ReAct · Plan → Research × N → Synthesize · Real-time execution

Research Agent

A multi-step AI agent running live on this server. It plans sub-tasks, retrieves context via BM25 RAG, researches each task with an LLM, then synthesizes a final answer — all visible in real-time below.

Research Question
Try a sample question
Enter a question and click Run Agent to start.
🔗 Agent Trace
Agent steps will appear here…
✨ Synthesized Answer
Final answer will stream here once synthesis begins…
How this agent works
📝
Input
Your question
→
📋
Planner Node
LLM breaks query
into 2-3 sub-tasks
→
🔍
Researcher Node
BM25 retrieval +
LLM per sub-task
↩ conditional edge loops
until all tasks done
→
✍️
Synthesizer Node
Streams final answer
from all findings
→
✅
END
Complete response
in your browser

Source: app/services/agent_service.py · View full architecture →