This repository provides a semantic search interface for the Voxel51 documentation, enabling users to query information using natural language. It targets developers and researchers working with the FiftyOne computer vision library, offering a more intuitive way to navigate documentation compared to traditional keyword search.

How It Works

The system leverages OpenAI's text-embedding-ada-002 model to generate vector embeddings for the FiftyOne documentation. These embeddings are stored and queried using the Qdrant vector search database. The approach uses LangChain splitters with custom Markdown parsing for robust document segmentation, enhancing the accuracy and relevance of search results.

Quick Start & Requirements

• Install via pip: pip install -e .
• Requires an OpenAI API key (export OPENAI_API_KEY=XXXXXXXX).
• Requires a running Qdrant server (docker run -d -p 6333:6333 qdrant/qdrant).
• Official docs: https://docs.voxel51.com/

Highlighted Details

• Command-line interface (fiftyone-docs-search query <query>) with options for result count, opening URLs, scores, and document types.
• Python API for programmatic access.
• Versioning aligns with FiftyOne documentation versions.
• Option to build the index locally from a cloned FiftyOne repository.

Maintenance & Community

• Project is associated with Voxel51, the creators of the FiftyOne library.
• Community engagement is encouraged via the FiftyOne Slack community.

Licensing & Compatibility

• The repository itself is not explicitly licensed in the README. The FiftyOne library is open source. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The project requires an external OpenAI API key, incurring potential costs. The README does not specify the license for the fiftyone-docs-search package itself, which may impact commercial use.