CLARA MCP Server
A hybrid pulmonary radiology diagnostic backend that provides MCP agent skills for vision inference, clinical RAG, report synthesis, and escalation, with multi-layer security.
README
CLARA — MCP Server (backend)
Backend pentru CLARA, sistem hibrid de diagnosticare radiologică pulmonară. Expune un model ViT-B/16 fine-tuned pe CheXpert (14 patologii), explicabilitate LRP-ViT, segmentare promptabilă LRP-to-SAM2, RAG clinic, generare de raport și 4 agent skills MCP, totul cu securitate agentică în 4 straturi.
⚠️ Instrument de augmentare, nu de înlocuire a radiologului. Output asistat de AI, de validat de medic.
Stack
| Componentă | Tehnologie |
|---|---|
| API + MCP | FastAPI + fastmcp (REST și MCP coexistente) |
| Pipeline asincron | Celery + Redis (progres prin SSE) |
| Bază de date | PostgreSQL + pgvector (audit log + index RAG) |
| CV | PyTorch 2.3 + transformers + peft |
| Orchestrare | Docker Compose |
| Frontend | Next.js (separat: ../clara-radiology-dashboard) |
Structura proiectului
clara-mcp-server/
├── app/
│ ├── main.py # FastAPI: montează REST + MCP
│ ├── config.py # setări (pydantic-settings)
│ ├── api/ # POST /analyze, GET /stream/{job_id}, scheme
│ ├── cv_engine/ # model, preprocess, LRP-ViT, LRP-to-SAM2, inference
│ ├── rag/ # embedder, knowledge base, retriever, TAM
│ ├── mcp/ # server fastmcp + cele 4 skills
│ ├── security/ # sanitize, prompt guard, guardrails, audit log
│ ├── tasks/ # Celery app + pipeline asincron
│ ├── db/ # SQLAlchemy engine + modele ORM
│ └── llm/ # generare raport radiologic
├── weights/ # vit_lora_chexpert.pt (montat ca volum)
├── scripts/init_db.sql # pgvector + tabele
├── demo/ # imagine demo pentru apărare
├── tests/test_e2e.py
├── Dockerfile
├── docker-compose.yml
└── requirements.txt
Pornire rapidă (docker compose up)
cp .env.example .env # completează OPENAI_API_KEY (opțional)
# pune modelul în weights/vit_lora_chexpert.pt (deja copiat dacă ai folosit scriptul)
docker compose up --build
Servicii disponibile după pornire:
- API + docs OpenAPI: http://localhost:8000/docs
- Endpoint MCP: http://localhost:8000/mcp
- Postgres:
localhost:5432, Redis:localhost:6379
API REST (rezumat)
| Metodă | Rută | Descriere |
|---|---|---|
POST |
/analyze |
Trimite o radiografie, creează un job asincron |
GET |
/stream/{job_id} |
Progres în timp real (SSE) + rezultat final |
GET |
/health |
Status model + dependențe |
Cele 4 agent skills (MCP)
- ExecuteVisionInference — clasificare ViT + LRP + LRP-to-SAM2
- QueryClinicalKnowledge — RAG ierarhic pe baza de cunoștințe clinică
- SynthesizeMedicalReport — raport radiologic via LLM
- EscalateToHumanExpert — decizie conservatoare de escalare
Securitate agentică (4 straturi)
- Sanitizare PII la ingestie (Presidio / regex fallback)
- Izolarea contextului în prompt (delimitare XML, context = date, nu comenzi)
- Validarea output-ului (Guardrails / heuristici anti-halucinație și anti-injecție)
- Audit log imutabil cu lanț de hash-uri (Postgres)
Note pentru demonstrația live
- SAM2 are fallback clasic (GrabCut) dacă pachetul Meta nu e instalat.
- BiomedCLIP are fallback la
sentence-transformersdacă nu se încarcă. - Fără
OPENAI_API_KEY, raportul revine la promptul structurat TAM (rulează oricum).
Status implementare: Pas 6/7 — backend complet (CV + REST/SSE + Celery + MCP + RAG + securitate) + integrare frontend (api.ts, SSE, /audit/recent, lrp_map_b64 + report în pipeline). Urmează test e2e + screenshot-uri.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.