RAG

AI agent that crawls GST, SEBI, RBI, MCA & IRDAI — analyzes circulars with Claude Sonnet, tracks deadlines, and alerts teams via WhatsApp/Email/Telegram. Day 5 of 27-day AI agent build challenge.

Retrieval-Augmented Generation (RAG) chatbot for PDF documents.

Fully local, GPU-powered hybrid memory search server for LLM agents — FAISS+BM25 RRF fusion, local cross-encoder reranking, MMR, time-decay. OpenAI-compatible API, no external API.

Local PostgreSQL Internals Assistant with Ollama, pgvector, and MCP

Agent-first unofficial NotebookLM client for Node.js — JSON CLI, citations, multi-turn chat, all studio artifacts, research & notes

A persistent memory function with customisable knowledge and in chart curating and retrieval.

Du-RAG: Durable RAG — persistent memory for AI agents built on retrieval-augmented generation

A website intelligence platform that converts a live website into a structured, queryable, inspectable knowledge system. It crawls, analyzes, and models a site as pages, forms, endpoints, auth signals, workflows, and retrieval-ready evidence.

Standalone Go CLI for answering questions with AI against Path of Building build.

Local-first legal document assistant for legal engineers — upload PDFs, search with BM25/CARP, and chat with citation-grade answers and bbox-grounded highlights. Next.js + FastAPI + SQLite.

SlimContext is a deterministic post-retrieval optimizer for RAG — dedupe, cluster, MMR, and token budgets. No LLM calls.

Open-source AI agent on Amazon Bedrock that turns regulator queries into audit-grade evidence packs — built by CreativeMinds Development (cmdev). NDPA · CBN CSAT · NIS2 · NIST.

A self-correcting AI memory engine with built-in memory quality governance, hybrid retrieval and multimodal support for text, images and emails. Not just storage & retrieval: An AI memory system that detects stale, conflicting and redundant content before feeding data to LLMs.

Houndex is a last-mile knowledge framework for building source-backed AI systems. This is the PY one.