siva-gunasehkaran/contextory-ai

Contextory 🧠 — Local-First AI Memory Assistant

🔗 Live App: https://www.contextory.app

Contextory is a production AI application designed to help users capture, search, and recall their work context using local-first LLMs.

🚀 What does Contextory do?

• —Semantic search over personal work history
• —Daily & weekly AI-generated summaries
• —“What did I work on?” queries with context retrieval
• —Built with privacy-first principles

🧠 AI Architecture

• —LLMs: Local models via Ollama (LLaMA / Mistral)
• —Embeddings: Local embedding models
• —Vector Database: Chroma
• —RAG: Retrieval-Augmented Generation
• —Agent patterns: Multi-step reasoning for summaries
• —Backend: FastAPI-based AI services

🔍 RAG Flow

User Query → Embedding → Vector Search (Chroma) → Context Retrieval → Prompt Construction → LLM Reasoning → Grounded Response + Citations

🔐 Why Local-First AI?

• —Strong privacy guarantees
• —No data sent to third-party APIs
• —Predictable cost
• —Works offline

🧑‍💻 About the Builder

Built by a full-stack engineer transitioning into AI Engineering, with a focus on:

• —LLM system design
• —RAG pipelines
• —Agent orchestration
• —Production-ready AI APIs

📌 Status

✅ Live & deployed 🧠 Actively evolving 📦 Open to feedback