Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| BM25_B | No | BM25 length normalization parameter. Example 0.75 | |
| BM25_K1 | No | BM25 saturation parameter. Example 1.2 | |
| LIBSQL_URL | No | Database URL (local file: file:./path/to/db.sqlite, remote libSQL: libsql://your-db.turso.io) | file:./libsql.db |
| BM25_ENABLE | No | Set to false or 0 to disable BM25 ordering | true |
| OLLAMA_HOST | No | Ollama host. Example http://localhost:11434 | |
| HYBRID_RRF_K | No | RRF K parameter for hybrid search | 60 |
| METRICS_PORT | No | Metrics HTTP port exposing /metrics and /healthz | 9090 |
| PROJECTS_DIR | No | Base directory for multi-project mode (can also be set via flag -projects-dir) | |
| HYBRID_SEARCH | No | Set to true/1 to enable hybrid search | |
| EMBEDDING_DIMS | No | Embedding dimension for new databases. Existing DBs are auto-detected and take precedence at runtime. | 4 |
| GOOGLE_API_KEY | No | Google API key for Gemini | |
| OPENAI_API_KEY | No | OpenAI API key | |
| VOYAGE_API_KEY | No | VoyageAI API key (alternative) | |
| LOCALAI_API_KEY | No | Optional LocalAI API key | |
| LOCALAI_BASE_URL | No | LocalAI base URL (OpenAI-compatible) | http://localhost:8080/v1 |
| VOYAGEAI_API_KEY | No | VoyageAI API key | |
| DB_MAX_IDLE_CONNS | No | Max idle DB connections (optional) | |
| DB_MAX_OPEN_CONNS | No | Max open DB connections (optional) | |
| LIBSQL_AUTH_TOKEN | No | Authentication token for remote databases | |
| HYBRID_TEXT_WEIGHT | No | Text weight for hybrid search | 0.4 |
| METRICS_PROMETHEUS | No | If set (e.g., true), expose Prometheus metrics | |
| EMBEDDINGS_PROVIDER | No | Optional embeddings source. Supported values: openai, ollama, gemini|google|google-gemini|google_genai, vertexai|vertex|google-vertex, localai|llamacpp|llama.cpp, voyageai|voyage|voyage-ai | |
| OLLAMA_HTTP_TIMEOUT | No | Ollama HTTP timeout to allow cold model load for larger models. Example 60s | |
| VERTEX_ACCESS_TOKEN | No | Vertex AI Bearer token | |
| DB_CONN_MAX_IDLE_SEC | No | Connection max idle time in seconds (optional) | |
| HYBRID_VECTOR_WEIGHT | No | Vector weight for hybrid search | 0.6 |
| EMBEDDINGS_ADAPT_MODE | No | How to adapt provider vectors to the DB size: pad_or_truncate | pad | truncate | pad_or_truncate |
| GEMINI_EMBEDDINGS_MODEL | No | Gemini embeddings model (default text-embedding-004, dims 768) | text-embedding-004 |
| OLLAMA_EMBEDDINGS_MODEL | No | Ollama embeddings model (default nomic-embed-text, dims 768) | nomic-embed-text |
| OPENAI_EMBEDDINGS_MODEL | No | OpenAI embeddings model | text-embedding-3-small |
| DB_CONN_MAX_LIFETIME_SEC | No | Connection max lifetime in seconds (optional) | |
| LOCALAI_EMBEDDINGS_MODEL | No | LocalAI embeddings model (default text-embedding-ada-002, dims 1536) | text-embedding-ada-002 |
| VOYAGEAI_EMBEDDINGS_DIMS | No | Optional VoyageAI embeddings dimensions to explicitly set expected output length | |
| VOYAGEAI_EMBEDDINGS_MODEL | No | VoyageAI embeddings model | voyage-3-lite |
| VERTEX_EMBEDDINGS_ENDPOINT | No | Vertex AI embeddings endpoint. Format: https://{location}-aiplatform.googleapis.com/v1/projects/{project}/locations/{location}/publishers/google/models/{model}:predict | |
| MULTI_PROJECT_AUTH_REQUIRED | No | Set to false/0 to disable per-project auth enforcement | true |
| MULTI_PROJECT_DEFAULT_TOKEN | No | Optional token value used when auto-initializing; if omitted, a random token is generated | |
| MULTI_PROJECT_AUTO_INIT_TOKEN | No | Set to true/1 to auto-create a token file on first access when none exists | false |
Schema
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
No tools | |