DataStax Docs Chat — Agentic (beta)

Beta An agentic assistant over the DataStax documentation, powered by Astra DB hybrid search + NVIDIA reranking and an agent (Langflow + DeepSeek) that decides when to search, can search more than once, and will ask to clarify when a question is ambiguous. You'll see its steps as it works.

How it works

This is a beta of an agentic DataStax docs assistant — the endgame of the docs-chat experiments. Your message goes to a Cloudflare Pages Function on this site, which streams to a Langflow agent. The agent (driven by DeepSeek) decides when to call a search tool that runs Astra DB hybrid search — vector + lexical (BM25) with NVIDIA reranking — over the documentation corpus, then writes a grounded answer. You can watch each step it takes in the panel above the answer.

Answers are grounded in the official documentation only, with source links.

This is the third iteration. The earlier versions are still live: v1Astra DB Docs Chat, plain RAG over the Astra-DB-only docs; and v2DataStax Docs Chat, folder-routed RAG over the full DataStax corpus.

×
Page views: