A collection of AI/ML mini projects exploring various AI capabilities.
| Project | Description |
|---|---|
| agent-memory-router | LLM-powered router that classifies queries to RAG, graph, or tabular backends |
| ai-vani | AI voice assistant |
| annual-report-ai | Vision AI for reading and analyzing annual reports |
| chat-with-csv | Chat with CSV data |
| desibot | Desi-focused chatbot |
| github-mcp | GitHub MCP server implementation |
| graphrag-project | Graph-based RAG with knowledge graphs |
| hybrid-rag | Hybrid RAG combining vector search (ChromaDB) and graph traversal (NetworkX) |
| marketmind | Multi-agent market analysis system |
| multimodal-agent | Agent with image understanding capabilities |
| lightrag-project | LightRAG-based knowledge graph retrieval with dual-level retrieval |
| pageindex-rag | Reasoning-based RAG over PDFs using PageIndex tree navigation (no vectors) |
| sarvam-chatbot | Sarvam AI chatbot |
| self-healing-rag | RAG system with self-correction capabilities |
| telegram-bot | Telegram bot with memory engine |
| video-rag | RAG system for querying video content via frame extraction and embeddings |
| vision-rag | Vision-based RAG over PDFs using image embeddings for visual document understanding |
| voice-support-bot | Voice support bot with token server |
| visual-search | Visual product search using CLIP |
| langsmith-tutorial | LangSmith observability tutorial — tracing, datasets, and LLM-as-judge evaluations |
| rag-eval | RAG evaluation pipeline using RAGAS v0.4 — faithfulness, answer relevancy, context precision, and recall scored via Gemini 2.5 Flash |
| memory-agent | Persistent memory chat agent — stores and retrieves user facts across sessions using mem0, Groq, and local Qdrant |
| parallel-news-agent | LangGraph map-reduce pipeline that researches multiple news topics in parallel and assembles them into an executive briefing using Groq and Tavily |
| a2a-agents | Multi-agent system on the A2A protocol — Research and Writer agents discover each other via Agent Cards and chain over JSON-RPC, powered by Groq |
| computer-use-agent | Desktop control agent — Gemini Flash sees the screen and decides actions in a LangGraph See→Think→Act loop, executed with PyAutoGUI |
| self-healing-code-agent | Code agent that writes Python, runs it in a subprocess, reads its own tracebacks, and fixes its own bugs until self-written tests pass — built on Pydantic AI and Groq |
Each project has its own setup instructions. Generally:
# Python projects
cd <project-name>
pip install -e .
# Node.js projects
cd <project>/frontend
npm install- Python 3.10+
- Node.js 18+