Feature: llamacpp support + strict collection naming (optional)

.env.example

@@ -90,3 +90,12 @@ SPLITTER_TYPE=ast
 
 # Whether to use hybrid search mode. If true, it will use both dense vector and BM25; if false, it will use only dense vector search.
 # HYBRID_MODE=true
+
+# =============================================================================
+# Collection Naming Configuration
+# =============================================================================
+
+# Whether to use strict collection naming that includes provider and model info
+# This prevents conflicts when switching between different embedding providers/models
+# If false (default), uses legacy naming for backward compatibility
+# EMBEDDING_STRICT_COLLECTION_NAMES=false

README.md

@@ -543,7 +543,7 @@ Claude Context is a monorepo containing three main packages:
 
 ### Supported Technologies
 
-- **Embedding Providers**: [OpenAI](https://openai.com), [VoyageAI](https://voyageai.com), [Ollama](https://ollama.ai), [Gemini](https://gemini.google.com)
+- **Embedding Providers**: [OpenAI](https://openai.com), [VoyageAI](https://voyageai.com), [Ollama](https://ollama.ai), [Gemini](https://gemini.google.com), [LlamaCpp](https://github.com/ggerganov/llama.cpp) (local inference on consumer hardware)
 - **Vector Databases**: [Milvus](https://milvus.io) or [Zilliz Cloud](https://zilliz.com/cloud)(fully managed vector database as a service)
 - **Code Splitters**: AST-based splitter (with automatic fallback), LangChain character-based splitter
 - **Languages**: TypeScript, JavaScript, Python, Java, C++, C#, Go, Rust, PHP, Ruby, Swift, Kotlin, Scala, Markdown

docs/getting-started/environment-variables.md

@@ -20,7 +20,7 @@ Claude Context supports a global configuration file at `~/.context/.env` to simp
 ### Embedding Provider
 | Variable | Description | Default |
 |----------|-------------|---------|
-| `EMBEDDING_PROVIDER` | Provider: `OpenAI`, `VoyageAI`, `Gemini`, `Ollama` | `OpenAI` |
+| `EMBEDDING_PROVIDER` | Provider: `OpenAI`, `VoyageAI`, `Gemini`, `Ollama`, `LlamaCpp` | `OpenAI` |
 | `EMBEDDING_MODEL` | Embedding model name (works for all providers) | Provider-specific default |
 | `OPENAI_API_KEY` | OpenAI API key | Required for OpenAI |
 | `OPENAI_BASE_URL` | OpenAI API base URL (optional, for custom endpoints) | `https://api.openai.com/v1` |
@@ -47,13 +47,29 @@ Claude Context supports a global configuration file at `~/.context/.env` to simp
 |----------|-------------|---------|
 | `MILVUS_TOKEN` | Milvus authentication token. Get [Zilliz Personal API Key](https://github.com/zilliztech/claude-context/blob/master/assets/signup_and_get_apikey.png) | Recommended |
 | `MILVUS_ADDRESS` | Milvus server address. Optional when using Zilliz Personal API Key | Auto-resolved from token |
+| `MILVUS_COLLECTION_NAME` | Custom collection name (optional, overrides automatic naming) | Auto-generated |
+| `EMBEDDING_STRICT_COLLECTION_NAMES` | Use strict collection naming with provider+model info to prevent conflicts | `false` |
+
+> **💡 Collection Naming:**
+> - **Legacy mode** (default): Collections named `hybrid_code_chunks_<hash>` - same name for all providers
+> - **Strict mode** (`EMBEDDING_STRICT_COLLECTION_NAMES=true`): Collections include provider and model, e.g. `hybrid_ollama_nomic_embed_text_<hash>_<unique>`
+> - **Benefits of strict mode**: Prevents data conflicts when switching between different embedding providers or models
+> - **Use case**: Enable strict mode when experimenting with multiple providers (Ollama, LlamaCpp, etc.) on the same codebase
 
 ### Ollama (Optional)
 | Variable | Description | Default |
 |----------|-------------|---------|
 | `OLLAMA_HOST` | Ollama server URL | `http://127.0.0.1:11434` |
 | `OLLAMA_MODEL`(alternative to `EMBEDDING_MODEL`) | Model name |  |
 
+### LlamaCpp (Optional)
+| Variable | Description | Default |
+|----------|-------------|---------|
+| `LLAMACPP_HOST` | LlamaCpp server URL | `http://localhost:8080` |
+| `LLAMACPP_MODEL` (alternative to `EMBEDDING_MODEL`) | Model name |  |
+| `LLAMACPP_TIMEOUT` | Request timeout in milliseconds | `30000` |
+| `LLAMACPP_CODE_PREFIX` | Enable automatic code prefix for embeddings | `true` |
+
 
 ### Advanced Configuration
 | Variable | Description | Default |

packages/core/src/context.ts

@@ -94,6 +94,7 @@ export interface ContextConfig {
     ignorePatterns?: string[];
     customExtensions?: string[]; // New: custom extensions from MCP
     customIgnorePatterns?: string[]; // New: custom ignore patterns from MCP
+    customCollectionName?: string; // New: custom collection name from MCP config
 }
 
 export class Context {
@@ -103,6 +104,7 @@ export class Context {
     private supportedExtensions: string[];
     private ignorePatterns: string[];
     private synchronizers = new Map<string, FileSynchronizer>();
+    private customCollectionName?: string;
 
     constructor(config: ContextConfig = {}) {
         // Initialize services
@@ -145,6 +147,9 @@ export class Context {
         // Remove duplicates
         this.ignorePatterns = [...new Set(allIgnorePatterns)];
 
+        // Store custom collection name if provided
+        this.customCollectionName = config.customCollectionName;
+
         console.log(`[Context] 🔧 Initialized with ${this.supportedExtensions.length} supported extensions and ${this.ignorePatterns.length} ignore patterns`);
         if (envCustomExtensions.length > 0) {
             console.log(`[Context] 📎 Loaded ${envCustomExtensions.length} custom extensions from environment: ${envCustomExtensions.join(', ')}`);
@@ -229,14 +234,37 @@ export class Context {
     }
 
     /**
-     * Generate collection name based on codebase path and hybrid mode
+     * Generate collection name based on codebase path, provider, model and hybrid mode
      */
     public getCollectionName(codebasePath: string): string {
+        // If custom collection name is provided, use it directly
+        if (this.customCollectionName) {
+            return this.customCollectionName;
+        }
+
         const isHybrid = this.getIsHybrid();
         const normalizedPath = path.resolve(codebasePath);
-        const hash = crypto.createHash('md5').update(normalizedPath).digest('hex');
-        const prefix = isHybrid === true ? 'hybrid_code_chunks' : 'code_chunks';
-        return `${prefix}_${hash.substring(0, 8)}`;
+        const pathHash = crypto.createHash('md5').update(normalizedPath).digest('hex');
+
+        // Check if strict collection naming is enabled
+        const strictNaming = envManager.get('EMBEDDING_STRICT_COLLECTION_NAMES')?.toLowerCase() === 'true';
+
+        if (strictNaming) {
+            // Generate collection name including provider and model to prevent conflicts
+            const provider = this.embedding.getProvider().toLowerCase();
+            const model = this.embedding.getModel().replace(/[^a-zA-Z0-9]/g, '_'); // Sanitize model name
+
+            // Create a comprehensive hash including provider and model to ensure uniqueness
+            const uniqueString = `${provider}_${model}_${normalizedPath}`;
+            const fullHash = crypto.createHash('md5').update(uniqueString).digest('hex');
+
+            const prefix = isHybrid === true ? 'hybrid' : 'code';
+            return `${prefix}_${provider}_${model}_${pathHash.substring(0, 8)}_${fullHash.substring(0, 8)}`;
+        } else {
+            // Legacy collection naming (default behavior)
+            const prefix = isHybrid === true ? 'hybrid_code_chunks' : 'code_chunks';
+            return `${prefix}_${pathHash.substring(0, 8)}`;
+        }
     }
 
     /**

packages/core/src/embedding/base-embedding.ts

@@ -73,4 +73,10 @@ export abstract class Embedding {
      * @returns Provider name
      */
     abstract getProvider(): string;
+
+    /**
+     * Get model name/identifier
+     * @returns Model name
+     */
+    abstract getModel(): string;
 } 

packages/core/src/embedding/gemini-embedding.ts

@@ -119,6 +119,10 @@ export class GeminiEmbedding extends Embedding {
         return 'Gemini';
     }
 
+    getModel(): string {
+        return this.config.model || 'gemini-embedding-001';
+    }
+
     /**
      * Set model type
      * @param model Model name

packages/core/src/embedding/index.ts

@@ -5,4 +5,5 @@ export * from './base-embedding';
 export * from './openai-embedding';
 export * from './voyageai-embedding';
 export * from './ollama-embedding';
-export * from './gemini-embedding'; 
+export * from './gemini-embedding';
+export * from './llamacpp-embedding';