Odysseus – self-hosted AI workspace

odysseus is a pragmatic take on the hosted LLM workspace problem. instead of another wrapper around openai/claude, it’s a full stack: agent orchestration, document editing with ai assistance (not generation-first), memory/skills persistence, and integrations for email/calendar/tasks. the architecture is solid—chromadb for embeddings, support for multiple inference backends (vllm, llama.cpp, ollama), and cookbook for model management. tradeoffs are real: apple silicon gpu support requires native install (docker runs linux vm), and the ui is feature-dense rather than minimal. worth evaluating if you’re tired of vendor lock-in and have the hardware budget. the cookbook model-fitting heuristic (vram-aware selection) is a nice touch for reducing friction on first run.