Git Oracle

An AI tool that summarizes GitHub PRs using Gemini and generates multilingual voice narration with Murf AI to make code reviews faster and more accessible.

Comment

Inspiration

As developers, we’ve all experienced the fatigue of reviewing large GitHub pull requests. Sifting through dozens of comments and complex code changes is time-consuming and mentally draining. Our inspiration came from a simple thought: "What if you could listen to a code review like a podcast?" We wanted to create a tool that would make understanding code changes faster, more accessible, and less of a chore, allowing developers to catch up on important updates while away from their keyboards.

What it does

Git Oracle is an AI-powered developer tool that transforms any GitHub pull request into a concise, multilingual audio summary. A user pastes a PR URL, selects a language, and instantly receives a clean text summary and a high-quality voice narration of the key changes, discussions, and code diffs. The tool also intelligently analyzes the sentiment of the conversation to select a voice with the appropriate tone.

How we built it

We architected Git Oracle as a full-stack application, focusing on a powerful backend pipeline and a clean user experience. Backend (Node.js/Express): We built a server to orchestrate a multi-step AI process. It first fetches all PR data (title, description, comments, and code diff) from the GitHub API. AI Core (Google Gemini): This data is then sent to the Gemini API. We engineered a series of sophisticated prompts to have the AI first analyze the conversation's sentiment, then generate a concise summary, and finally translate it into the user's chosen language while preserving technical terms. Voice Generation (Murf AI): The final, clean text is sent to the Murf AI API, which generates a natural-sounding audio file using a voice selected based on the detected sentiment. Frontend (React): The user interface was built with React and TailwindCSS. It provides a simple input form and a custom-built audio player that features "karaoke-style" word highlighting, synchronizing the text with the audio playback for an engaging experience.

Challenges we ran into

Our biggest challenge was a persistent 404 Not Found error from the GitHub API that only occurred on one developer's machine. After hours of methodical debugging—checking headers, token scopes, and even testing from a different network—we proved the issue was environmental. This forced us to adopt a hybrid development strategy and taught us a valuable lesson in isolating platform-specific bugs. Our second challenge was in prompt engineering; we had to refine our prompts multiple times to prevent the AI from translating technical terms and to ensure it produced clean, voice-friendly text.

Accomplishments that we're proud of

We are incredibly proud of successfully building a seamless, end-to-end pipeline that integrates three complex, distinct APIs. Implementing the sentiment analysis to dynamically change the narrator's voice was a major accomplishment that makes the tool feel truly intelligent. On the frontend, we're proud of building a polished UI with a custom audio player and word-by-word highlighting, which elevates the user experience far beyond a basic prototype.

What we learned

This project was a deep dive into the practicalities of building an AI application. We learned how to write sophisticated, multi-step prompts to steer a large language model to a precise output. We mastered the integration of multiple asynchronous APIs and learned how to handle complex authentication and data transformation between them. Most importantly, we learned the value of strategic pivoting and methodical debugging when faced with persistent, real-world development roadblocks.

What's next for Git Oracle

The vision for Git Oracle is to become an indispensable tool for development teams. Our next steps would be to:

  1. Add support for more languages and voices.
  2. Develop smarter AI insights to automatically detect risky code changes or potential security issues.
  3. Create a browser extension or a GitHub App for even tighter integration into the developer workflow.
  4. Introduce collaboration features, allowing users to share and comment on the audio summaries with their team.

Built With

Share this project:

Updates