FirstCommit

Inspiration

As developers, we remember how intimidating it was to make our first open-source contribution. FirstCommit was built to remove that friction — to help new contributors find meaningful repositories, understand the codebase quickly, and make their first PR with confidence.

What it does

FirstCommit recommends active open-source repositories tailored to the developer's skills and interests. It breaks down GitHub issues, explains the code relevant to the task using AI, and provides a guided walkthrough to help contributors make their first commit.

How we built it

We used:

• MongoDB Atlas Vector Search to store and retrieve code embeddings.
• Vertex AI to generate context-aware answers and summaries.
• GCP Cloud Run to host the backend services.
• Next.js on the frontend for a responsive developer interface.
• A custom data pipeline to embed and index the top 1,000 GitHub repositories from the GH Archive dataset.

Challenges we ran into

• Efficiently chunking and embedding large codebases without exceeding memory or timeout limits.
• Designing a guide system that was specific enough to be helpful, but general enough to scale across diverse repositories.
• Managing latency and cost while working with large language models and embedding APIs.

Accomplishments that we're proud of

• Automatically processed and embedded over 70 repositories, enabling real-time semantic search.
• Created a full-stack experience that bridges GitHub issues and code context using AI.
• Made open source feel more accessible to newcomers without sacrificing code understanding.

What we learned

• Practical applications of retrieval-augmented generation (RAG) in real-world codebases.
• How to combine vector search, LLMs, and structured metadata to guide users through unfamiliar repositories.
• That the open-source barrier to entry isn't technical it's about context. And we can solve that.

What's next for FirstCommit

• Expanding the dataset to include more repositories.
• Letting user's upload their own preferred repositories.
• Supporting personalised contributor onboarding based on repo context and contributor history.

Built With

• ai
• archive
• atlas
• cloud
• gcp
• gh
• google
• llms
• mongodb
• next.js
• python
• run
• search
• typescript
• vector
• vertex