Letta-Deepseek

About

The project demonstrates wiring Letta's stateful-memory framework to DeepSeek models for a division-of-labor coding pipeline: a research agent with Tavily web search, a DeepSeek-powered coding agent, a documentation agent, and an orchestrator coordinating them. Memory features emphasize optimization, pattern recognition, and knowledge accumulation across runs, with deployment documented for Lightning AI. At nine commits, eleven stars, and no issues or releases, it is a personal reference implementation rather than a supported product. Developers use it as a worked example of wiring Letta's memory framework to DeepSeek and Tavily rather than as a daily-driver tool.