|
|
--- |
|
|
title: ADMP-LS |
|
|
emoji: 🤖 |
|
|
colorFrom: blue |
|
|
colorTo: purple |
|
|
sdk: docker |
|
|
app_port: 7860 |
|
|
--- |
|
|
|
|
|
# ADMP-LS |
|
|
|
|
|
A multi-service MCP (Model Context Protocol) application containing independent MCP servers and a Streamlit client. |
|
|
|
|
|
## Description & Citation |
|
|
|
|
|
ADMP-LS is an agent-based platform for life sciences that unifies literature review, evidence-grounded QA, and parameter extraction with transparent provenance. |
|
|
|
|
|
|
|
|
 |
|
|
|
|
|
 |
|
|
|
|
|
 |
|
|
 |
|
|
## Service Architecture |
|
|
|
|
|
- **Streamlit Client** (Port 7860, Python 3.11): Main user interface |
|
|
- **Retrieve Service** (Port 9487, Python 3.11): Biomedical RAG service |
|
|
- **Review Service** (Port 8880, Python 3.11): Biomedical Review service |
|
|
|
|
|
Related docs: |
|
|
- Client details: `client/README.md` |
|
|
- Retrieve server: `servers/Retrieve/readme.md` |
|
|
- Review server (EN): `servers/Review/readme.md` |
|
|
|
|
|
## Technical Features |
|
|
|
|
|
- ✅ Multi-stage Docker build |
|
|
- ✅ Multi-Python version support (3.11 + 3.12) |
|
|
- ✅ Virtual environment isolation |
|
|
- ✅ HF Spaces compliant |
|
|
- ✅ GPU support (optional) |
|
|
|
|
|
## Deployment |
|
|
|
|
|
This Space uses Docker deployment, with all services running in the same container but using independent Python virtual environments to avoid dependency conflicts. |
|
|
|
|
|
## Environment Variables |
|
|
|
|
|
You can set the following environment variables through HF Spaces: |
|
|
|
|
|
### Basic Configuration |
|
|
- `PORT`: Streamlit client port (default 7860) |
|
|
- `RETRIEVE_PORT`: Retrieve service port (default 9487) |
|
|
- `REVIEW_PORT`: Review service port (default 8880) |
|
|
|
|
|
### Retrieve Service LLM Configuration |
|
|
- `QA_LLM_MAIN_API_KEY`: QA main model API key |
|
|
- `QA_LLM_MAIN_BASE_URL`: QA main model base URL |
|
|
- `QA_LLM_BACKUP_API_KEY`: QA backup model API key |
|
|
- `QA_LLM_BACKUP_BASE_URL`: QA backup model base URL |
|
|
- `REWRITE_LLM_MAIN_API_KEY`: Rewrite main model API key |
|
|
- `REWRITE_LLM_MAIN_BASE_URL`: Rewrite main model base URL |
|
|
- `REWRITE_LLM_BACKUP_API_KEY`: Rewrite backup model API key |
|
|
- `REWRITE_LLM_BACKUP_BASE_URL`: Rewrite backup model base URL |
|
|
|
|
|
### Retrieve Service Web Search Configuration |
|
|
- `SERPER_API_KEY`: Serper API key (for web search) |
|
|
|
|
|
### Review Service Configuration |
|
|
- `OPENAI_BASE_URL`: OpenAI API base URL |
|
|
- `OPENAI_API_KEY`: OpenAI API key |
|
|
- `QIANWEN_BASE_URL`: Qianwen API base URL |
|
|
- `QIANWEN_API_KEY`: Qianwen API key |
|
|
- `SEARCH_URL`: Search service URL |
|
|
- `LOG_DIR`: Log directory |
|
|
- `LOG_LEVEL`: Log level (DEBUG, INFO, WARNING, ERROR, CRITICAL) |
|
|
- `LOG_MAX_SIZE`: Log file maximum size (bytes) |
|
|
- `LOG_BACKUP_COUNT`: Log backup file count |
|
|
- `LOG_ENABLE_CONSOLE`: Enable console logging (true/false) |
|
|
- `LOG_ENABLE_FILE`: Enable file logging (true/false) |
|
|
- `DEBUG_MODE`: Debug mode (true/false) |
|
|
|
|
|
### Client Configuration (providers) |
|
|
- `OPENAI_API_KEY`, `OPENAI_BASE_URL` |
|
|
- `ANTHROPIC_API_KEY`, `ANTHROPIC_BASE_URL` |
|
|
- `GOOGLE_API_KEY`, `GOOGLE_BASE_URL` |
|
|
- `AWS_REGION`, `AWS_ACCESS_KEY_ID`, `AWS_SECRET_ACCESS_KEY` |
|
|
- `GROQ_API_KEY`, `GROQ_BASE_URL` |
|
|
|
|
|
Note: place a `.env` file under `client/` (see `client/.env-example.txt`). The client also reads `client/servers_config.json` for MCP/server integration. |
|
|
|
|
|
## Ports |
|
|
- Space app port (client UI): `PORT` (default 7860) |
|
|
- Internal Retrieve service: 9487 (HTTP APIs: `/retrieve`, `/stream-chat`) |
|
|
- Internal Review service: 8880 (HTTP APIs: `/health`, `/review_generate`; MCP mounts: `/review`, `/check`) |
|
|
|
|
|
## Quick Links |
|
|
- Start client locally: see `client/README.md` |
|
|
- Start Retrieve locally or with Docker: see `servers/Retrieve/readme.md` |
|
|
- Start Review locally or with Docker: see `servers/Review/readme.md` |
|
|
|
|
|
## 🔗 Links |
|
|
- **GitHub**: [https://github.com/JackKuo666/ADMP-LS](https://github.com/JackKuo666/ADMP-LS) |
|
|
- **Hugging Face Spaces**: [https://huggingface.co/spaces/jackkuo/ADMP-LS](https://huggingface.co/spaces/jackkuo/ADMP-LS) |
|
|
- **Demo Video**: [https://youtu.be/wabS5Q6TMBk](https://youtu.be/wabS5Q6TMBk) |
|
|
|
|
|
|
|
|
## 🙏 Acknowledgements |
|
|
|
|
|
- [mcp-playground](https://github.com/Elkhn/mcp-playground) |
|
|
- [LangChain](https://github.com/langchain-ai/langchain) |
|
|
- [LangGraph](https://github.com/langchain-ai/langgraph) |
|
|
- [FastMCP](https://github.com/jlowin/fastmcp) |
|
|
- [Streamlit](https://github.com/streamlit/streamlit) |
|
|
|
|
|
## Citation |
|
|
Citation: Guo M., Sun Z., Xie S., Hu J., Sun S., Li X., Feng L., Jiang J. "ADMP-LS: Agent-based Dialogue and Mining Platform for Evidence-Grounded QA, Extraction, and Literature Review in Life Science," Zhejiang Lab & Western Carolina University. |
|
|
|
|
|
|
|
|
|
