Documentation search with AI assistant

[skovy]

Hey all, having an AI assistant with access to specific data has become very popular. The primary benefit being you don't have to copy/paste content like documentation into an LLM, it can help find an answer faster and can fill in missing pieces between disparate pieces of documentation. It's not necessarily complex but does require a few pieces of infrastructure (server, database, vector store, LLM API). That's where Asktro (https://asktro.com) comes in, it's a hosted solution to unlock natural language search and an AI assistant that is specific to your docs.

You can think of it as Supabase-style docs search as a service.

It's leveraging Docusaurus's (and Nextra's) opinionated approach to ingest the documentation at build time to chunk, embed, and persist the documentation in a vector store. This allows both natural language search and an AI assistant.

You can see a demo of it on the Docusaurus starter project here: https://docusaurus.asktro.com

Docs on how it works to integrate: https://asktro.com/docs/docusaurus

Would love to hear your thoughts and feedback, thanks!

[slorber]

Looks nice 👍

We already have a other solutions proposed in this discussion: #8835

[ntucker]

Is this project dead?