Show HN: GitHub-assistant – Natural language questions from your GitHub data (github.com)
Simon(sfarshid) and I spend a lot of time on GitHub. As data nerds we put together a quick tool to explore your repository’s data.
How it works:
- Data Loading: We use dlt to pull data (issues, PRs, commits, stars) from GitHub
- Semantic Layer: Relta wraps the underlying dataset into a semantic layer so the LLM doesn’t hallucinate.
- Text-to-SQL: A text-to-SQL agent transforms your plain-English question into a query using the semantic layer
- Generative Charts: assistant-ui dynamically generates a chart based on the SQL query
- Refinements: If the semantic layer can’t handle your question, our agent submits semantic layer improvements via pull requests
Hosted version: https://github-assistant.com
Demo Video: https://youtu.be/ATaf98nID5c
Check out the repo + hosted version and let us know what you think.
16 comments
[ 3.0 ms ] story [ 61.8 ms ] threadPut the video in the Readme!
The idea is for the system to provide answers that have close to 100% accuracy, but make it a single click for developers to to improve the semantic layer.
[0] - https://popsia.com
- The GitHub graphql API limits to 100 items to be queried at a time and has pretty opaque secondary rate limits. Building this with cURL would take effort. dlt handles all this complexity to set up a robust pipeline by providing a connector to the GitHub API. - Creating semantic layers manually from a relational dataset and leveraging it in a text-to-sql pipeline to prevent hallucinations (similar to those we highlighted in our Medium post) would take lots of manual effort, which Relta streamlines. - Creating a chat front-end with charts was made easy by assistant-ui
Hope this makes sense.