How mgalen007/atlas works
Atlas is a FastAPI service that accepts a research topic and returns a structured summary. Its pipeline gathers findings, validates them, then summarizes the validated findings; the service orchestrates each stage. The README also documents a health-check endpoint. Source excerpts support the API routing, orchestration and agent/model configuration shown here; the Gemini provider’s internal behavior is not sampled.
Main entry points: service.py Pipeline service · research_agent.py Research agent · critic_agent.py Critic agent · summarizer_agent.py Summarizer agent
Components 9
API
- FastAPI app
server/main.py - Research endpoint
server/features/research/router.py
Research pipeline
- Pipeline service
server/features/agents/service.py - Research agent
server/features/agents/research_agent.py - Critic agent
server/features/agents/critic_agent.py - Summarizer agent
server/features/agents/summarizer_agent.py
Agents and model
- Gemini model config
server/features/agents/config/models.py
Other components
- API caller
- Google Gemini
Connections 12
- API caller → FastAPI app: HTTP requests
- FastAPI app → Research endpoint: mounts router
- Research endpoint → Pipeline service: runs pipeline
- Pipeline service → Research agent: invokes research
- Pipeline service → Critic agent: invokes validation
- Pipeline service → Summarizer agent: invokes summary
- Research agent → Gemini model config: uses model
- Critic agent → Gemini model config: uses model
- Summarizer agent → Gemini model config: uses model
- Research agent → Google Gemini: generates findings
- Critic agent → Google Gemini: validates findings
- Summarizer agent → Google Gemini: generates summary