AI-powered startup intelligence agent — hand-built ReAct loop, parallel tool execution, live web search, and RAG-grounded document analysis. No frameworks. Full understanding.
-
Updated
Aug 28, 2026 - Python
AI-powered startup intelligence agent — hand-built ReAct loop, parallel tool execution, live web search, and RAG-grounded document analysis. No frameworks. Full understanding.
Lightweight Retrieval‑Augmented Generation (RAG) demo with Streamlit UI. Ingest PDF/TXT docs, chunk & embed with Google GenAI or other providers, store in Pinecone, and query via chat. Optional MongoDB logging. Simple backend + front‑end for fast document search, retrieval, and generative answers.
Agent-first document ingestion layer on Docling with Markdown, image sidecars, manifests, OCR, and quality gating.
Blood-work/lab levels tracking app that is both privacy oriented and visually appealing
AI revision platform for university past papers, with PDF ingestion, hierarchical chunking, vector semantic search, query planning, and citation-grounded RAG study guide generation.
Hybrid Q&A
Evidence-first event intelligence prototype for policy monitoring, with source-grounded RAG, change detection, event briefs, and PDF ingestion.
MédiRDV — Academic medical appointment assistant chatbot powered by a local LLM (phi4-mini via Ollama) with RAG on healthcare PDFs. Built with React, TypeScript, and shadcn/ui. Epitech B-AIA-210 project.
Multilingual campus chatbot that grounds Gemini responses in uploaded PDFs, built with React, FastAPI, Express, MongoDB, LangChain, ChromaDB, and Ollama.
Local AI Question-Answering bot built with Ollama and LangChain using a RAG (Retrieval-Augmented Generation) pipeline. Answers queries from local PDF/text documents with fast, private inference.
Fully local Retrieval-Augmented Generation (RAG) chatbot powered by FAISS vector search and Ollama LLMs. Supports PDF, TXT, and Markdown ingestion, fast similarity search, model switching, and intelligent document-aware Q&A. Everything runs locally with zero external API calls, wrapped in a clean Streamlit interface.
Hybrid RAG pipeline — ingest PDFs into Qdrant (vector) + Neo4j (graph), query via a Claude agent skill
LLM-agnostic persistent wiki-memory for any AI agent. OpenClaw plugin, D3.js knowledge graph, PDF ingestion, WebSocket live updates, 124 tests.
To associate your repository with the pdf-ingestion topic, visit your repo's landing page and select "manage topics."