ProjectBuild an Agentic Content Studio: the explanation · Module 9: The research agent and web search · part 11 of 22
Part 11 · Module 9: The research agent and web search
Module 9: The research agent and web search
21 min read·
About 40 minutes · Needs: Module 8 (the LangGraph pipeline), Module 3 (the Ollama model client) · Code: backend/app/ai/agents/research.py, backend/app/ai/search.py, backend/app/ai/prompts.py (research_queries, research_facts)
Where this fits
The research agent (blue: this module) is the first node of the pipeline. It asks the model (amber) for queries and facts and calls Ollama's web search API (amber: outside our control).
Research is the first node of the generate graph. Before the strategist plans a single slide, this agent looks up the topic on the web, reads a few pages and hands the next agents a short list of facts, each tied to the page it came from. It is optional by design: when it is off or fails, the deck is still made, and the deck says honestly that its claims are unverified.
The rest of this project is yours to keep
This project is bought on its own, once, and stays readable afterwards, including the parts added to it later.