Multi-Source Research Agent
A LangGraph agent that searches Google, Bing & Reddit in parallel and cites its sources.
The Multi-Source Research Agent answers a question by fanning out across several sources at once — Google and Bing (via Bright Data SERP) plus Reddit (via the free PullPush archive) — rather than trusting any single source.
It runs as a LangGraph state machine: fan-out search, then a Reddit drill-down where the LLM selects the most relevant threads via structured Pydantic output and pulls their comments, then per-source analysis, and finally a synthesis step that folds everything into one coherent, cited answer using Gemini 2.5 Flash.
Because each stage is an explicit node in the graph, the flow is easy to reason about and extend — adding a new source or a verification step is a matter of wiring in another node rather than rewriting a monolithic prompt. Third-party data is parsed, normalized, and filtered before it reaches the model context.