zebir-sir/agentic-rag-paper-assistant

Agentic RAG Paper Assistant is an open-source workspace for research paper ingestion, evidence-traced QA and multi-source retrieval, intended for researchers or teams who need local/private paper analysis (summaries, method breakdowns, experiment interpretation, related-work and multi-paper comparison). It combines FastAPI + Streamlit frontend, PostgreSQL+pgvector, LangChain/LangGraph and OpenAI-compatible models. Key features: - Section-aware PDF ingestion (Markdown-like structure, line/section metadata) - Artifact extraction (tables/figures/algorithms) stored as searchable chunks - Source-aware Intent Planner to constrain tools and provenance - LangGraph multi-stage retrieval: plan → retrieve → evaluate → rewrite → generate - Evidence tracing with source, section, line and similarity scores - OpenAlex academic search, optional RabbitMQ async ingestion and Redis query-embedding cache - Streaming UI, long-session memory compression, and local-first design for private deployment

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.