ScholarAgent RAG MCP Server
by NacyZh
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| PAPERS_DIR | No | Path to the directory containing PDF papers. | path/to/papers |
| ENABLE_HYDE | No | Enable HyDE query expansion. | false |
| LLM_API_KEY | No | API key for HyDE LLM. | |
| HYDE_BACKEND | No | Backend for HyDE: 'api' or 'vllm'. | api |
| LLM_BASE_URL | No | Base URL for HyDE LLM API. | |
| LLM_MODEL_NAME | No | Model name for HyDE. | |
| BGE_OFFLINE_MODE | No | Enable offline mode for BGE models; no downloads. | false |
| HYDE_API_TIMEOUT | No | Timeout in seconds for HyDE API calls. | 60 |
| BGE_AUTO_DOWNLOAD | No | Automatically download missing BGE models from Hugging Face. | true |
| BGE_M3_MODEL_PATH | No | Path to the BGE-M3 embedding model. | ./rag/models/bge-m3 |
| RAG_SUMMARY_TOP_P | No | Top-p for summarization. | 0.8 |
| PAPER_PARSER_DEVICE | No | Device for PDF parsing (e.g., cpu, cuda). | cuda |
| RAG_SUMMARY_API_KEY | No | API key for the summary API. | |
| RAG_SUMMARY_BACKEND | No | Backend for paper summaries: 'api' or 'vllm'. | api |
| HYDE_VLLM_MODEL_PATH | No | Path to local vLLM model for HyDE. | |
| TOOL_MAX_QUERY_CHARS | No | Maximum characters for tool queries. | |
| BGE_M3_MODEL_REVISION | No | Revision of the BGE-M3 model. | main |
| RAG_SUMMARY_API_MODEL | No | Model name used in API requests. | Qwen/Qwen3-8B-AWQ |
| TOOL_MAX_RETURN_CHARS | No | Maximum characters for tool return text. | |
| RAG_SUMMARY_MODEL_NAME | No | Name of the summary model. | Qwen/Qwen3-8B-AWQ |
| RAG_SUMMARY_MODEL_PATH | No | Path to the local summary model directory. | ./rag/models/Qwen3-8B-AWQ |
| RAG_SUMMARY_MODEL_REPO | No | Hugging Face repository for the summary model. | Qwen/Qwen3-8B-AWQ |
| BGE_RERANKER_MODEL_PATH | No | Path to the BGE reranker model. | ./rag/models/bge-reranker-v2-m3 |
| RAG_SUMMARY_CONCURRENCY | No | Concurrency for background summary jobs. | 1 |
| RAG_SUMMARY_TEMPERATURE | No | Temperature for summarization. | 0.2 |
| RAG_SUMMARY_API_BASE_URL | No | Base URL for the OpenAI-compatible API for summaries. | http://127.0.0.1:8001/v1 |
| RAG_SUMMARY_OFFLINE_MODE | No | Enable offline mode for the summary model. | false |
| RAG_SUMMARY_AUTO_DOWNLOAD | No | Automatically download the summary model if missing. | true |
| RAG_SUMMARY_ENFORCE_EAGER | No | Enforce eager execution for vLLM. | true |
| RAG_SUMMARY_CPU_OFFLOAD_GB | No | CPU offload in GB for vLLM. | 0 |
| RAG_SUMMARY_MODEL_REVISION | No | Revision of the summary model. | main |
| RETRIEVAL_EVAL_RESULTS_DIR | No | Directory for retrieval evaluation results. | ./rag/evaluation/results |
| BGE_RERANKER_MODEL_REVISION | No | Revision of the BGE reranker model. | main |
| RAG_SUMMARY_API_MAX_RETRIES | No | Maximum retries for API calls. | 2 |
| RETRIEVAL_EVAL_DATASET_PATH | No | Path to retrieval evaluation dataset. | ./rag/evaluation/data/retrieval_eval.jsonl |
| RAG_SUMMARY_MAX_INPUT_TOKENS | No | Maximum input tokens for summarization. | 4096 |
| RAG_SUMMARY_MAX_OUTPUT_TOKENS | No | Maximum output tokens for summarization. | 2048 |
| RAG_SUMMARY_API_TIMEOUT_SECONDS | No | Timeout in seconds for API calls. | 180 |
| TOOL_JOB_STATUS_WAIT_MAX_SECONDS | No | Maximum wait seconds for job status polling. | |
| TOOL_JOB_STATUS_WAIT_STEP_SECONDS | No | Step seconds for job status polling backoff. | |
| RAG_SUMMARY_GPU_MEMORY_UTILIZATION | No | GPU memory utilization fraction for vLLM. | 0.75 |
| TOOL_JOB_STATUS_WAIT_INITIAL_SECONDS | No | Initial wait seconds for job status polling. |
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 | |
Latest Blog Posts
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/NacyZh/MCP_Server_for_papers'
If you have feedback or need assistance with the MCP directory API, please join our Discord server