|
|
--- |
|
|
title: OpenSorus - OSS Maintainer Agent |
|
|
emoji: ⚡️ |
|
|
colorFrom: yellow |
|
|
colorTo: red |
|
|
sdk: gradio |
|
|
sdk_version: 5.33.0 |
|
|
app_file: app.py |
|
|
pinned: true |
|
|
license: mit |
|
|
short_description: GitHub Issues Maintainer Agent for Open Source Projects |
|
|
--- |
|
|
# OpenSorus - AI Maintainer Agent for GitHub Issues |
|
|
<p align="center"><img width="65%" src="https://res.cloudinary.com/ivolve/image/upload/v1749392337/opensorus-banner_wel6po.jpg"/></p> |
|
|
|
|
|
> The OSS Copilot for automated triage & instant GitHub issue support with helpful context aware replies. |
|
|
|
|
|
OpenSorus is an AI Agent distributed as a GitHub App that reads your GitHub issues, understands your codebase, and responds with a comment to help your users/contributors. It acts like a first-level dev support assistant of your project that never sleeps and actually understands your codebase. |
|
|
|
|
|
> A submission for agent-demo-track. |
|
|
|
|
|
## Features |
|
|
|
|
|
Once installed and triggered on a GitHub issue (either via @mention or through the Gradio interface), OpenSorus autonomously: |
|
|
|
|
|
- Pulls issue context from GitHub |
|
|
|
|
|
- Indexes relevant parts of your codebase to match with the issue using semantic similarity search mechanisms. |
|
|
|
|
|
- Retrieves relevant code snippets or docs from the repo. |
|
|
|
|
|
- Crafts a useful, relevant reply by using the context from the codebase and generative capabilities of its own. |
|
|
|
|
|
- Posts the response back as a GitHub comment. |
|
|
|
|
|
## Goal |
|
|
Open source projects often get overwhelmed with open issues, thanks to their inclusive and collaborative nature. But that openness should feel empowering, not exhausting, for all, including maintainers & contributors. |
|
|
|
|
|
OpenSorus aims to make the life of open-source projects a little easier by handling the P1-level issues. |
|
|
Built for open source maintainers & teams who want to reduce issue backlog, increase community engagement and free up time to focus on critical issues. |
|
|
|
|
|
## Demo Video |
|
|
🎥 Watch the [Demo Video here](https://www.loom.com/share/d39697a60b944dbb938c3952d66cdc62?sid=1e730996-c912-4089-b717-a42ac9fbfe25). |
|
|
|
|
|
|
|
|
## Usage |
|
|
- Install the [OpenSorus GitHub App](https://github.com/apps/opensorus). |
|
|
|
|
|
- Configure this app for a particular repository by giving access to the repo from the dropdown. |
|
|
|
|
|
> Follow this guide to learn more about how to [install GitHub apps](https://docs.github.com/en/apps/using-github-apps/installing-a-github-app-from-a-third-party#installing-a-github-app). |
|
|
|
|
|
Once, you're done installing, there are primarily two ways to use the agent: |
|
|
### 🔁 Option 1 (Quicker): Auto-trigger via GitHub mention |
|
|
1. In any issue, simply comment `@opensorus`. |
|
|
|
|
|
2. The agent reads your issue, understands your repo, and replies back within seconds. |
|
|
|
|
|
### 💻 Option 2: Use the Gradio UI on HF Spaces |
|
|
1. Visit this [space](https://huggingface.co/spaces/Agents-MCP-Hackathon/OpenSorus). |
|
|
2. Paste your GitHub issue URL. |
|
|
|
|
|
3. Enter the primary branch name (e.g., main or master). |
|
|
|
|
|
4. Click Run Agent 🚀. |
|
|
5. Agent logs back a success message, & you're done! |
|
|
|
|
|
(Check back the issue's comments for updates). |
|
|
|
|
|
## Tech Stack |
|
|
|
|
|
Here’s the tech stack that made it all possible: |
|
|
|
|
|
🧠 LLM: Devstral (via Mistral API 🧡) |
|
|
|
|
|
🧬 Embeddings Generation: Codestral-Embed (via Mistral API 🧡) |
|
|
|
|
|
🗂️ Indexing, Embeddings storage & retrieval: LlamaIndex 🦙 |
|
|
|
|
|
🔎 Querying Context: Codestral (via Mistral API 🧡) + LlamaIndex 🦙 |
|
|
|
|
|
🧱 Infra Provider: Modal Labs |
|
|
|
|
|
🛠️ GitHub Integration: GitHub REST API |
|
|
|
|
|
✨ Web Interface: Gradio UI + Hugging Face Spaces |
|
|
|
|
|
|
|
|
## Limitations |
|
|
- Expects a well defined issue description for a better response. Vague descriptions may result in unhelpful/vague/irrelevant comment. |
|
|
|
|
|
- Rate limits may apply if too many requests come frequently. |
|
|
|
|
|
- Currently optimized for small-mid scale repositories. Might lag or take longer on large/heavy codebases. |
|
|
|
|
|
- Only reads issue description as a context, not any further comments on the issue. |
|
|
|
|
|
- Doesn’t handle PRs or other discussions yet (only issues for now). |
|
|
|
|
|
## Support |
|
|
For any feedback, support or bug report: |
|
|
- Feel free to [open a discussion](https://huggingface.co/spaces/Agents-MCP-Hackathon/OpenSorus/discussions?status=open&type=discussion&sort=recently-created) on HF. |
|
|
|
|
|
- You can also [open an issue](https://github.com/aditi-dsi/opensorus) on GitHub. |
|
|
|
|
|
- Or feel free to reach out via DM on [X](https://x.com/halfacupoftea_) or [LinkedIn](https://www.linkedin.com/in/aditi-bindal/). |
|
|
|
|
|
|
|
|
## Acknowledgements |
|
|
Made with ❤️ by [Aditi Bindal](https://huggingface.co/halfacupoftea) |
|
|
|
|
|
Big Thanks to Mistral AI & LlamaIndex for incredible LLMs and Agentic tools and to Hugging Face for providing the opportunitiy to build this agent. |
|
|
|
|
|
And special shoutout to ChatGPT for designing such a cute logo! |
|
|
|
|
|
|
|
|
## License |
|
|
This project is licensed under the MIT License – see the LICENSE file for details. |
|
|
|
|
|
|
|
|
|
