Argus
An enterprise knowledge-base platform that combines retrieval-augmented generation with an AI agent, answering questions from uploaded documents with citations back to their sources.
Argus is an enterprise knowledge-base platform built around retrieval-augmented generation and an AI agent, meant to keep every answer traceable to a source document.
Uploaded documents are parsed, chunked, embedded and indexed; a query planner rewrites or decomposes user questions, then a hybrid retriever combines vector search and keyword search, fuses the results by reciprocal rank fusion, and grades the evidence across four sufficiency levels before generating an answer with citations. An agent built on a graph execution engine switches between plain chat and knowledge-base search within one conversation, compressing conversation memory in three stages to fit a limited context window.
It has role-based permissions, group data isolation, access and refresh tokens, and audit logging. The backend is Java with Spring Boot and PostgreSQL/pgvector; the AI layer uses Spring AI Alibaba with a ReactAgent graph engine; storage and search use MinIO and Elasticsearch; the frontend is Vue 3 and TypeScript.