RepoAssist is an AI-driven platform that connects with your Git repositories, leveraging the capabilities of Gemini 2.5 flash, Langchain, and RAG to provide real-time insights and interactions with your codebase. Whether you're working solo or collaborating with a team, RepoAssist makes managing, querying, and understanding your codebase easier and more efficient.
Watch the demo video below:
RepoAssist seamlessly integrates with your GitHub repository to allow you to interact with your codebase using advanced AI models. By linking your repository, you can query your codebase, see your last 15 commits with summarized insights, and use speech-to-text AI to transcribe meetings and generate summaries.
The repo files are treated as documents in Langchain.
Use getSummary(document.pageContent) for example, bankers-algo.c.
This reduces the load on Gemini and ensures better embeddings.
Output:
This C code implements the Banker's Algorithm, a deadlock avoidance algorithm. It simulates resource allocation among five processes (P0-P4) using three resource types. The algorithm checks if a safe sequence exists, meaning all processes can finish without deadlock. It uses allocation, maximum, and available resource matrices to determine if a safe sequence exists and prints the sequence if one is found; otherwise, it indicates that the system is unsafe (a deadlock could occur).
Use getEmbeddings(document.summary) to generate vector embeddings like:
[0.345, 1.445, 0.455, 2.578]
Use getEmbeddings(document.summary) --> queryVector.
The more similar the query vector is to different lines, the more similar they are. For example, service.ts is quite similar to query vector meanwhile index.ts doesn't matches at all, showing they are at opposite directions.
a. The files are first uploaded to Cloudinary, which in return provides you with a URL.
b. Then, we create a meeting based on its defined schema.
c. After processing, the status is updated, and any issues generated for summaries are recorded.
Easily link your GitHub repository to RepoAssist. Once connected, you can ask AI-driven questions about your codebase. Whether it's understanding complex functions or tracking changes, Langchain and RAG provide accurate, contextual answers about your code.
Gain deep insights into your codebase with Langchain and Gemini 1.5. You can query specific parts of your code, understand logic, and ask complex questions, all with quick, clear responses. The system analyzes your code's context and offers relevant, actionable insights in real-time.
Upload audio files of your team meetings to Cloudinary and have them automatically transcribed with Assembly AI. The system generates accurate summaries of discussions, helping you stay on top of key takeaways, action items, and next steps, without having to sift through long audio files.
Review and summarize your last 15 commits with detailed notes, thanks to Gemini 1.5. Understand what was changed, why, and how it fits into your project without diving deep into the code manually. Cloudinary provides secure storage for any related media files, offering a comprehensive view of your project history.
Start with free credits and easily purchase additional credits via Stripe when you run out. Premium features like unlimited queries, document uploads, and meeting transcriptions are available with a one-time or recurring payment, ensuring a seamless experience.
Save answers to your questions or meeting summaries for later reference. Although the system primarily processes real-time queries, it can also save valuable insights and summaries for future use, helping you track decisions and ideas over time.
Built with Next.js and TRPC Protocol, RepoAssist offers a highly responsive and fast experience. Whether you're interacting with your repository or uploading audio files for transcriptions, expect swift processing times and a smooth user experience.
With Cloudinary as the media storage provider, RepoAssist ensures secure storage for your audio files, commit history media, and other important documents. The cloud-based solution offers scalable and reliable storage that can handle large amounts of data.
- Frontend: Next.js
- Backend: TRPC Protocol
- AI & Embedding: Langchain, Gemini 1.5, RAG
- Database: PostgreSQL, Neon DB
- Cloud Storage: Cloudinary (for audio file storage)
- Authentication: NextAuth.js
- Payment Integration: Stripe
RepoAssist is a powerful AI-driven tool for efficient codebase management, and we’re excited about its future! Here are some key features and improvements we’re working on:
Currently, RepoAssist is constrained by GitHub's API rate limits, which may restrict the frequency and volume of interactions with large repositories or organizations. We are actively optimizing our API calls and exploring alternative methods to reduce dependency on GitHub's rate limits. This will improve performance and allow more seamless interactions with repositories.
While RepoAssist provides real-time answers to questions, we plan to introduce advanced query optimizations that will better handle large codebases. Features like code context preservation across sessions will allow for even more accurate, detailed, and contextually aware responses.
In future versions of RepoAssist, we plan to introduce AI-driven code refactoring suggestions. RepoAssist will not only answer questions but will also suggest code improvements, optimizations, and best practices, helping you maintain clean and efficient code.
RepoAssist currently integrates with GitHub, but we plan to expand support for other version control platforms like GitLab and Bitbucket. This will make RepoAssist a more versatile tool for developers working across different ecosystems.
We’re continuously working on improving our meeting transcription and summarization feature. Upcoming enhancements include support for multi-speaker transcription, better summary accuracy, and automatic action item extraction. This will make it even easier to stay on top of key takeaways and next steps from meetings.
We aim to introduce real-time collaboration, where teams can simultaneously query the codebase, share insights, and collaborate on solutions. This feature will enhance the dynamic nature of team workflows and improve collaboration within projects.
As part of our broader vision, we plan to develop a mobile app version of RepoAssist. This will allow users to interact with their repositories, query code, and access meeting summaries on the go, making RepoAssist even more accessible and useful.
Stay tuned for more updates as we continue to enhance RepoAssist!