zrx2231050-lab/ai-project-agent

AI Project Agent is a runnable prototype for AI-driven personal knowledge management and project execution. It ingests scattered project artifacts (docs, markdown, code notes, meeting records, risks, todos) and builds a structured project state using multi-agent reasoning and RAG-backed citations. Intended users: individual contributors, product managers and small teams who want local-first project memory, structured tracking, and reproducible QA without mandatory cloud tokens. Key features: - Four agents: DocumentParser, Planning, Execution, Review for extraction, prioritization, drafting and verification. - Structured memory: facts, tasks, risks, decisions, briefs, run history. - RAG question answering with source titles, chunk headings, line ranges, scores and confidence; explicit insufficient-evidence handling. - Import: pasted text, Markdown, .docx files. - Local-first operation with Xiaomi-compatible model layer and graceful model-fallback behavior. - APIs for ingest, ask, state, brief; quick start via npm; focused smoke tests. MIT license.