Summary

Argus is an enterprise-grade RAG knowledge platform designed to integrate private documents with large language models. It addresses LLM challenges like hallucination, knowledge fragmentation, and lack of conversational memory, enabling traceable and factual AI-powered Q&A for technical users.

How It Works

The platform implements a full RAG pipeline, from document ingestion and intelligent parsing to hybrid retrieval and AI-driven generation. It employs a novel approach combining semantic search (PostgreSQL/pgvector HNSW) with keyword search (Elasticsearch IK/BM25), fused via Reciprocal Rank Fusion (RRF). An AI Agent layer, built on Spring AI Alibaba ReactAgent, orchestrates queries, supports dual CHAT/KB_SEARCH modes, and manages short-term memory through progressive compression.

Quick Start & Requirements

• Primary Install/Run: Backend: ./mvnw spring-boot:run (Java 21). Frontend: npm run dev (Node.js >= 20.19).
• Prerequisites: JDK 21, Node.js >= 20.19, PostgreSQL 16+ (with pgvector extension), Elasticsearch 8.x (with IK Chinese analyzer plugin), MinIO (optional), DashScope API Key.
• Setup: Requires configuring PostgreSQL, Elasticsearch, and MinIO instances, then updating application-local.yml with credentials and endpoints. Docker commands are provided for middleware setup.
• Links: API Documentation: http://localhost:10001/doc.html.

Highlighted Details

• RAG Pipeline: Comprehensive end-to-end RAG flow featuring self-developed query planning, RRF fusion, and a four-level evidence sufficiency evaluation mechanism.
• AI Agent Capabilities: Dynamic tool orchestration via ReactAgent, seamless switching between CHAT and KB_SEARCH modes within a session, and three-level short-term memory compression for extended context awareness.
• Hybrid Retrieval: Parallel semantic (PGvector HNSW, COSINE_DISTANCE) and keyword (Elasticsearch BM25, IK analyzer) search channels, unified by RRF for robust result ranking.
• Enterprise Security: Robust security features including role-based access control (Admin/Owner/Member), JWT dual tokens, BCrypt password hashing, group data isolation, and comprehensive operation logging.

Maintenance & Community

The project is developed iteratively by the "Argus team," with a clear version evolution (V1.0-V4.0) indicating ongoing feature development. No specific community channels (e.g., Discord, Slack) or notable contributors/sponsorships are listed in the README.

Licensing & Compatibility

Argus is released under the MIT License, which permits broad usage, including commercial applications and integration into closed-source projects.

Limitations & Caveats

Setup involves deploying and configuring multiple complex middleware components (PostgreSQL, Elasticsearch, MinIO). The system relies on external LLM and embedding services (DashScope), necessitating API key management. While described as enterprise-grade, the rapid versioning suggests active development, and specific performance benchmarks or detailed bug tracking are not presented.