# Building Fabric: How I Automated Stakeholder Reports from GitHub

I’ve always hated weekly meetings with non-technical stakeholders.

Explaining technical work is hard. If you’re experienced, you know the drill: don’t mention webhooks or backend code. You say, *"We added payments so customers can buy items,"* and keep it moving.

But then comes the request nobody talks about: **the written report.**

In most small startups, there is no product manager. You have to write everything down for compliance or tracking. I found myself wasting hours re-reading my own code, checking the backlog, and filling out Notion pages to remember what I built.

It was a waste of time because everything I built was already logged in my Git history.

I wondered: *If my work is already tracked in my commits and pull requests, why am I writing these reports manually?*

So, I built a tool to do it for me.

That is how **Fabric** started. An AI engine that turns your GitHub activity into finished reports for your stakeholders.

## Core Features

I focused on building a tool that handles the heavy lifting of reporting so you don't have to:

*   **Deep Repository Sync:** Goes beyond simple line counts. It extracts real-time commit metadata to capture the actual intent of your work.
    
*   **Guided Onboarding:** A clean, multi-step interface designed to get you set up quickly and reduce cognitive load.
    
*   **The Transparency Engine:** Generates high-fidelity Markdown reports focused on business outcomes.
    
*   **Report Persistence:** A centralized dashboard that lets you visualize and manage your historical reporting data.
    
*   **Persona-Driven Synthesis:** Features custom tone mapping, allowing you to tailor reports specifically for CTOs, Founders, or Board Members.
    

## The Tech Stack

I kept the architecture simple to focus on the AI logic rather than managing complex infrastructure:

*   **Next.js 16:** The backbone of the application. Its speed and routing make it the perfect choice for a developer-focused tool.
    
*   **Groq API:** This is where the magic happens. I chose Groq for its incredibly fast inference speeds, which are essential when feeding it batches of commit logs for summarization.
    
*   **Octokit:** The official GitHub SDK. It handles the heavy lifting of interacting with the GitHub API.
    
*   **Cloudflare R2:** Fabric extracts screenshots from PRs and re-hosts them on R2 (S3-compatible) to keep your internal tooling private without relying on GitHub authentication.
    

## What’s Next

Building Fabric has been about staying focused. To allow users to use Fabric faster, I’ve had to make some hard choices about what to ship later:

*   **Platform Focus:** Fabric is GitHub-only for now to prioritize the most common use cases. Support for GitLab and BitBucket is on the roadmap.
    
*   **Security & Scaling:** Currently, Fabric uses a simple, single-session experience without authentication. This lets individual developers and freelancers get immediate value without the friction of complex auth strategies.
    

### Conclusion

I'm already working on part two of this series, where I’ll dive into the technical architecture, prompt engineering and the core features.

Fabric is available at: https://github.com/franciscoluna-28/fabric-ai

Follow my journey and see my other work at [itsfranciscoluna.com](http://itsfranciscoluna.com).