Bio-Intel Agent

Automated biomedical literature monitoring pipeline.

PubMed → LLM Summarization → Slack Alerts

---

What It Does

Biomedical research publishes over 3,000 papers per day. No human can keep up. Bio-Intel Agent automates the Search → Read → Summarize → Alert loop so researchers get actionable intelligence without reading every abstract.

Give it a keyword (e.g., "CRISPR", "mRNA vaccine", "intermittent fasting"). The pipeline queries PubMed for recent papers, summarizes each abstract into 3 bullet points using a local LLM or extractive algorithm, and delivers the results to Slack or console. Redis caching prevents redundant fetches. The entire system runs locally with zero paid API keys.

---

Architecture

POST /trigger-pipeline?keyword=CRISPR
  → FastAPI (returns 200 immediately)
  → Background task:
      ┌─────────────┐    ┌──────────────┐    ┌───────────┐
      │ Redis Cache  │───▶│  PubMed API  │───▶│ Summarize │
      │  (1hr TTL)   │    │  (Biopython) │    │ (LLM/ext) │
      └─────────────┘    └──────────────┘    └─────┬─────┘
                                                    ▼
                                             ┌───────────────┐
                                             │ Slack / stdout │
                                             └───────────────┘

Cache hit → skip PubMed + LLM, serve cached results directly.

---

Endpoints

Method	Path	Description
GET	/	Root status check
GET	/health	Full system health (mode, LLM backend, Redis, Slack, uptime)
POST	/trigger-pipeline?keyword=X	Run full pipeline in background
GET	/articles/search?keyword=X&limit=5	Search PubMed directly (no summarization)
POST	/pipeline/batch	Run pipeline for multiple keywords
GET	/pipeline/history	Last 10 pipeline runs
GET	/demo	Run full pipeline with "CRISPR" in mock mode — always works

Interactive API docs at /docs (Swagger UI).

---

Setup

Zero Config (demo mode)

git clone https://github.com/schoudhary90210/Bio-Intel-Agent.git
cd Bio-Intel-Agent
pip install -r requirements.txt
uvicorn app.main:app --reload

Then test:

curl http://localhost:8000/demo | python3 -m json.tool

With Docker

docker-compose up --build

This starts the API on port 8000 and Redis on port 6379.

With Local LLM (Ollama)

# Install ollama: https://ollama.com
ollama pull llama3.2
pip install ollama

MOCK_MODE=false LLM_BACKEND=ollama uvicorn app.main:app --reload

With Slack Alerts

Add your webhook URL to .env:

SLACK_WEBHOOK_URL=https://hooks.slack.com/services/T.../B.../xxx

Without a webhook URL, results print to console.

---

LLM Backends

Backend	Config	Requirements	Description
auto	LLM_BACKEND=auto	—	Tries ollama → extractive (default)
ollama	LLM_BACKEND=ollama	Local ollama install	Local LLM summarization via llama3.2
extractive	LLM_BACKEND=extractive	None	Sentence scoring algorithm, zero dependencies
mock	MOCK_MODE=true	None	Static responses for testing

The extractive backend scores sentences by biomedical term density, position weighting, and numeric data detection (p-values, percentages), then returns the top 3 as bullet points.

---

Configuration

Variable	Default	Description
MOCK_MODE	true	Bypass all real API calls
REDIS_URL	redis://localhost:6379/0	Redis connection string
PUBMED_EMAIL	[email protected]	Required by NCBI Entrez API
LLM_BACKEND	auto	LLM strategy: auto, ollama, extractive, mock
SLACK_WEBHOOK_URL	—	Slack incoming webhook (console fallback if empty)
OPENAI_API_KEY	—	Optional: OpenAI key for GPT-4 backend

Set via .env file or environment variables. Copy .env.example to get started.

---

Running Tests

pip install -r requirements.txt
pytest tests/ -v

Tests cover all services (PubMed, LLM, Slack, Redis) and all API endpoints. All tests run in mock mode with no external dependencies.

---

Project Structure

Bio-Intel-Agent/
├── app/
│   ├── main.py              # FastAPI routes + pipeline orchestration
│   ├── config.py             # Pydantic Settings configuration
│   ├── services/
│   │   ├── pubmed.py         # PubMed/Entrez abstract fetching
│   │   ├── llm.py            # LLM summarization (ollama/extractive/mock)
│   │   └── slack.py          # Slack Block Kit delivery + console fallback
│   └── utils/
│       └── redis_client.py   # Redis cache wrapper with graceful degradation
├── tests/
│   ├── test_api.py           # FastAPI endpoint tests
│   ├── test_pubmed.py        # PubMed service tests
│   ├── test_llm.py           # LLM summarization tests
│   ├── test_slack.py         # Slack delivery tests
│   └── test_redis.py         # Redis caching tests
├── requirements.txt
├── Dockerfile
├── docker-compose.yml
├── .env.example
└── README.md

---

Troubleshooting

Issue	Fix
ModuleNotFoundError	Run pip install -r requirements.txt
Redis connection refused	Start Redis: docker run -p 6379:6379 redis:alpine or skip — app works without it
Ollama not found	Install from ollama.com, then ollama pull llama3.2
PubMed returns 0 results	Try a broader keyword or check NCBI status. Mock mode returns sample data.
Slack not sending	Check SLACK_WEBHOOK_URL in .env. Without it, output goes to console.
Port 8000 in use	lsof -ti:8000 | xargs kill -9 then restart