self_rag_mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| USE_VLLM | No | true to use local vLLM backend, false to use OpenRouter for chat. | true |
| CHUNK_SIZE | No | Child chunk size (chars). | 600 |
| QDRANT_URL | No | URL of the Qdrant instance. | |
| VLLM_MODEL | No | Model served by vLLM. | qwen2.5-7b-instruct-awq |
| MAX_RETRIES | No | Max Self-RAG retry loops. | 3 |
| WEAVIATE_URL | No | URL of the Weaviate instance. | |
| CHUNK_OVERLAP | No | Child chunk overlap. | 150 |
| VLLM_BASE_URL | No | vLLM OpenAI-compatible endpoint. | http://localhost:8000/v1 |
| QDRANT_API_KEY | No | API key for Qdrant, if required. | |
| EMBEDDING_MODEL | No | Dense embedding model (1536-dim) — always via OpenRouter. | openai/text-embedding-3-small |
| LLM_TEMPERATURE | No | LLM temperature (0 = deterministic). | 0.0 |
| RETRIEVAL_K_MMR | No | Documents after MMR diversity filter. | 15 |
| OPENROUTER_MODEL | No | Chat model when routed through OpenRouter. | openai/gpt-4.1-mini |
| PINECONE_API_KEY | No | API key for Pinecone, required if VECTORDB_PROVIDER=pinecone. | |
| RETRIEVAL_LAMBDA | No | MMR relevance-vs-diversity balance (0-1). | 0.5 |
| PARENT_CHUNK_SIZE | No | Parent chunk size (chars). | 1200 |
| VECTORDB_PROVIDER | No | Vector DB provider: qdrant, pinecone, weaviate, chroma, or pgvector. | qdrant |
| CHROMA_PERSIST_DIR | No | Directory for Chroma persistence. | |
| OPENROUTER_API_KEY | Yes | Required for embeddings (via OpenRouter) and for chat when USE_VLLM=false. | |
| RETRIEVAL_K_RERANK | No | Final documents after FlashRank. | 4 |
| PINECONE_INDEX_NAME | No | Index name for Pinecone. | |
| RETRIEVAL_K_INITIAL | No | Hybrid search candidate pool size. | 20 |
| RETRIEVAL_COLLECTION | No | Collection name for retrieval. | self_rag_documents |
| VECTORDB_HYBRID_STRATEGY | No | Hybrid search strategy: rrf, weighted, semantic, two_pass. | rrf |
| PARENT_EXPANSION_COLLECTION | No | Collection for parent expansion. Set to 'null' to disable. | self_rag_parents |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Server capabilities have not been inspected yet.
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
No tools | |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
This server cannot be deployed
Maintenance
ActivityMaintained
ResponsivenessNo issues