vector-rag

A production-ready Hybrid RAG system with ChromaDB, BM25, Multi-Query Retrieval, MMR, CrossEncoder reranking, and Ollama-powered LLM responses.

Enterprise-grade Vector RAG platform engineered for semantic knowledge retrieval, AI-powered document intelligence, vector search infrastructure, and contextual conversational AI workflows.

Enterprise Autonomous Multi-Agent Orchestration & Real-Time Vector RAG Intelligence Platform | Python • FastAPI • Next.js 14 • WebSockets • Vector Store • Docker

Vector RAG chatbot for YouTube videos using LlamaIndex. Fetches the transcript with a self-written loader, embeds it with Gemini into ChromaDB Cloud, reranks with a cross-encoder, and answers from the video in a CLI chat with PostgreSQL memory. Includes Phoenix tracing, token counting, and cost tracking. Vector baseline of a RAG series.

Vector RAG vs. Vectorless RAG vs. OKF

⚡ Jinx — 100% Offline Sovereign AI Agent for Android. Local WebGPU LLM inference, NotebookLM-grade vector RAG, Pyodide Python sandbox and blackout survival triage architecture.

Where does Graph RAG beat Vector RAG? A stratified evaluation on financial contract Q&A (MSc dissertation, DBS)

Graph-based multi-hop document retrieval via citation network search — a lightweight alternative between Vector RAG and GraphRAG, evaluated on OGBN-Arxiv with 12 seeding strategies and 6 traversal algorithms.

The structural memory integrity layer for AI agents. Built with Cognee 1.0. Coherence extends the memify() pipeline to detect, flag, and surgically prune temporal supersessions and semantic contradictions from agent memory networks. This ensures agents never believe two conflicting things at once.