+		for _, t := range s.Tokens {
+			tokens = append(tokens, int32(t.Id))
+		}
+
+		segment := &pb.TranscriptSegment{Id: int32(s.Num), Text: s.Text, Start: int64(s.Start), End: int64(s.End), Tokens: tokens}
+		segments = append(segments, segment)
+
+		text += s.Text
+	}
+
+	return pb.TranscriptResult{
+		Segments: segments,
+		Text:     text,
+	}, nil
+
+}

backend/index.yaml

@@ -45,7 +45,6 @@
     default: "cpu-whisper"
     nvidia: "cuda12-whisper"
     intel: "intel-sycl-f16-whisper"
-    metal: "metal-whisper"
     amd: "rocm-whisper"
     vulkan: "vulkan-whisper"
     nvidia-l4t: "nvidia-l4t-arm64-whisper"
@@ -72,7 +71,7 @@
     # amd: "rocm-stablediffusion-ggml"
     vulkan: "vulkan-stablediffusion-ggml"
     nvidia-l4t: "nvidia-l4t-arm64-stablediffusion-ggml"
-    metal: "metal-stablediffusion-ggml"
+    # metal: "metal-stablediffusion-ggml"
     # darwin-x86: "darwin-x86-stablediffusion-ggml"
 - &rfdetr
   name: "rfdetr"
@@ -148,7 +147,7 @@
   uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-mlx-vlm"
   icon: https://avatars.githubusercontent.com/u/102832242?s=200&v=4
   urls:
-    - https://github.com/Blaizzy/mlx-vlm
+    - https://github.com/ml-explore/mlx-vlm
   mirrors:
     - localai/localai-backends:latest-metal-darwin-arm64-mlx-vlm
   license: MIT
@@ -160,23 +159,6 @@
     - vision-language
     - LLM
     - MLX
-- &mlx-audio
-  name: "mlx-audio"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-mlx-audio"
-  icon: https://avatars.githubusercontent.com/u/102832242?s=200&v=4
-  urls:
-    - https://github.com/Blaizzy/mlx-audio
-  mirrors:
-    - localai/localai-backends:latest-metal-darwin-arm64-mlx-audio
-  license: MIT
-  description: |
-      Run Audio Models with MLX
-  tags:
-    - audio-to-text
-    - audio-generation
-    - text-to-audio
-    - LLM
-    - MLX
 - &rerankers
   name: "rerankers"
   alias: "rerankers"
@@ -201,6 +183,8 @@
     nvidia: "cuda12-transformers"
     intel: "intel-transformers"
     amd: "rocm-transformers"
+    metal: "metal-transformers"
+    default: "cpu-transformers"
 - &diffusers
   name: "diffusers"
   icon: https://raw.githubusercontent.com/huggingface/diffusers/main/docs/source/en/imgs/diffusers_library.jpg
@@ -270,7 +254,6 @@
     nvidia: "cuda12-kokoro"
     intel: "intel-kokoro"
     amd: "rocm-kokoro"
-    nvidia-l4t: "nvidia-l4t-kokoro"
 - &coqui
   urls:
     - https://github.com/idiap/coqui-ai-TTS
@@ -351,9 +334,6 @@
   alias: "chatterbox"
   capabilities:
     nvidia: "cuda12-chatterbox"
-    metal: "metal-chatterbox"
-    default: "cpu-chatterbox"
-    nvidia-l4t: "nvidia-l4t-arm64-chatterbox"
 - &piper
   name: "piper"
   uri: "quay.io/go-skynet/local-ai-backends:latest-piper"
@@ -427,68 +407,6 @@
     - text-to-speech
     - TTS
   license: apache-2.0
-- &neutts
-  name: "neutts"
-  urls:
-    - https://github.com/neuphonic/neutts-air
-  description: |
-    NeuTTS Air is the world’s first super-realistic, on-device, TTS speech language model with instant voice cloning. Built off a 0.5B LLM backbone, NeuTTS Air brings natural-sounding speech, real-time performance, built-in security and speaker cloning to your local device - unlocking a new category of embedded voice agents, assistants, toys, and compliance-safe apps.
-  tags:
-    - text-to-speech
-    - TTS
-  license: apache-2.0
-  capabilities:
-    default: "cpu-neutts"
-    nvidia: "cuda12-neutts"
-    amd: "rocm-neutts"
-    nvidia-l4t: "nvidia-l4t-neutts"
-- !!merge <<: *neutts
-  name: "neutts-development"
-  capabilities:
-    default: "cpu-neutts-development"
-    nvidia: "cuda12-neutts-development"
-    amd: "rocm-neutts-development"
-    nvidia-l4t: "nvidia-l4t-neutts-development"
-- !!merge <<: *neutts
-  name: "cpu-neutts"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-neutts"
-  mirrors:
-    - localai/localai-backends:latest-cpu-neutts
-- !!merge <<: *neutts
-  name: "cuda12-neutts"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-neutts"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-cuda-12-neutts
-- !!merge <<: *neutts
-  name: "rocm-neutts"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-neutts"
-  mirrors:
-    - localai/localai-backends:latest-gpu-rocm-hipblas-neutts
-- !!merge <<: *neutts
-  name: "nvidia-l4t-neutts"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-arm64-neutts"
-  mirrors:
-    - localai/localai-backends:latest-nvidia-l4t-arm64-neutts
-- !!merge <<: *neutts
-  name: "cpu-neutts-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-cpu-neutts"
-  mirrors:
-    - localai/localai-backends:master-cpu-neutts
-- !!merge <<: *neutts
-  name: "cuda12-neutts-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-neutts"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-cuda-12-neutts
-- !!merge <<: *neutts
-  name: "rocm-neutts-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-rocm-hipblas-neutts"
-  mirrors:
-    - localai/localai-backends:master-gpu-rocm-hipblas-neutts
-- !!merge <<: *neutts
-  name: "nvidia-l4t-neutts-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-nvidia-l4t-arm64-neutts"
-  mirrors:
-    - localai/localai-backends:master-nvidia-l4t-arm64-neutts
 - !!merge <<: *mlx
   name: "mlx-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-mlx"
@@ -499,11 +417,6 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-mlx-vlm"
   mirrors:
     - localai/localai-backends:master-metal-darwin-arm64-mlx-vlm
-- !!merge <<: *mlx-audio
-  name: "mlx-audio-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-mlx-audio"
-  mirrors:
-    - localai/localai-backends:master-metal-darwin-arm64-mlx-audio
 - !!merge <<: *kitten-tts
   name: "kitten-tts-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-kitten-tts"
@@ -646,16 +559,6 @@
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-whisper"
   mirrors:
     - localai/localai-backends:latest-cpu-whisper
-- !!merge <<: *whispercpp
-  name: "metal-whisper"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-whisper"
-  mirrors:
-    - localai/localai-backends:latest-metal-darwin-arm64-whisper
-- !!merge <<: *whispercpp
-  name: "metal-whisper-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-whisper"
-  mirrors:
-    - localai/localai-backends:master-metal-darwin-arm64-whisper
 - !!merge <<: *whispercpp
   name: "cpu-whisper-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-whisper"
@@ -742,16 +645,6 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-stablediffusion-ggml"
   mirrors:
     - localai/localai-backends:master-cpu-stablediffusion-ggml
-- !!merge <<: *stablediffusionggml
-  name: "metal-stablediffusion-ggml"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:latest-metal-darwin-arm64-stablediffusion-ggml
-- !!merge <<: *stablediffusionggml
-  name: "metal-stablediffusion-ggml-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-stablediffusion-ggml"
-  mirrors:
-    - localai/localai-backends:master-metal-darwin-arm64-stablediffusion-ggml
 - !!merge <<: *stablediffusionggml
   name: "vulkan-stablediffusion-ggml"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-vulkan-stablediffusion-ggml"
@@ -960,6 +853,28 @@
     nvidia: "cuda12-transformers-development"
     intel: "intel-transformers-development"
     amd: "rocm-transformers-development"
+    default: "cpu-transformers-development"
+    metal: "metal-transformers-development"
+- !!merge <<: *transformers
+  name: "cpu-transformers"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-transformers"
+  mirrors:
+    - localai/localai-backends:latest-cpu-transformers
+- !!merge <<: *transformers
+  name: "cpu-transformers-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-cpu-transformers"
+  mirrors:
+    - localai/localai-backends:master-cpu-transformers
+- !!merge <<: *transformers
+  name: "metal-transformers"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-transformers"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-transformers
+- !!merge <<: *transformers
+  name: "metal-transformers-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-transformers"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-transformers
 - !!merge <<: *transformers
   name: "cuda12-transformers"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-transformers"
@@ -1113,7 +1028,6 @@
     nvidia: "cuda12-kokoro-development"
     intel: "intel-kokoro-development"
     amd: "rocm-kokoro-development"
-    nvidia-l4t: "nvidia-l4t-kokoro-development"
 - !!merge <<: *kokoro
   name: "cuda11-kokoro-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-11-kokoro"
@@ -1139,16 +1053,6 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-intel-kokoro"
   mirrors:
     - localai/localai-backends:master-gpu-intel-kokoro
-- !!merge <<: *kokoro
-  name: "nvidia-l4t-kokoro"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-l4t-kokoro"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-l4t-kokoro
-- !!merge <<: *kokoro
-  name: "nvidia-l4t-kokoro-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-l4t-kokoro"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-l4t-kokoro
 - !!merge <<: *kokoro
   name: "cuda11-kokoro"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-11-kokoro"
@@ -1300,39 +1204,6 @@
   name: "chatterbox-development"
   capabilities:
     nvidia: "cuda12-chatterbox-development"
-    metal: "metal-chatterbox-development"
-    default: "cpu-chatterbox-development"
-    nvidia-l4t: "nvidia-l4t-arm64-chatterbox"
-- !!merge <<: *chatterbox
-  name: "cpu-chatterbox"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-chatterbox"
-  mirrors:
-    - localai/localai-backends:latest-cpu-chatterbox
-- !!merge <<: *chatterbox
-  name: "cpu-chatterbox-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-cpu-chatterbox"
-  mirrors:
-    - localai/localai-backends:master-cpu-chatterbox
-- !!merge <<: *chatterbox
-  name: "nvidia-l4t-arm64-chatterbox"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-l4t-arm64-chatterbox"
-  mirrors:
-    - localai/localai-backends:latest-gpu-nvidia-l4t-arm64-chatterbox
-- !!merge <<: *chatterbox
-  name: "nvidia-l4t-arm64-chatterbox-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-l4t-arm64-chatterbox"
-  mirrors:
-    - localai/localai-backends:master-gpu-nvidia-l4t-arm64-chatterbox
-- !!merge <<: *chatterbox
-  name: "metal-chatterbox"
-  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-chatterbox"
-  mirrors:
-    - localai/localai-backends:latest-metal-darwin-arm64-chatterbox
-- !!merge <<: *chatterbox
-  name: "metal-chatterbox-development"
-  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-chatterbox"
-  mirrors:
-    - localai/localai-backends:master-metal-darwin-arm64-chatterbox
 - !!merge <<: *chatterbox
   name: "cuda12-chatterbox-development"
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-chatterbox"

backend/python/README.md

@@ -1,190 +1,38 @@
-# Python Backends for LocalAI
+# Common commands about conda environment
 
-This directory contains Python-based AI backends for LocalAI, providing support for various AI models and hardware acceleration targets.
+## Create a new empty conda environment
 
-## Overview
+```
+conda create --name <env-name> python=<your version> -y
 
-The Python backends use a unified build system based on `libbackend.sh` that provides:
-- **Automatic virtual environment management** with support for both `uv` and `pip`
-- **Hardware-specific dependency installation** (CPU, CUDA, Intel, MLX, etc.)
-- **Portable Python support** for standalone deployments
-- **Consistent backend execution** across different environments
-
-## Available Backends
-
-### Core AI Models
-- **transformers** - Hugging Face Transformers framework (PyTorch-based)
-- **vllm** - High-performance LLM inference engine
-- **mlx** - Apple Silicon optimized ML framework
-- **exllama2** - ExLlama2 quantized models
-
-### Audio & Speech
-- **bark** - Text-to-speech synthesis
-- **coqui** - Coqui TTS models
-- **faster-whisper** - Fast Whisper speech recognition
-- **kitten-tts** - Lightweight TTS
-- **mlx-audio** - Apple Silicon audio processing
-- **chatterbox** - TTS model
-- **kokoro** - TTS models
-
-### Computer Vision
-- **diffusers** - Stable Diffusion and image generation
-- **mlx-vlm** - Vision-language models for Apple Silicon
-- **rfdetr** - Object detection models
-
-### Specialized
-
-- **rerankers** - Text reranking models
-
-## Quick Start
-
-### Prerequisites
-- Python 3.10+ (default: 3.10.18)
-- `uv` package manager (recommended) or `pip`
-- Appropriate hardware drivers for your target (CUDA, Intel, etc.)
-
-### Installation
-
-Each backend can be installed individually:
-
-```bash
-# Navigate to a specific backend
-cd backend/python/transformers
-
-# Install dependencies
-make transformers
-# or
-bash install.sh
-
-# Run the backend
-make run
-# or
-bash run.sh
+conda create --name autogptq python=3.11 -y
 ```
 
-### Using the Unified Build System
+## To activate the environment
 
-The `libbackend.sh` script provides consistent commands across all backends:
-
-```bash
-# Source the library in your backend script
-source $(dirname $0)/../common/libbackend.sh
-
-# Install requirements (automatically handles hardware detection)
-installRequirements
-
-# Start the backend server
-startBackend $@
-
-# Run tests
-runUnittests
+As of conda 4.4
+```
+conda activate autogptq
 ```
 
-## Hardware Targets
+The conda version older than 4.4
 
-The build system automatically detects and configures for different hardware:
-
-- **CPU** - Standard CPU-only builds
-- **CUDA** - NVIDIA GPU acceleration (supports CUDA 11/12)
-- **Intel** - Intel XPU/GPU optimization
-- **MLX** - Apple Silicon (M1/M2/M3) optimization
-- **HIP** - AMD GPU acceleration
-
-### Target-Specific Requirements
-
-Backends can specify hardware-specific dependencies:
-- `requirements.txt` - Base requirements
-- `requirements-cpu.txt` - CPU-specific packages
-- `requirements-cublas11.txt` - CUDA 11 packages
-- `requirements-cublas12.txt` - CUDA 12 packages
-- `requirements-intel.txt` - Intel-optimized packages
-- `requirements-mps.txt` - Apple Silicon packages
-
-## Configuration Options
-
-### Environment Variables
-
-- `PYTHON_VERSION` - Python version (default: 3.10)
-- `PYTHON_PATCH` - Python patch version (default: 18)
-- `BUILD_TYPE` - Force specific build target
-- `USE_PIP` - Use pip instead of uv (default: false)
-- `PORTABLE_PYTHON` - Enable portable Python builds
-- `LIMIT_TARGETS` - Restrict backend to specific targets
-
-### Example: CUDA 12 Only Backend
-
-```bash
-# In your backend script
-LIMIT_TARGETS="cublas12"
-source $(dirname $0)/../common/libbackend.sh
+```
+source activate autogptq
 ```
 
-### Example: Intel-Optimized Backend
+## Install the packages to your environment
 
-```bash
-# In your backend script
-LIMIT_TARGETS="intel"
-source $(dirname $0)/../common/libbackend.sh
+Sometimes you need to install the packages from the conda-forge channel
+
+By using `conda`
+```
+conda install <your-package-name>
+
+conda install -c conda-forge <your package-name>
 ```
 
-## Development
-
-### Adding a New Backend
-
-1. Create a new directory in `backend/python/`
-2. Copy the template structure from `common/template/`
-3. Implement your `backend.py` with the required gRPC interface
-4. Add appropriate requirements files for your target hardware
-5. Use `libbackend.sh` for consistent build and execution
-
-### Testing
-
-```bash
-# Run backend tests
-make test
-# or
-bash test.sh
+Or by using `pip`
 ```
-
-### Building
-
-```bash
-# Install dependencies
-make <backend-name>
-
-# Clean build artifacts
-make clean
+pip install <your-package-name>
 ```
-
-## Architecture
-
-Each backend follows a consistent structure:
-```
-backend-name/
-├── backend.py          # Main backend implementation
-├── requirements.txt    # Base dependencies
-├── requirements-*.txt  # Hardware-specific dependencies
-├── install.sh         # Installation script
-├── run.sh            # Execution script
-├── test.sh           # Test script
-├── Makefile          # Build targets
-└── test.py           # Unit tests
-```
-
-## Troubleshooting
-
-### Common Issues
-
-1. **Missing dependencies**: Ensure all requirements files are properly configured
-2. **Hardware detection**: Check that `BUILD_TYPE` matches your system
-3. **Python version**: Verify Python 3.10+ is available
-4. **Virtual environment**: Use `ensureVenv` to create/activate environments
-
-## Contributing
-
-When adding new backends or modifying existing ones:
-1. Follow the established directory structure
-2. Use `libbackend.sh` for consistent behavior
-3. Include appropriate requirements files for all target hardware
-4. Add comprehensive tests
-5. Update this README if adding new backend types

backend/python/bark/requirements.txt

@@ -1,4 +1,4 @@
 bark==0.1.5
-grpcio==1.76.0
+grpcio==1.74.0
 protobuf
 certifi

backend/python/chatterbox/backend.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 """
-This is an extra gRPC server of LocalAI for Chatterbox TTS
+This is an extra gRPC server of LocalAI for Bark TTS
 """
 from concurrent import futures
 import time
@@ -14,98 +14,15 @@ import backend_pb2_grpc
 import torch
 import torchaudio as ta
 from chatterbox.tts import ChatterboxTTS
-from chatterbox.mtl_tts import ChatterboxMultilingualTTS
+
 import grpc
-import tempfile
 
-def is_float(s):
-    """Check if a string can be converted to float."""
-    try:
-        float(s)
-        return True
-    except ValueError:
-        return False
-def is_int(s):
-    """Check if a string can be converted to int."""
-    try:
-        int(s)
-        return True
-    except ValueError:
-        return False
-
-def split_text_at_word_boundary(text, max_length=250):
-    """
-    Split text at word boundaries without truncating words.
-    Returns a list of text chunks.
-    """
-    if not text or len(text) <= max_length:
-        return [text]
-    
-    chunks = []
-    words = text.split()
-    current_chunk = ""
-    
-    for word in words:
-        # Check if adding this word would exceed the limit
-        if len(current_chunk) + len(word) + 1 <= max_length:
-            if current_chunk:
-                current_chunk += " " + word
-            else:
-                current_chunk = word
-        else:
-            # If current chunk is not empty, add it to chunks
-            if current_chunk:
-                chunks.append(current_chunk)
-                current_chunk = word
-            else:
-                # If a single word is longer than max_length, we have to include it anyway
-                chunks.append(word)
-                current_chunk = ""
-    
-    # Add the last chunk if it's not empty
-    if current_chunk:
-        chunks.append(current_chunk)
-    
-    return chunks
-
-def merge_audio_files(audio_files, output_path, sample_rate):
-    """
-    Merge multiple audio files into a single audio file.
-    """
-    if not audio_files:
-        return
-    
-    if len(audio_files) == 1:
-        # If only one file, just copy it
-        import shutil
-        shutil.copy2(audio_files[0], output_path)
-        return
-    
-    # Load all audio files
-    waveforms = []
-    for audio_file in audio_files:
-        waveform, sr = ta.load(audio_file)
-        if sr != sample_rate:
-            # Resample if necessary
-            resampler = ta.transforms.Resample(sr, sample_rate)
-            waveform = resampler(waveform)
-        waveforms.append(waveform)
-    
-    # Concatenate all waveforms
-    merged_waveform = torch.cat(waveforms, dim=1)
-    
-    # Save the merged audio
-    ta.save(output_path, merged_waveform, sample_rate)
-    
-    # Clean up temporary files
-    for audio_file in audio_files:
-        if os.path.exists(audio_file):
-            os.remove(audio_file)
 
 _ONE_DAY_IN_SECONDS = 60 * 60 * 24
 
 # If MAX_WORKERS are specified in the environment use it, otherwise default to 1
 MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
+COQUI_LANGUAGE = os.environ.get('COQUI_LANGUAGE', None)
 
 # Implement the BackendServicer class with the service methods
 class BackendServicer(backend_pb2_grpc.BackendServicer):
@@ -130,28 +47,6 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
         if not torch.cuda.is_available() and request.CUDA:
             return backend_pb2.Result(success=False, message="CUDA is not available")
 
-
-        options = request.Options
-
-        # empty dict
-        self.options = {}
-
-        # The options are a list of strings in this form optname:optvalue
-        # We are storing all the options in a dict so we can use it later when
-        # generating the images
-        for opt in options:
-            if ":" not in opt:
-                continue
-            key, value = opt.split(":")
-            # if value is a number, convert it to the appropriate type
-            if is_float(value):
-                value = float(value)
-            elif is_int(value):
-                value = int(value)
-            elif value.lower() in ["true", "false"]:
-                value = value.lower() == "true"
-            self.options[key] = value
-
         self.AudioPath = None
 
         if os.path.isabs(request.AudioPath):
@@ -161,14 +56,10 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
             modelFileBase = os.path.dirname(request.ModelFile)
             # modify LoraAdapter to be relative to modelFileBase
             self.AudioPath = os.path.join(modelFileBase, request.AudioPath)
+
         try:
             print("Preparing models, please wait", file=sys.stderr)
-            if "multilingual" in self.options:
-                # remove key from options
-                del self.options["multilingual"]
-                self.model = ChatterboxMultilingualTTS.from_pretrained(device=device)
-            else:
-                self.model = ChatterboxTTS.from_pretrained(device=device)
+            self.model = ChatterboxTTS.from_pretrained(device=device)
         except Exception as err:
             return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
         # Implement your logic here for the LoadModel service
@@ -177,43 +68,14 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
 
     def TTS(self, request, context):
         try:
-            kwargs = {}
-
-            if "language" in self.options:
-                kwargs["language_id"] = self.options["language"]
+            # Generate audio using ChatterboxTTS
             if self.AudioPath is not None:
-                kwargs["audio_prompt_path"] = self.AudioPath
-
-            # add options to kwargs
-            kwargs.update(self.options)
-
-            # Check if text exceeds 250 characters
-            # (chatterbox does not support long text)
-            # https://github.com/resemble-ai/chatterbox/issues/60
-            # https://github.com/resemble-ai/chatterbox/issues/110
-            if len(request.text) > 250:
-                # Split text at word boundaries
-                text_chunks = split_text_at_word_boundary(request.text, max_length=250)
-                print(f"Splitting text into chunks of 250 characters: {len(text_chunks)}", file=sys.stderr)
-                # Generate audio for each chunk
-                temp_audio_files = []
-                for i, chunk in enumerate(text_chunks):
-                    # Generate audio for this chunk
-                    wav = self.model.generate(chunk, **kwargs)
-                    
-                    # Create temporary file for this chunk
-                    temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.wav')
-                    temp_file.close()
-                    ta.save(temp_file.name, wav, self.model.sr)
-                    temp_audio_files.append(temp_file.name)
-                
-                # Merge all audio files
-                merge_audio_files(temp_audio_files, request.dst, self.model.sr)
+                wav = self.model.generate(request.text, audio_prompt_path=self.AudioPath)
             else:
-                # Generate audio using ChatterboxTTS for short text
-                wav = self.model.generate(request.text, **kwargs)
-                # Save the generated audio
-                ta.save(request.dst, wav, self.model.sr)
+                wav = self.model.generate(request.text)
+            
+            # Save the generated audio
+            ta.save(request.dst, wav, self.model.sr)
             
         except Exception as err:
             return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")

backend/python/chatterbox/install.sh

@@ -15,6 +15,5 @@ fi
 if [ "x${BUILD_PROFILE}" == "xintel" ]; then
     EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
 fi
-EXTRA_PIP_INSTALL_FLAGS+=" --no-build-isolation"
 
 installRequirements

backend/python/chatterbox/requirements-cpu.txt

@@ -1,9 +1,5 @@
---extra-index-url https://download.pytorch.org/whl/cpu
 accelerate
-torch
-torchaudio
-numpy>=1.24.0,<1.26.0
-transformers
-# https://github.com/mudler/LocalAI/pull/6240#issuecomment-3329518289
-chatterbox-tts@git+https://git@github.com/mudler/chatterbox.git@faster
-#chatterbox-tts==0.1.4
+torch==2.6.0
+torchaudio==2.6.0
+transformers==4.46.3
+chatterbox-tts

backend/python/chatterbox/requirements-cublas11.txt

@@ -2,7 +2,5 @@
 torch==2.6.0+cu118
 torchaudio==2.6.0+cu118
 transformers==4.46.3
-numpy>=1.24.0,<1.26.0
-# https://github.com/mudler/LocalAI/pull/6240#issuecomment-3329518289
-chatterbox-tts@git+https://git@github.com/mudler/chatterbox.git@faster
+chatterbox-tts
 accelerate

backend/python/chatterbox/requirements-cublas12.txt

@@ -1,7 +1,5 @@
-torch
-torchaudio
-transformers
-numpy>=1.24.0,<1.26.0
-# https://github.com/mudler/LocalAI/pull/6240#issuecomment-3329518289
-chatterbox-tts@git+https://git@github.com/mudler/chatterbox.git@faster
+torch==2.6.0
+torchaudio==2.6.0
+transformers==4.46.3
+chatterbox-tts
 accelerate

backend/python/chatterbox/requirements-hipblas.txt

@@ -1,8 +1,6 @@
 --extra-index-url https://download.pytorch.org/whl/rocm6.0
 torch==2.6.0+rocm6.1
 torchaudio==2.6.0+rocm6.1
-transformers
-numpy>=1.24.0,<1.26.0
-# https://github.com/mudler/LocalAI/pull/6240#issuecomment-3329518289
-chatterbox-tts@git+https://git@github.com/mudler/chatterbox.git@faster
+transformers==4.46.3
+chatterbox-tts
 accelerate

backend/python/chatterbox/requirements-intel.txt

@@ -2,11 +2,10 @@
 intel-extension-for-pytorch==2.3.110+xpu
 torch==2.3.1+cxx11.abi
 torchaudio==2.3.1+cxx11.abi
-transformers
-numpy>=1.24.0,<1.26.0
-# https://github.com/mudler/LocalAI/pull/6240#issuecomment-3329518289
-chatterbox-tts@git+https://git@github.com/mudler/chatterbox.git@faster
+transformers==4.46.3
+chatterbox-tts
 accelerate
 oneccl_bind_pt==2.3.100+xpu
 optimum[openvino]
-setuptools
+setuptools
+accelerate

backend/python/chatterbox/requirements-l4t.txt

@@ -1,7 +0,0 @@
---extra-index-url https://pypi.jetson-ai-lab.io/jp6/cu126/
-torch
-torchaudio
-transformers
-numpy>=1.24.0,<1.26.0
-chatterbox-tts@git+https://git@github.com/mudler/chatterbox.git@faster
-accelerate

backend/python/chatterbox/requirements.txt

@@ -2,5 +2,4 @@ grpcio==1.71.0
 protobuf
 certifi
 packaging
-setuptools
-poetry
+setuptools

backend/python/common/libbackend.sh

@@ -286,8 +286,7 @@ _makeVenvPortable() {
 function ensureVenv() {
     local interpreter=""
 
-    if [ "x${PORTABLE_PYTHON}" == "xtrue" ] || [ -e "$(_portable_python)" ]; then
-        echo "Using portable Python"
+    if [ "x${PORTABLE_PYTHON}" == "xtrue" ]; then
         ensurePortablePython
         interpreter="$(_portable_python)"
     else
@@ -385,11 +384,6 @@ function installRequirements() {
         requirementFiles+=("${EDIR}/requirements-${BUILD_PROFILE}-after.txt")
     fi
 
-    # This is needed to build wheels that e.g. depends on Python.h
-    if [ "x${PORTABLE_PYTHON}" == "xtrue" ]; then
-        export C_INCLUDE_PATH="${C_INCLUDE_PATH:-}:$(_portable_dir)/include/python${PYTHON_VERSION}"
-    fi
-
     for reqFile in ${requirementFiles[@]}; do
         if [ -f "${reqFile}" ]; then
             echo "starting requirements install for ${reqFile}"

backend/python/common/template/requirements.txt

@@ -1,3 +1,3 @@
-grpcio==1.76.0
+grpcio==1.74.0
 protobuf
 grpcio-tools

backend/python/coqui/requirements.txt

@@ -1,4 +1,4 @@
-grpcio==1.76.0
+grpcio==1.74.0
 protobuf
 certifi
 packaging==24.1

backend/python/diffusers/backend.py

@@ -18,7 +18,7 @@ import backend_pb2_grpc
 import grpc
 
 from diffusers import SanaPipeline, StableDiffusion3Pipeline, StableDiffusionXLPipeline, StableDiffusionDepth2ImgPipeline, DPMSolverMultistepScheduler, StableDiffusionPipeline, DiffusionPipeline, \
-    EulerAncestralDiscreteScheduler, FluxPipeline, FluxTransformer2DModel, QwenImageEditPipeline, AutoencoderKLWan, WanPipeline, WanImageToVideoPipeline
+    EulerAncestralDiscreteScheduler, FluxPipeline, FluxTransformer2DModel, QwenImageEditPipeline
 from diffusers import StableDiffusionImg2ImgPipeline, AutoPipelineForText2Image, ControlNetModel, StableVideoDiffusionPipeline, Lumina2Text2ImgPipeline
 from diffusers.pipelines.stable_diffusion import safety_checker
 from diffusers.utils import load_image, export_to_video
@@ -66,21 +66,19 @@ from diffusers.schedulers import (
 )
 
 def is_float(s):
-    """Check if a string can be converted to float."""
     try:
         float(s)
         return True
     except ValueError:
         return False
+
 def is_int(s):
-    """Check if a string can be converted to int."""
     try:
         int(s)
         return True
     except ValueError:
         return False
 
-
 # The scheduler list mapping was taken from here: https://github.com/neggles/animatediff-cli/blob/6f336f5f4b5e38e85d7f06f1744ef42d0a45f2a7/src/animatediff/schedulers.py#L39
 # Credits to https://github.com/neggles
 # See https://github.com/huggingface/diffusers/issues/4167 for more details on sched mapping from A1111
@@ -189,8 +187,6 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
                     value = float(value)
                 elif is_int(value):
                     value = int(value)
-                elif value.lower() in ["true", "false"]:
-                    value = value.lower() == "true"
                 self.options[key] = value
 
             # From options, extract if present "torch_dtype" and set it to the appropriate type
@@ -338,32 +334,6 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
                     torch_dtype=torch.bfloat16)
                 self.pipe.vae.to(torch.bfloat16)
                 self.pipe.text_encoder.to(torch.bfloat16)
-            elif request.PipelineType == "WanPipeline":
-                # WAN2.2 pipeline requires special VAE handling
-                vae = AutoencoderKLWan.from_pretrained(
-                    request.Model, 
-                    subfolder="vae", 
-                    torch_dtype=torch.float32
-                )
-                self.pipe = WanPipeline.from_pretrained(
-                    request.Model,
-                    vae=vae,
-                    torch_dtype=torchType
-                )
-                self.txt2vid = True  # WAN2.2 is a text-to-video pipeline
-            elif request.PipelineType == "WanImageToVideoPipeline":
-                # WAN2.2 image-to-video pipeline
-                vae = AutoencoderKLWan.from_pretrained(
-                    request.Model, 
-                    subfolder="vae", 
-                    torch_dtype=torch.float32
-                )
-                self.pipe = WanImageToVideoPipeline.from_pretrained(
-                    request.Model,
-                    vae=vae,
-                    torch_dtype=torchType
-                )
-                self.img2vid = True  # WAN2.2 image-to-video pipeline
 
             if CLIPSKIP and request.CLIPSkip != 0:
                 self.clip_skip = request.CLIPSkip
@@ -505,24 +475,11 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
             "num_inference_steps": steps,
         }
 
-        # Handle image source: prioritize RefImages over request.src
-        image_src = None
-        if hasattr(request, 'ref_images') and request.ref_images and len(request.ref_images) > 0:
-            # Use the first reference image if available
-            image_src = request.ref_images[0]
-            print(f"Using reference image: {image_src}", file=sys.stderr)
-        elif request.src != "":
-            # Fall back to request.src if no ref_images
-            image_src = request.src
-            print(f"Using source image: {image_src}", file=sys.stderr)
-        else:
-            print("No image source provided", file=sys.stderr)
-        
-        if image_src and not self.controlnet and not self.img2vid:
-            image = Image.open(image_src)
+        if request.src != "" and not self.controlnet and not self.img2vid:
+            image = Image.open(request.src)
             options["image"] = image
-        elif self.controlnet and image_src:
-            pose_image = load_image(image_src)
+        elif self.controlnet and request.src:
+            pose_image = load_image(request.src)
             options["image"] = pose_image
 
         if CLIPSKIP and self.clip_skip != 0:
@@ -564,11 +521,7 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
 
         if self.img2vid:
             # Load the conditioning image
-            if image_src:
-                image = load_image(image_src)
-            else:
-                # Fallback to request.src for img2vid if no ref_images
-                image = load_image(request.src)
+            image = load_image(request.src)
             image = image.resize((1024, 576))
 
             generator = torch.manual_seed(request.seed)
@@ -605,96 +558,6 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
 
         return backend_pb2.Result(message="Media generated", success=True)
 
-    def GenerateVideo(self, request, context):
-        try:
-            prompt = request.prompt
-            if not prompt:
-                return backend_pb2.Result(success=False, message="No prompt provided for video generation")
-
-            # Set default values from request or use defaults
-            num_frames = request.num_frames if request.num_frames > 0 else 81
-            fps = request.fps if request.fps > 0 else 16
-            cfg_scale = request.cfg_scale if request.cfg_scale > 0 else 4.0
-            num_inference_steps = request.step if request.step > 0 else 40
-            
-            # Prepare generation parameters
-            kwargs = {
-                "prompt": prompt,
-                "negative_prompt": request.negative_prompt if request.negative_prompt else "",
-                "height": request.height if request.height > 0 else 720,
-                "width": request.width if request.width > 0 else 1280,
-                "num_frames": num_frames,
-                "guidance_scale": cfg_scale,
-                "num_inference_steps": num_inference_steps,
-            }
-
-            # Add custom options from self.options (including guidance_scale_2 if specified)
-            kwargs.update(self.options)
-
-            # Set seed if provided
-            if request.seed > 0:
-                kwargs["generator"] = torch.Generator(device=self.device).manual_seed(request.seed)
-
-            # Handle start and end images for video generation
-            if request.start_image:
-                kwargs["start_image"] = load_image(request.start_image)
-            if request.end_image:
-                kwargs["end_image"] = load_image(request.end_image)
-
-            print(f"Generating video with {kwargs=}", file=sys.stderr)
-
-            # Generate video frames based on pipeline type
-            if self.PipelineType == "WanPipeline":
-                # WAN2.2 text-to-video generation
-                output = self.pipe(**kwargs)
-                frames = output.frames[0]  # WAN2.2 returns frames in this format
-            elif self.PipelineType == "WanImageToVideoPipeline":
-                # WAN2.2 image-to-video generation
-                if request.start_image:
-                    # Load and resize the input image according to WAN2.2 requirements
-                    image = load_image(request.start_image)
-                    # Use request dimensions or defaults, but respect WAN2.2 constraints
-                    request_height = request.height if request.height > 0 else 480
-                    request_width = request.width if request.width > 0 else 832
-                    max_area = request_height * request_width
-                    aspect_ratio = image.height / image.width
-                    mod_value = self.pipe.vae_scale_factor_spatial * self.pipe.transformer.config.patch_size[1]
-                    height = round((max_area * aspect_ratio) ** 0.5 / mod_value) * mod_value
-                    width = round((max_area / aspect_ratio) ** 0.5 / mod_value) * mod_value
-                    image = image.resize((width, height))
-                    kwargs["image"] = image
-                    kwargs["height"] = height
-                    kwargs["width"] = width
-                
-                output = self.pipe(**kwargs)
-                frames = output.frames[0]
-            elif self.img2vid:
-                # Generic image-to-video generation
-                if request.start_image:
-                    image = load_image(request.start_image)
-                    image = image.resize((request.width if request.width > 0 else 1024, 
-                                       request.height if request.height > 0 else 576))
-                    kwargs["image"] = image
-                
-                output = self.pipe(**kwargs)
-                frames = output.frames[0]
-            elif self.txt2vid:
-                # Generic text-to-video generation
-                output = self.pipe(**kwargs)
-                frames = output.frames[0]
-            else:
-                return backend_pb2.Result(success=False, message=f"Pipeline {self.PipelineType} does not support video generation")
-
-            # Export video
-            export_to_video(frames, request.dst, fps=fps)
-            
-            return backend_pb2.Result(message="Video generated successfully", success=True)
-
-        except Exception as err:
-            print(f"Error generating video: {err}", file=sys.stderr)
-            traceback.print_exc()
-            return backend_pb2.Result(success=False, message=f"Error generating video: {err}")
-
 
 def serve(address):
     server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS),

backend/python/diffusers/requirements-cpu.txt

@@ -8,5 +8,4 @@ compel
 peft
 sentencepiece
 torch==2.7.1
-optimum-quanto
-ftfy
+optimum-quanto

backend/python/diffusers/requirements-cublas11.txt

@@ -1,12 +1,11 @@
 --extra-index-url https://download.pytorch.org/whl/cu118
+torch==2.7.1+cu118
+torchvision==0.22.1+cu118
 git+https://github.com/huggingface/diffusers
 opencv-python
 transformers
-torchvision==0.22.1
 accelerate
 compel
 peft
 sentencepiece
-torch==2.7.1
-optimum-quanto
-ftfy
+optimum-quanto

backend/python/diffusers/requirements-cublas12.txt

@@ -1,12 +1,10 @@
---extra-index-url https://download.pytorch.org/whl/cu121
+torch==2.7.1
+torchvision==0.22.1
 git+https://github.com/huggingface/diffusers
 opencv-python
 transformers
-torchvision
 accelerate
 compel
 peft
 sentencepiece
-torch
-ftfy
-optimum-quanto
+optimum-quanto

backend/python/diffusers/requirements-hipblas.txt

@@ -8,5 +8,4 @@ accelerate
 compel
 peft
 sentencepiece
-optimum-quanto
-ftfy
+optimum-quanto

backend/python/diffusers/requirements-intel.txt

@@ -12,5 +12,4 @@ accelerate
 compel
 peft
 sentencepiece
-optimum-quanto
-ftfy
+optimum-quanto

backend/python/diffusers/requirements-l4t.txt

@@ -8,5 +8,4 @@ peft
 optimum-quanto
 numpy<2
 sentencepiece
-torchvision
-ftfy
+torchvision

backend/python/diffusers/requirements-mps.txt

@@ -7,5 +7,4 @@ accelerate
 compel
 peft
 sentencepiece
-optimum-quanto
-ftfy
+optimum-quanto

backend/python/diffusers/requirements.txt

@@ -1,5 +1,5 @@
 setuptools
-grpcio==1.76.0
+grpcio==1.74.0
 pillow
 protobuf
 certifi

backend/python/diffusers/test.py

@@ -31,7 +31,7 @@ class TestBackendServicer(unittest.TestCase):
         """
         This method tests if the server starts up successfully
         """
-        time.sleep(20)
+        time.sleep(10)
         try:
             self.setUp()
             with grpc.insecure_channel("localhost:50051") as channel:
@@ -48,7 +48,7 @@ class TestBackendServicer(unittest.TestCase):
         """
         This method tests if the model is loaded successfully
         """
-        time.sleep(20)
+        time.sleep(10)
         try:
             self.setUp()
             with grpc.insecure_channel("localhost:50051") as channel:
@@ -66,7 +66,7 @@ class TestBackendServicer(unittest.TestCase):
         """
         This method tests if the backend can generate images
         """
-        time.sleep(20)
+        time.sleep(10)
         try:
             self.setUp()
             with grpc.insecure_channel("localhost:50051") as channel:

backend/python/exllama2/requirements.txt

@@ -1,4 +1,4 @@
-grpcio==1.76.0
+grpcio==1.74.0
 protobuf
 certifi
 wheel

backend/python/kokoro/backend.py

@@ -64,15 +64,15 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
             # Generate audio using Kokoro pipeline
             generator = self.pipeline(request.text, voice=voice)
             
-            speechs = []
-            # Get all the audio segment
+            # Get the first (and typically only) audio segment
             for i, (gs, ps, audio) in enumerate(generator):
-                speechs.append(audio)
+                # Save audio to the destination file
+                sf.write(request.dst, audio, 24000)
                 print(f"Generated audio segment {i}: gs={gs}, ps={ps}", file=sys.stderr)
-            # Merges the audio segments and writes them to the destination
-            speech = torch.cat(speechs, dim=0)
-            sf.write(request.dst, speech, 24000)
-
+                # For now, we only process the first segment
+                # If you need to handle multiple segments, you might want to modify this
+                break
+                
         except Exception as err:
             return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
         

backend/python/kokoro/requirements-l4t.txt

@@ -1,7 +0,0 @@
---extra-index-url https://pypi.jetson-ai-lab.io/jp6/cu126/
-torch
-torchaudio
-transformers
-accelerate
-kokoro
-soundfile

backend/python/mlx-audio/Makefile

@@ -1,23 +0,0 @@
-.PHONY: mlx-audio
-mlx-audio:
-	bash install.sh
-
-.PHONY: run
-run: mlx-audio
-	@echo "Running mlx-audio..."
-	bash run.sh
-	@echo "mlx run."
-
-.PHONY: test
-test: mlx-audio
-	@echo "Testing mlx-audio..."
-	bash test.sh
-	@echo "mlx tested."
-
-.PHONY: protogen-clean
-protogen-clean:
-	$(RM) backend_pb2_grpc.py backend_pb2.py
-
-.PHONY: clean
-clean: protogen-clean
-	rm -rf venv __pycache__

backend/python/mlx-audio/backend.py

@@ -1,465 +0,0 @@
-#!/usr/bin/env python3
-import asyncio
-from concurrent import futures
-import argparse
-import signal
-import sys
-import os
-import shutil
-import glob
-from typing import List
-import time
-import tempfile
-
-import backend_pb2
-import backend_pb2_grpc
-
-import grpc
-from mlx_audio.tts.utils import load_model
-import soundfile as sf
-import numpy as np
-import uuid
-
-def is_float(s):
-    """Check if a string can be converted to float."""
-    try:
-        float(s)
-        return True
-    except ValueError:
-        return False
-def is_int(s):
-    """Check if a string can be converted to int."""
-    try:
-        int(s)
-        return True
-    except ValueError:
-        return False
-
-_ONE_DAY_IN_SECONDS = 60 * 60 * 24
-
-# If MAX_WORKERS are specified in the environment use it, otherwise default to 1
-MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
-
-# Implement the BackendServicer class with the service methods
-class BackendServicer(backend_pb2_grpc.BackendServicer):
-    """
-    A gRPC servicer that implements the Backend service defined in backend.proto.
-    This backend provides TTS (Text-to-Speech) functionality using MLX-Audio.
-    """
-
-    def Health(self, request, context):
-        """
-        Returns a health check message.
-
-        Args:
-            request: The health check request.
-            context: The gRPC context.
-
-        Returns:
-            backend_pb2.Reply: The health check reply.
-        """
-        return backend_pb2.Reply(message=bytes("OK", 'utf-8'))
-
-    async def LoadModel(self, request, context):
-        """
-        Loads a TTS model using MLX-Audio.
-
-        Args:
-            request: The load model request.
-            context: The gRPC context.
-
-        Returns:
-            backend_pb2.Result: The load model result.
-        """
-        try:
-            print(f"Loading MLX-Audio TTS model: {request.Model}", file=sys.stderr)
-            print(f"Request: {request}", file=sys.stderr)
-            
-            # Parse options like in the kokoro backend
-            options = request.Options
-            self.options = {}
-            
-            # The options are a list of strings in this form optname:optvalue
-            # We store all the options in a dict for later use
-            for opt in options:
-                if ":" not in opt:
-                    continue
-                key, value = opt.split(":", 1)  # Split only on first colon to handle values with colons
-                
-                # Convert numeric values to appropriate types
-                if is_float(value):
-                    value = float(value)
-                elif is_int(value):
-                    value = int(value)
-                elif value.lower() in ["true", "false"]:
-                    value = value.lower() == "true"
-                    
-                self.options[key] = value
-            
-            print(f"Options: {self.options}", file=sys.stderr)
-            
-            # Load the model using MLX-Audio's load_model function
-            try:
-                self.tts_model = load_model(request.Model)
-                self.model_path = request.Model
-                print(f"TTS model loaded successfully from {request.Model}", file=sys.stderr)
-            except Exception as model_err:
-                print(f"Error loading TTS model: {model_err}", file=sys.stderr)
-                return backend_pb2.Result(success=False, message=f"Failed to load model: {model_err}")
-                
-        except Exception as err:
-            print(f"Error loading MLX-Audio TTS model {err=}, {type(err)=}", file=sys.stderr)
-            return backend_pb2.Result(success=False, message=f"Error loading MLX-Audio TTS model: {err}")
-
-        print("MLX-Audio TTS model loaded successfully", file=sys.stderr)
-        return backend_pb2.Result(message="MLX-Audio TTS model loaded successfully", success=True)
-
-    def TTS(self, request, context):
-        """
-        Generates TTS audio from text using MLX-Audio.
-
-        Args:
-            request: A TTSRequest object containing text, model, destination, voice, and language.
-            context: A grpc.ServicerContext object that provides information about the RPC.
-
-        Returns:
-            A Result object indicating success or failure.
-        """
-        try:
-            # Check if model is loaded
-            if not hasattr(self, 'tts_model') or self.tts_model is None:
-                return backend_pb2.Result(success=False, message="TTS model not loaded. Please call LoadModel first.")
-            
-            print(f"Generating TTS with MLX-Audio - text: {request.text[:50]}..., voice: {request.voice}, language: {request.language}", file=sys.stderr)
-            
-            # Handle speed parameter based on model type
-            speed_value = self._handle_speed_parameter(request, self.model_path)
-            
-            # Map language names to codes if needed
-            lang_code = self._map_language_code(request.language, request.voice)
-            
-            # Prepare generation parameters
-            gen_params = {
-                "text": request.text,
-                "speed": speed_value,
-                "verbose": False,
-            }
-            
-            # Add model-specific parameters
-            if request.voice and request.voice.strip():
-                gen_params["voice"] = request.voice
-            
-            # Check if model supports language codes (primarily Kokoro)
-            if "kokoro" in self.model_path.lower():
-                gen_params["lang_code"] = lang_code
-            
-            # Add pitch and gender for Spark models
-            if "spark" in self.model_path.lower():
-                gen_params["pitch"] = 1.0  # Default to moderate
-                gen_params["gender"] = "female"  # Default to female
-            
-            print(f"Generation parameters: {gen_params}", file=sys.stderr)
-            
-            # Generate audio using the loaded model
-            try:
-                results = self.tts_model.generate(**gen_params)
-            except Exception as gen_err:
-                print(f"Error during TTS generation: {gen_err}", file=sys.stderr)
-                return backend_pb2.Result(success=False, message=f"TTS generation failed: {gen_err}")
-            
-            # Process the generated audio segments
-            audio_arrays = []
-            for segment in results:
-                audio_arrays.append(segment.audio)
-            
-            # If no segments, return error
-            if not audio_arrays:
-                print("No audio segments generated", file=sys.stderr)
-                return backend_pb2.Result(success=False, message="No audio generated")
-            
-            # Concatenate all segments
-            cat_audio = np.concatenate(audio_arrays, axis=0)
-            
-            # Generate output filename and path
-            if request.dst:
-                output_path = request.dst
-            else:
-                unique_id = str(uuid.uuid4())
-                filename = f"tts_{unique_id}.wav"
-                output_path = filename
-            
-            # Write the audio as a WAV
-            try:
-                sf.write(output_path, cat_audio, 24000)
-                print(f"Successfully wrote audio file to {output_path}", file=sys.stderr)
-                
-                # Verify the file exists and has content
-                if not os.path.exists(output_path):
-                    print(f"File was not created at {output_path}", file=sys.stderr)
-                    return backend_pb2.Result(success=False, message="Failed to create audio file")
-                
-                file_size = os.path.getsize(output_path)
-                if file_size == 0:
-                    print("File was created but is empty", file=sys.stderr)
-                    return backend_pb2.Result(success=False, message="Generated audio file is empty")
-                
-                print(f"Audio file size: {file_size} bytes", file=sys.stderr)
-                
-            except Exception as write_err:
-                print(f"Error writing audio file: {write_err}", file=sys.stderr)
-                return backend_pb2.Result(success=False, message=f"Failed to save audio: {write_err}")
-            
-            return backend_pb2.Result(success=True, message=f"TTS audio generated successfully: {output_path}")
-                
-        except Exception as e:
-            print(f"Error in MLX-Audio TTS: {e}", file=sys.stderr)
-            return backend_pb2.Result(success=False, message=f"TTS generation failed: {str(e)}")
-
-    async def Predict(self, request, context):
-        """
-        Generates TTS audio based on the given prompt using MLX-Audio TTS.
-        This is a fallback method for compatibility with the Predict endpoint.
-
-        Args:
-            request: The predict request.
-            context: The gRPC context.
-
-        Returns:
-            backend_pb2.Reply: The predict result.
-        """
-        try:
-            # Check if model is loaded
-            if not hasattr(self, 'tts_model') or self.tts_model is None:
-                context.set_code(grpc.StatusCode.FAILED_PRECONDITION)
-                context.set_details("TTS model not loaded. Please call LoadModel first.")
-                return backend_pb2.Reply(message=bytes("", encoding='utf-8'))
-            
-            # For TTS, we expect the prompt to contain the text to synthesize
-            if not request.Prompt:
-                context.set_code(grpc.StatusCode.INVALID_ARGUMENT)
-                context.set_details("Prompt is required for TTS generation")
-                return backend_pb2.Reply(message=bytes("", encoding='utf-8'))
-            
-            # Handle speed parameter based on model type
-            speed_value = self._handle_speed_parameter(request, self.model_path)
-            
-            # Map language names to codes if needed
-            lang_code = self._map_language_code(None, None)  # Use defaults for Predict
-            
-            # Prepare generation parameters
-            gen_params = {
-                "text": request.Prompt,
-                "speed": speed_value,
-                "verbose": False,
-            }
-            
-            # Add model-specific parameters
-            if hasattr(self, 'options') and 'voice' in self.options:
-                gen_params["voice"] = self.options['voice']
-            
-            # Check if model supports language codes (primarily Kokoro)
-            if "kokoro" in self.model_path.lower():
-                gen_params["lang_code"] = lang_code
-            
-            print(f"Generating TTS with MLX-Audio - text: {request.Prompt[:50]}..., params: {gen_params}", file=sys.stderr)
-            
-            # Generate audio using the loaded model
-            try:
-                results = self.tts_model.generate(**gen_params)
-            except Exception as gen_err:
-                print(f"Error during TTS generation: {gen_err}", file=sys.stderr)
-                context.set_code(grpc.StatusCode.INTERNAL)
-                context.set_details(f"TTS generation failed: {gen_err}")
-                return backend_pb2.Reply(message=bytes("", encoding='utf-8'))
-            
-            # Process the generated audio segments
-            audio_arrays = []
-            for segment in results:
-                audio_arrays.append(segment.audio)
-            
-            # If no segments, return error
-            if not audio_arrays:
-                print("No audio segments generated", file=sys.stderr)
-                return backend_pb2.Reply(message=bytes("No audio generated", encoding='utf-8'))
-            
-            # Concatenate all segments
-            cat_audio = np.concatenate(audio_arrays, axis=0)
-            duration = len(cat_audio) / 24000  # Assuming 24kHz sample rate
-            
-            # Return success message with audio information
-            response = f"TTS audio generated successfully. Duration: {duration:.2f}s, Sample rate: 24000Hz"
-            return backend_pb2.Reply(message=bytes(response, encoding='utf-8'))
-            
-        except Exception as e:
-            print(f"Error in MLX-Audio TTS Predict: {e}", file=sys.stderr)
-            context.set_code(grpc.StatusCode.INTERNAL)
-            context.set_details(f"TTS generation failed: {str(e)}")
-            return backend_pb2.Reply(message=bytes("", encoding='utf-8'))
-
-    def _handle_speed_parameter(self, request, model_path):
-        """
-        Handle speed parameter based on model type.
-        
-        Args:
-            request: The TTSRequest object.
-            model_path: The model path to determine model type.
-            
-        Returns:
-            float: The processed speed value.
-        """
-        # Get speed from options if available
-        speed = 1.0
-        if hasattr(self, 'options') and 'speed' in self.options:
-            speed = self.options['speed']
-        
-        # Handle speed parameter based on model type
-        if "spark" in model_path.lower():
-            # Spark actually expects float values that map to speed descriptions
-            speed_map = {
-                "very_low": 0.0,
-                "low": 0.5,
-                "moderate": 1.0,
-                "high": 1.5,
-                "very_high": 2.0,
-            }
-            if isinstance(speed, str) and speed in speed_map:
-                speed_value = speed_map[speed]
-            else:
-                # Try to use as float, default to 1.0 (moderate) if invalid
-                try:
-                    speed_value = float(speed)
-                    if speed_value not in [0.0, 0.5, 1.0, 1.5, 2.0]:
-                        speed_value = 1.0  # Default to moderate
-                except:
-                    speed_value = 1.0  # Default to moderate
-        else:
-            # Other models use float speed values
-            try:
-                speed_value = float(speed)
-                if speed_value < 0.5 or speed_value > 2.0:
-                    speed_value = 1.0  # Default to 1.0 if out of range
-            except ValueError:
-                speed_value = 1.0  # Default to 1.0 if invalid
-        
-        return speed_value
-
-    def _map_language_code(self, language, voice):
-        """
-        Map language names to codes if needed.
-        
-        Args:
-            language: The language parameter from the request.
-            voice: The voice parameter from the request.
-            
-        Returns:
-            str: The language code.
-        """
-        if not language:
-            # Default to voice[0] if not found
-            return voice[0] if voice else "a"
-        
-        # Map language names to codes if needed
-        language_map = {
-            "american_english": "a",
-            "british_english": "b",
-            "spanish": "e",
-            "french": "f",
-            "hindi": "h",
-            "italian": "i",
-            "portuguese": "p",
-            "japanese": "j",
-            "mandarin_chinese": "z",
-            # Also accept direct language codes
-            "a": "a", "b": "b", "e": "e", "f": "f", "h": "h", "i": "i", "p": "p", "j": "j", "z": "z",
-        }
-        
-        return language_map.get(language.lower(), language)
-
-    def _build_generation_params(self, request, default_speed=1.0):
-        """
-        Build generation parameters from request attributes and options for MLX-Audio TTS.
-
-        Args:
-            request: The gRPC request.
-            default_speed: Default speed if not specified.
-
-        Returns:
-            dict: Generation parameters for MLX-Audio
-        """
-        # Initialize generation parameters for MLX-Audio TTS
-        generation_params = {
-            'speed': default_speed,
-            'voice': 'af_heart',  # Default voice
-            'lang_code': 'a',     # Default language code
-        }
-        
-        # Extract parameters from request attributes
-        if hasattr(request, 'Temperature') and request.Temperature > 0:
-            # Temperature could be mapped to speed variation
-            generation_params['speed'] = 1.0 + (request.Temperature - 0.5) * 0.5
-        
-        # Override with options if available
-        if hasattr(self, 'options'):
-            # Speed from options
-            if 'speed' in self.options:
-                generation_params['speed'] = self.options['speed']
-            
-            # Voice from options
-            if 'voice' in self.options:
-                generation_params['voice'] = self.options['voice']
-            
-            # Language code from options
-            if 'lang_code' in self.options:
-                generation_params['lang_code'] = self.options['lang_code']
-            
-            # Model-specific parameters
-            param_option_mapping = {
-                'temp': 'speed',
-                'temperature': 'speed',
-                'top_p': 'speed',  # Map top_p to speed variation
-            }
-            
-            for option_key, param_key in param_option_mapping.items():
-                if option_key in self.options:
-                    if param_key == 'speed':
-                        # Ensure speed is within reasonable bounds
-                        speed_val = float(self.options[option_key])
-                        if 0.5 <= speed_val <= 2.0:
-                            generation_params[param_key] = speed_val
-        
-        return generation_params
-
-async def serve(address):
-    # Start asyncio gRPC server
-    server = grpc.aio.server(migration_thread_pool=futures.ThreadPoolExecutor(max_workers=MAX_WORKERS),
-        options=[
-            ('grpc.max_message_length', 50 * 1024 * 1024),  # 50MB
-            ('grpc.max_send_message_length', 50 * 1024 * 1024),  # 50MB
-            ('grpc.max_receive_message_length', 50 * 1024 * 1024),  # 50MB
-        ])
-    # Add the servicer to the server
-    backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
-    # Bind the server to the address
-    server.add_insecure_port(address)
-
-    # Gracefully shutdown the server on SIGTERM or SIGINT
-    loop = asyncio.get_event_loop()
-    for sig in (signal.SIGINT, signal.SIGTERM):
-        loop.add_signal_handler(
-            sig, lambda: asyncio.ensure_future(server.stop(5))
-        )
-
-    # Start the server
-    await server.start()
-    print("MLX-Audio TTS Server started. Listening on: " + address, file=sys.stderr)
-    # Wait for the server to be terminated
-    await server.wait_for_termination()
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description="Run the MLX-Audio TTS gRPC server.")
-    parser.add_argument(
-        "--addr", default="localhost:50051", help="The address to bind the server to."
-    )
-    args = parser.parse_args()
-
-    asyncio.run(serve(args.addr))

backend/python/mlx-audio/install.sh

@@ -1,14 +0,0 @@
-#!/bin/bash
-set -e
-
-USE_PIP=true
-
-backend_dir=$(dirname $0)
-
-if [ -d $backend_dir/common ]; then
-    source $backend_dir/common/libbackend.sh
-else
-    source $backend_dir/../common/libbackend.sh
-fi
-
-installRequirements

backend/python/mlx-audio/requirements-mps.txt

@@ -1 +0,0 @@
-git+https://github.com/Blaizzy/mlx-audio

backend/python/mlx-audio/requirements.txt

@@ -1,7 +0,0 @@
-grpcio==1.71.0
-protobuf
-certifi
-setuptools
-mlx-audio
-soundfile
-numpy

backend/python/mlx-audio/run.sh

@@ -1,11 +0,0 @@
-#!/bin/bash
-
-backend_dir=$(dirname $0)
-
-if [ -d $backend_dir/common ]; then
-    source $backend_dir/common/libbackend.sh
-else
-    source $backend_dir/../common/libbackend.sh
-fi
-
-startBackend $@

backend/python/mlx-audio/test.py

@@ -1,142 +0,0 @@
-import unittest
-import subprocess
-import time
-import backend_pb2
-import backend_pb2_grpc
-
-import grpc
-
-import unittest
-import subprocess
-import time
-import grpc
-import backend_pb2_grpc
-import backend_pb2
-
-class TestBackendServicer(unittest.TestCase):
-    """
-    TestBackendServicer is the class that tests the gRPC service.
-
-    This class contains methods to test the startup and shutdown of the gRPC service.
-    """
-    def setUp(self):
-        self.service = subprocess.Popen(["python", "backend.py", "--addr", "localhost:50051"])
-        time.sleep(10)
-
-    def tearDown(self) -> None:
-        self.service.terminate()
-        self.service.wait()
-
-    def test_server_startup(self):
-        try:
-            self.setUp()
-            with grpc.insecure_channel("localhost:50051") as channel:
-                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.Health(backend_pb2.HealthMessage())
-                self.assertEqual(response.message, b'OK')
-        except Exception as err:
-            print(err)
-            self.fail("Server failed to start")
-        finally:
-            self.tearDown()
-    def test_load_model(self):
-        """
-        This method tests if the TTS model is loaded successfully
-        """
-        try:
-            self.setUp()
-            with grpc.insecure_channel("localhost:50051") as channel:
-                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.LoadModel(backend_pb2.ModelOptions(Model="mlx-community/Kokoro-82M-4bit"))
-                self.assertTrue(response.success)
-                self.assertEqual(response.message, "MLX-Audio TTS model loaded successfully")
-        except Exception as err:
-            print(err)
-            self.fail("LoadModel service failed")
-        finally:
-            self.tearDown()
-
-    def test_tts_generation(self):
-        """
-        This method tests if TTS audio is generated successfully
-        """
-        try:
-            self.setUp()
-            with grpc.insecure_channel("localhost:50051") as channel:
-                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.LoadModel(backend_pb2.ModelOptions(Model="mlx-community/Kokoro-82M-4bit"))
-                self.assertTrue(response.success)
-                
-                # Test TTS generation
-                tts_req = backend_pb2.TTSRequest(
-                    text="Hello, this is a test of the MLX-Audio TTS system.",
-                    model="mlx-community/Kokoro-82M-4bit",
-                    voice="af_heart",
-                    language="a"
-                )
-                tts_resp = stub.TTS(tts_req)
-                self.assertTrue(tts_resp.success)
-                self.assertIn("TTS audio generated successfully", tts_resp.message)
-        except Exception as err:
-            print(err)
-            self.fail("TTS service failed")
-        finally:
-            self.tearDown()
-
-    def test_tts_with_options(self):
-        """
-        This method tests if TTS works with various options and parameters
-        """
-        try:
-            self.setUp()
-            with grpc.insecure_channel("localhost:50051") as channel:
-                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.LoadModel(backend_pb2.ModelOptions(
-                    Model="mlx-community/Kokoro-82M-4bit",
-                    Options=["voice:af_soft", "speed:1.2", "lang_code:b"]
-                ))
-                self.assertTrue(response.success)
-                
-                # Test TTS generation with different voice and language
-                tts_req = backend_pb2.TTSRequest(
-                    text="Hello, this is a test with British English accent.",
-                    model="mlx-community/Kokoro-82M-4bit",
-                    voice="af_soft",
-                    language="b"
-                )
-                tts_resp = stub.TTS(tts_req)
-                self.assertTrue(tts_resp.success)
-                self.assertIn("TTS audio generated successfully", tts_resp.message)
-        except Exception as err:
-            print(err)
-            self.fail("TTS with options service failed")
-        finally:
-            self.tearDown()
-
-
-    def test_tts_multilingual(self):
-        """
-        This method tests if TTS works with different languages
-        """
-        try:
-            self.setUp()
-            with grpc.insecure_channel("localhost:50051") as channel:
-                stub = backend_pb2_grpc.BackendStub(channel)
-                response = stub.LoadModel(backend_pb2.ModelOptions(Model="mlx-community/Kokoro-82M-4bit"))
-                self.assertTrue(response.success)
-                
-                # Test Spanish TTS
-                tts_req = backend_pb2.TTSRequest(
-                    text="Hola, esto es una prueba del sistema TTS MLX-Audio.",
-                    model="mlx-community/Kokoro-82M-4bit",
-                    voice="af_heart",
-                    language="e"
-                )
-                tts_resp = stub.TTS(tts_req)
-                self.assertTrue(tts_resp.success)
-                self.assertIn("TTS audio generated successfully", tts_resp.message)
-        except Exception as err:
-            print(err)
-            self.fail("Multilingual TTS service failed")
-        finally:
-            self.tearDown()

backend/python/mlx-audio/test.sh

@@ -1,12 +0,0 @@
-#!/bin/bash
-set -e
-
-backend_dir=$(dirname $0)
-
-if [ -d $backend_dir/common ]; then
-    source $backend_dir/common/libbackend.sh
-else
-    source $backend_dir/../common/libbackend.sh
-fi
-
-runUnittests

backend/python/mlx-vlm/backend.py

@@ -21,21 +21,6 @@ import io
 from PIL import Image
 import tempfile
 
-def is_float(s):
-    """Check if a string can be converted to float."""
-    try:
-        float(s)
-        return True
-    except ValueError:
-        return False
-def is_int(s):
-    """Check if a string can be converted to int."""
-    try:
-        int(s)
-        return True
-    except ValueError:
-        return False
-
 _ONE_DAY_IN_SECONDS = 60 * 60 * 24
 
 # If MAX_WORKERS are specified in the environment use it, otherwise default to 1
@@ -47,6 +32,22 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
     A gRPC servicer that implements the Backend service defined in backend.proto.
     """
 
+    def _is_float(self, s):
+        """Check if a string can be converted to float."""
+        try:
+            float(s)
+            return True
+        except ValueError:
+            return False
+
+    def _is_int(self, s):
+        """Check if a string can be converted to int."""
+        try:
+            int(s)
+            return True
+        except ValueError:
+            return False
+
     def Health(self, request, context):
         """
         Returns a health check message.
@@ -86,9 +87,10 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
                     continue
                 key, value = opt.split(":", 1)  # Split only on first colon to handle values with colons
                 
-                if is_float(value):
+                # Convert numeric values to appropriate types
+                if self._is_float(value):
                     value = float(value)
-                elif is_int(value):
+                elif self._is_int(value):
                     value = int(value)
                 elif value.lower() in ["true", "false"]:
                     value = value.lower() == "true"

backend/python/mlx/backend.py

@@ -24,27 +24,28 @@ _ONE_DAY_IN_SECONDS = 60 * 60 * 24
 # If MAX_WORKERS are specified in the environment use it, otherwise default to 1
 MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
 
-def is_float(s):
-    """Check if a string can be converted to float."""
-    try:
-        float(s)
-        return True
-    except ValueError:
-        return False
-def is_int(s):
-    """Check if a string can be converted to int."""
-    try:
-        int(s)
-        return True
-    except ValueError:
-        return False
-
 # Implement the BackendServicer class with the service methods
 class BackendServicer(backend_pb2_grpc.BackendServicer):
     """
     A gRPC servicer that implements the Backend service defined in backend.proto.
     """
 
+    def _is_float(self, s):
+        """Check if a string can be converted to float."""
+        try:
+            float(s)
+            return True
+        except ValueError:
+            return False
+
+    def _is_int(self, s):
+        """Check if a string can be converted to int."""
+        try:
+            int(s)
+            return True
+        except ValueError:
+            return False
+
     def Health(self, request, context):
         """
         Returns a health check message.
@@ -85,9 +86,9 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
                 key, value = opt.split(":", 1)  # Split only on first colon to handle values with colons
                 
                 # Convert numeric values to appropriate types
-                if is_float(value):
+                if self._is_float(value):
                     value = float(value)
-                elif is_int(value):
+                elif self._is_int(value):
                     value = int(value)
                 elif value.lower() in ["true", "false"]:
                     value = value.lower() == "true"

backend/python/neutts/Makefile

@@ -1,23 +0,0 @@
-.PHONY: neutts
-neutts:
-	bash install.sh
-
-.PHONY: run
-run: neutts
-	@echo "Running neutts..."
-	bash run.sh
-	@echo "neutts run."
-
-.PHONY: test
-test: neutts
-	@echo "Testing neutts..."
-	bash test.sh
-	@echo "neutts tested."
-
-.PHONY: protogen-clean
-protogen-clean:
-	$(RM) backend_pb2_grpc.py backend_pb2.py
-
-.PHONY: clean
-clean: protogen-clean
-	rm -rf venv __pycache__

backend/python/neutts/backend.py

@@ -1,162 +0,0 @@
-#!/usr/bin/env python3
-"""
-This is an extra gRPC server of LocalAI for NeuTTSAir
-"""
-from concurrent import futures
-import time
-import argparse
-import signal
-import sys
-import os
-import backend_pb2
-import backend_pb2_grpc
-import torch
-from neuttsair.neutts import NeuTTSAir
-import soundfile as sf
-
-import grpc
-
-def is_float(s):
-    """Check if a string can be converted to float."""
-    try:
-        float(s)
-        return True
-    except ValueError:
-        return False
-def is_int(s):
-    """Check if a string can be converted to int."""
-    try:
-        int(s)
-        return True
-    except ValueError:
-        return False
-
-_ONE_DAY_IN_SECONDS = 60 * 60 * 24
-
-# If MAX_WORKERS are specified in the environment use it, otherwise default to 1
-MAX_WORKERS = int(os.environ.get('PYTHON_GRPC_MAX_WORKERS', '1'))
-
-# Implement the BackendServicer class with the service methods
-class BackendServicer(backend_pb2_grpc.BackendServicer):
-    """
-    BackendServicer is the class that implements the gRPC service
-    """
-    def Health(self, request, context):
-        return backend_pb2.Reply(message=bytes("OK", 'utf-8'))
-    def LoadModel(self, request, context):
-
-        # Get device
-        # device = "cuda" if request.CUDA else "cpu"
-        if torch.cuda.is_available():
-            print("CUDA is available", file=sys.stderr)
-            device = "cuda"
-        else:
-            print("CUDA is not available", file=sys.stderr)
-            device = "cpu"
-        mps_available = hasattr(torch.backends, "mps") and torch.backends.mps.is_available()
-        if mps_available:
-            device = "mps"
-        if not torch.cuda.is_available() and request.CUDA:
-            return backend_pb2.Result(success=False, message="CUDA is not available")
-
-
-        options = request.Options
-
-        # empty dict
-        self.options = {}
-        self.ref_text = None
-
-        # The options are a list of strings in this form optname:optvalue
-        # We are storing all the options in a dict so we can use it later when
-        # generating the images
-        for opt in options:
-            if ":" not in opt:
-                continue
-            key, value = opt.split(":")
-            # if value is a number, convert it to the appropriate type
-            if is_float(value):
-                value = float(value)
-            elif is_int(value):
-                value = int(value)
-            elif value.lower() in ["true", "false"]:
-                value = value.lower() == "true"
-            self.options[key] = value
-
-        codec_repo = "neuphonic/neucodec"
-        if "codec_repo" in self.options:
-            codec_repo = self.options["codec_repo"]
-            del self.options["codec_repo"]
-        if "ref_text" in self.options:
-            self.ref_text = self.options["ref_text"]
-            del self.options["ref_text"]
-
-        self.AudioPath = None
-
-        if os.path.isabs(request.AudioPath):
-            self.AudioPath = request.AudioPath
-        elif request.AudioPath and request.ModelFile != "" and not os.path.isabs(request.AudioPath):
-            # get base path of modelFile
-            modelFileBase = os.path.dirname(request.ModelFile)
-            # modify LoraAdapter to be relative to modelFileBase
-            self.AudioPath = os.path.join(modelFileBase, request.AudioPath)
-        try:
-            print("Preparing models, please wait", file=sys.stderr)
-            self.model = NeuTTSAir(backbone_repo=request.Model, backbone_device=device, codec_repo=codec_repo, codec_device=device)
-        except Exception as err:
-            return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
-        # Implement your logic here for the LoadModel service
-        # Replace this with your desired response
-        return backend_pb2.Result(message="Model loaded successfully", success=True)
-
-    def TTS(self, request, context):
-        try:
-            kwargs = {}
-
-            # add options to kwargs
-            kwargs.update(self.options)
-
-            ref_codes = self.model.encode_reference(self.AudioPath)
-
-            wav = self.model.infer(request.text, ref_codes, self.ref_text)
-
-            sf.write(request.dst, wav, 24000)            
-        except Exception as err:
-            return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
-        return backend_pb2.Result(success=True)
-
-def serve(address):
-    server = grpc.server(futures.ThreadPoolExecutor(max_workers=MAX_WORKERS),
-        options=[
-            ('grpc.max_message_length', 50 * 1024 * 1024),  # 50MB
-            ('grpc.max_send_message_length', 50 * 1024 * 1024),  # 50MB
-            ('grpc.max_receive_message_length', 50 * 1024 * 1024),  # 50MB
-        ])
-    backend_pb2_grpc.add_BackendServicer_to_server(BackendServicer(), server)
-    server.add_insecure_port(address)
-    server.start()
-    print("Server started. Listening on: " + address, file=sys.stderr)
-
-    # Define the signal handler function
-    def signal_handler(sig, frame):
-        print("Received termination signal. Shutting down...")
-        server.stop(0)
-        sys.exit(0)
-
-    # Set the signal handlers for SIGINT and SIGTERM
-    signal.signal(signal.SIGINT, signal_handler)
-    signal.signal(signal.SIGTERM, signal_handler)
-
-    try:
-        while True:
-            time.sleep(_ONE_DAY_IN_SECONDS)
-    except KeyboardInterrupt:
-        server.stop(0)
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser(description="Run the gRPC server.")
-    parser.add_argument(
-        "--addr", default="localhost:50051", help="The address to bind the server to."
-    )
-    args = parser.parse_args()
-
-    serve(args.addr)

backend/python/neutts/install.sh

@@ -1,33 +0,0 @@
-#!/bin/bash
-set -e
-
-backend_dir=$(dirname $0)
-if [ -d $backend_dir/common ]; then
-    source $backend_dir/common/libbackend.sh
-else
-    source $backend_dir/../common/libbackend.sh
-fi
-
-# This is here because the Intel pip index is broken and returns 200 status codes for every package name, it just doesn't return any package links.
-# This makes uv think that the package exists in the Intel pip index, and by default it stops looking at other pip indexes once it finds a match.
-# We need uv to continue falling through to the pypi default index to find optimum[openvino] in the pypi index
-# the --upgrade actually allows us to *downgrade* torch to the version provided in the Intel pip index
-if [ "x${BUILD_PROFILE}" == "xintel" ]; then
-    EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
-fi
-
-if [ "x${BUILD_TYPE}" == "xcublas" ] || [ "x${BUILD_TYPE}" == "xl4t" ]; then
-    export CMAKE_ARGS="-DGGML_CUDA=on"
-fi
-
-if [ "x${BUILD_TYPE}" == "xhipblas" ]; then
-    export CMAKE_ARGS="-DGGML_HIPBLAS=on"
-fi
-
-EXTRA_PIP_INSTALL_FLAGS+=" --no-build-isolation"
-
-git clone https://github.com/neuphonic/neutts-air neutts-air
-
-cp -rfv neutts-air/neuttsair ./
-
-installRequirements

backend/python/neutts/requirements-after.txt

@@ -1,2 +0,0 @@
-datasets==4.1.1
-torchtune==0.6.1