Flagship project plan · Planned
Local AI Companion
A planned privacy-first AI companion designed to run locally, retain conversations, and retrieve information from private documents.
Proposed architecture: Next.js interface → FastAPI backend → LangGraph agent, with Ollama for local models and SQLite or PostgreSQL for persistence.
Planned first version
- Local chat and conversation history
- Persistent memory and semantic search
- Tool calling and retrieval over private files
- Tests, evaluation examples, and a Docker run setup
Implementation, demos, and measured results will be added as milestones are completed.