Reference

Key upstream libraries, protocols, and external service docs referenced by this project (grouped by capability area).

Workflow / Agent orchestration

• LangGraph: workflow graphs and node orchestration (the foundation of BaseGraph / MDGraph / HTMLGraph). See LangGraph Documentation.
• LangChain: messaging/tool abstractions and model interfaces (widely used via langchain_core and langchain.tools). See LangChain Documentation.

Configuration & validation

• Pydantic: schema definition and validation (tool inputs, plan validation, etc.). See Pydantic Documentation.

Web / networking

• aiohttp: async HTTP client used in fetch/image-related paths. See aiohttp Documentation.
• Requests: sync HTTP client used in parts of model calling / streaming. See Requests Documentation.

Retrieval / search / academic sources

• arXiv API: paper search and metadata. See arXiv API User Manual.
• Tavily Search API: optional search source (enabled via config). See Tavily Documentation.
• Jina AI Reader / Fetch: optional fetch/reader service (enabled via config). See Jina AI Reader.

Storage / state / queue

• Redis (redis-py): session params, task status, and conversation logs (best-effort / environment dependent). See redis-py Documentation.
• PostgreSQL: LangGraph checkpointer persistence for workflow checkpoint/resume. See PostgreSQL Documentation.
• Apache RocketMQ: message queue for event publishing in non-test environments. See RocketMQ Documentation.

Protocols

• Model Context Protocol (MCP): tool-based capability integration (e.g., filesystem) with health checking and client management. See Model Context Protocol.

Documentation site

• MkDocs: documentation site generator. See MkDocs Documentation.
• Material for MkDocs: theme and UI components used by this wiki. See Material for MkDocs.