This repository has been archived by the owner on Dec 4, 2024. It is now read-only.

dbt-labs / dbot Public archive

Notifications You must be signed in to change notification settings
Fork 4
Star 37

An LLM-powered chatbot with the added context of the dbt knowledge base.

Notifications

Name		Name	Last commit message	Last commit date
Latest commit History 19 Commits
context-sources		context-sources
test-docs		test-docs
.gitignore		.gitignore
.python-version		.python-version
README.md		README.md
app.py		app.py
code_crawler.py		code_crawler.py
csv_crawler.py		csv_crawler.py
document_crawler.py		document_crawler.py
markdown_crawler.py		markdown_crawler.py
question_answerer.py		question_answerer.py
requirements.txt		requirements.txt
test_code_crawler.py		test_code_crawler.py
test_csv_crawler.py		test_csv_crawler.py
test_markdown_crawler.py		test_markdown_crawler.py
test_question_answerer.py		test_question_answerer.py
test_vector_store.py		test_vector_store.py
vector_store.py		vector_store.py
vercel.json		vercel.json

Repository files navigation

Archival Notice

This repository has been archived.

As a result all of its historical issues and PRs have been closed.

Please do not clone this repo without understanding the risk in doing so:

It may have unaddressed security vulnerabilities
It may have unaddressed bugs

Click for historical readme

dbot

An LLM-powered chatbot with the added context of the dbt knowledge base.

An experiment

This is currently a basic CLI prototype that has been supplied with the knowledge of the dbt Developer Hub (docs, guides, etc) and uses this to add context to questions asked by users.

How does it work?

It stores embeddings of the material on the Developer Hub (chunked by markdown heading sections) in a vector database, and finds documents similar to the question that is asked. It then prepends as much of that context onto the prompt as it can fit in ChatGPT's context window.