---
pretty_name: AI Tutor Vector DB (public docs-only bundle)
viewer: false
---

# AI Tutor vector DB — public docs-only bundle

Prebuilt retrieval artifacts for the [Towards AI tutor chatbot](https://github.com/towardsai/ai-tutor-app):
a ChromaDB collection (Cohere `embed-v4.0` embeddings), BM25 index and
document dictionary, plus the file-based knowledge base (`kb/`) the agent
browses at runtime.

This is the **public fallback** for the private
`towardsai-tutors/ai-tutor-vector-db` bundle. It contains **documentation
sources only** (`claude_code_docs`, `deep_agents`, `langchain`, `langgraph`, `llama_index`, `openai_docs`, `peft`, `transformers`, `trl`); the Towards AI course content that ships in
the private bundle is not included. The app downloads this bundle
automatically on first start when no `HF_TOKEN` (or one without access to the
private dataset) is available.

Derived from the production artifacts by
`data/scraping_scripts/build_public_docs_bundle.py` — an allowlist filter over
the prod bundle (same embeddings, same tree layout), followed by a leak audit.
Documentation content belongs to its upstream projects and keeps their
respective licenses.