Build In public Agent

AI agent built during AgentHack2025 hackathon by WeMakeDevs.

This Agent gets triggered by updates in the default branch of the repository (like pushes, PR merges etc) and sends the code sumary and code snippet image to the slack channel only if the significance of the code changes are high enough, or else it gets ignored.

Architechure (Agent Workflow)

Here's Demo Video: https://www.youtube.com/watch?v=B7jQonYjjrk

1.First when the user pushes or merges a PR, our flask backend gets code changes data from the github webhook.

2.This server performs some checks to confirm that this is the smae repo and branch as the original one , then it cleans the data and pushes the data into a Redis Queue.

3.The AI agent pulls this data out from the Queue and uses it's significance tool to find the significance of the code changes.

4.If the significance is low, it gets ignored but if it is medium or higher, then the AI agent decides to use a code summary and snippet tool to get the code summary and the condensed code snippet of the changes that were pushed.

5.Next AI agent uses get code snippet image Tool to get the image of the code snippet generated by the previous tool. This tool uses fetches from a custom backend we built in nodejs + express.

6.Then the code summary and code snippet image is sent to slack channel of the company or developer.

Problems Solved

• Companies: They can have a code changes channel where they can have all the changes updated by the developers.

• Open Source Orgs: They can use this to show their transperency in development by having their updates sent to public channels, where others can get these updates.

• Solo Developers: Devs can have their build in public work done by this AI Agent. They can also track their high significance code changes.

How we are unique?

• This isnt your typical AI automation. This is an Agent which is intelligent enough to make it's own decisions.

• This Agent doesn't just put all your changes in the slack. It only selects those changes which have high significance. It ignores all the minor changes like comments, formatting etc.

• This also provides cool code snippet images for which we have built a backend image generator REST API.

Tools

• ngrok – Expose local server for webhooks
• GitHub Webhook – Trigger events on code changes
• Node.js + Express – Handle and route incoming requests
• Flask – Backend logic and AI orchestration
• Docker – Containerize and manage services
• Redis – Message broker for task queue
• RQ – Process background jobs
• Portia AI Agent – Run automated AI workflows
• OpenAI – Text generation and processing
• Gemini LLM – NLP tasks and analysis
• Perplexity AI – Research and information retrieval
• GPT – Querying and summarization
• Portia SDK – Build/customize AI agents
• Slack – Notifications and communication
• Slack SDK – Develop Slack tools/bots
• Puppeteer – Generate code snippet images
• Eraser.io – Create workflow diagrams

How to setup locally?

1. clone the project and create a python virtual environment in the root dir and activate it.

2. create a .env file in the root directory and fill it while referring .env.example file

3. install all the dependencies from requirements.txt

   pip install -r requirements.txt

4. Configure the ngrok and run a ngrok tunnel on port 3000.Then copy the link provided by ngrok and paste it in github webhook settings.

5. Then run a docker container with redis/redis-stack image and map port 6379 to 6379.

6. Open the img-gen-backend in another terminal and run:

   node server.js

7. Run server.py, worker.py and server.js in seperate terminals.

8. Cool. You are ready to go. Happy building.

Note: For optimal performance, use API keys from a paid instance. Free-tier limits often cause interruptions due to restricted token capacity.