Agentic Search
A team of AI agents that researches a question and checks its own work. Plan, search, scrape, report, review, loop.
- Python
- LangGraph
- LangChain
- Groq
- Serper
- Streamlit
- Turns one research question into a full pipeline: a Planner picks the query, Google Serper fetches results, a Selector chooses the best source, and a scraper reads the page.
- A Reporter writes a cited answer, then a Reviewer judges its comprehensiveness, citations and relevance.
- A Router acts on the review and sends the work back to the Planner, Selector or Reporter, or finishes, so weak answers get another pass.
- Agents exchange JSON-schema outputs (Groq JSON mode) so routing never depends on parsing free text.
- A Streamlit chat interface shows each agent's output live as the graph runs.