ThoughtDAG

A desktop canvas organizes conversations into connectable nodes, with a local session index, topic materials, and CLI search. An optional Jev layer assists with topic identification and ranking past excerpts; a user-configured language model generates answers. Data is stored locally, but remote model calls transmit relevant content. Rule-based retrieval remains available when Jev is disabled. It can connect to local Ollama. Jev's cloud service requires its API key; a self-hosted decision service can operate without one. Source code is available at no charge; hosted features, online models, and third-party services may have separate fees. Sources: https://github.com/chenxiachan/thoughtdag/blob/main/README.md https://www.jev.store/projects/thoughtdag Only source code or documentation was reviewed. The application was not installed or run, and its effectiveness was not verified. Project demonstration video: