PyIndex — AI-Powered Enterprise Search

An enterprise search demo: real documentation ingested from a public API, indexed for search, and answered with Claude-generated, cited responses — built with FastAPI, React, and the Anthropic API.

📊 Summary

Built a FastAPI backend that ingests real package documentation from the public PyPI JSON API for 25 popular Python libraries.
Indexed the content with a TF-IDF search index (scikit-learn) for fast, ranked keyword search over the corpus.
Integrated the Anthropic Claude API to generate grounded, cited answers from the top retrieved passages — mirroring how enterprise search tools like Glean surface AI answers on top of raw results.
Built a React (Vite) frontend with a search bar, an "AI Answer" panel with inline citation chips linking back to sources, and a ranked list of raw search results.

🧰 Tools Used

FastAPI, scikit-learn, Python
Anthropic Claude API
React, Vite
PyPI JSON API (public data source)

📷 Screenshots

PyIndex search bar with the index status showing passages indexed across 25 packages

Claude-generated AI Answer with inline citation chips for a Pydantic validation question

📁 Source Code

View the full source, including the ingestion pipeline, search index, and Claude-powered answer generation, on GitHub.

View on GitHub