Continuous Retrieval
Fetches newly published papers from arXiv using configurable search queries and polling intervals.
Automatic Retrieval and Summarizing Agent Team
Professional Research Operations for Reinforcement Learning
This project runs a multi-agent system that continuously discovers new RL papers, generates concise summaries, publishes markdown reports, and exposes real-time monitoring telemetry.
Structured Summaries
Transforms abstracts into compact bullet summaries designed for fast scanning and daily research updates.
Live Monitoring
Exposes dashboard, JSON API, and SSE stream so you can observe runtime state, events, and queue health.
Agent Workflow
Agent 01
Retriever
Collects latest candidate papers and stores canonical metadata in SQLite state tables.
Agent 02
Summarizer
Produces concise, structured summaries from abstract text with deterministic formatting.
Agent 03
Publisher
Writes markdown reports into generated/summaries/ and updates the summary index.
Run Commands
python -m agent_team run
python -m agent_team run-once --json
Monitoring Endpoints
- Dashboard:
http://127.0.0.1:8787/ - Status API:
/api/status - Event Stream:
/events(SSE)
Latest Generated Summaries
loading...Sub-topic Statistical Analysis
waiting for data...Distribution Diagram
0
papers