agentic-research-workflow

Automating an AI research workflow involves using a FastAPI-based orchestration service that coordinates planning, researching, writing, and editing agents. This Skill integrates Tavily web search, arXiv papers, and Wikipedia summaries to execute multi-step research tasks automatically.

Multi-agent planning for research tasks works by passing a topic through a sequential pipeline: a planner agent structures the approach, a research agent gathers sources, a writer drafts the content, and an editor refines the final output. A PostgreSQL-backed state machine tracks each step.

Yes, you can deploy this AI research workflow using a single-container Docker-ready deployment. The workflow requires a PostgreSQL database to maintain persistent state management and track task progress via a REST API and simple web UI.

Orchestrating tool-calling agents requires a FastAPI service that routes tool requests to external APIs like Tavily for live web search, arXiv for scholarly papers, and Wikipedia for general knowledge, aggregating the results for the writing agent.

FastAPI is used to build the REST API that manages stateful task tracking for the research workflow. Combined with a PostgreSQL backend, it ensures that the multi-step process from planning to editing persists across sessions reliably.

Single-container deployment for multi-agent research workflows limits horizontal scaling of individual agents. While the PostgreSQL state machine handles task tracking, resource-intensive tool-calling across multiple simultaneous research tasks may create processing bottlenecks within one container.