agentic-research-agent-fastapi

Automate multi-step research workflows using a FastAPI service that decomposes prompts into subtasks, executes them with external tools, and iteratively refines results through a reflection loop backed by PostgreSQL task tracking.

Structure research tasks using a planning agent that decomposes complex prompts into structured subtasks, enabling an executor agent to reliably run research and gather results with tools like Tavily, arXiv, and Wikipedia.

Yes, Docker is required for reproducible deployment runs, and an OpenAI-compatible LLM is required to power the agentic reasoning. The system operates as a FastAPI service with PostgreSQL-backed task tracking.

Yes, the FastAPI research agent integrates with arXiv and Wikipedia, alongside Tavily. The executor agent uses these external tools to gather results, while PostgreSQL tracks the workflow status and outputs.

A reflection loop improves research outcomes by iteratively evaluating and refining gathered findings after the executor agent completes its tasks. This cycle ensures the multi-step workflow corrects errors and enhances quality across multiple iterations.

Monitor research task progress through a lightweight web UI and PostgreSQL-backed task tracking. This setup allows you to trigger workflows and observe the planning, execution, and reflection stages in real time.