- Project
- 03
- Category
- Developer Infrastructure
- Period
- December 2025
File Intelligence
AI-Powered File Intelligence Infrastructure
Overview
Local-first semantic search across your own files. Embeddings and vector search run on the machine, with cloud models as an option rather than a requirement.
- 01
Developed a privacy-aware semantic search system using embeddings and ChromaDB for local-first organization.
- 02
Designed a modular FastAPI backend supporting offline/local and cloud-based LLM workflows.
- 03
Built a React/TypeScript interface for file exploration, semantic search, and AI-assisted understanding.
Problem
Understanding your own files usually means uploading them somewhere. That is an unacceptable trade for anything sensitive, and an unnecessary one when the hardware to run the model is already on the desk.
Approach
- 01
Make local the default path: Ollama for inference, ChromaDB for vectors, nothing leaves the machine unless asked.
- 02
Keep the backend provider-agnostic so local and cloud LLM workflows are swappable behind one FastAPI interface.
- 03
Index by meaning rather than filename — embeddings drive exploration and search.
- 04
Build the client in React and TypeScript with Zustand holding search and exploration state.
Architecture
- 01IngestFile traversal and chunking
- 02EmbedLocal embedding model
- 03StoreChromaDB vector index
- 04ServeModular FastAPI — local or cloud provider
- 05ExploreReact + TypeScript + Zustand client
Technology
- React.js
- FastAPI
- Python
- ChromaDB
- Ollama
- TypeScript
- Zustand
Project details
Details available on request.
Gallery
