DevSync

Inspiration

Software teams spend 40%+ of their time just understanding codebases instead of building features. We noticed that project knowledge is scattered across GitHub commits, meetings, Slack chats, and documents. This fragmentation slows down onboarding, reduces clarity, and often leads to duplicated work. We wanted a tool that acts as a single intelligent brain for the entire project—summarizing code, understanding architecture, analyzing commits, and making meetings instantly actionable. That idea sparked the creation of DevSync.

What it does

DevSync is an AI-powered platform that unifies: Code Intelligence – file-level summaries, commit explanation, architecture insights Meeting Intelligence – real-time transcription, summaries, task extraction Automated Documentation – system design, change logs, Q&A, and more AI Project Brain – ask anything about your codebase or meetings and get contextual answers.

In short, it turns your repository + meetings into an always-updated, searchable knowledge hub.

How we built it

We built Dionysus2 using a fully modern, scalable stack:

Frontend: Next.js 15, React 19 RC, Tailwind, Radix UI Backend: tRPC, Next.js API Routes Database: PostgreSQL + Prisma (with support for vector embeddings) AI: OpenAI, Google Gemini, AssemblyAI Auth: Clerk Storage: Firebase Deployment: Docker + Fly.io

Our architecture follows a modular T3 pattern, integrating AI pipelines for code parsing, embedding generation, semantic search, and meeting transcription. We also use math-based vector similarity to power contextual project Q&A.

Challenges we ran into

1.Handling large repository ingestion efficiently 2.GitHub API rate limits during recursive crawling 3.Maintaining LLM accuracy on complex architectures 4.Ensuring reliable real-time meeting transcription 5.Designing a unified UI/UX for code + meetings + documentation 6.Optimizing Prisma for vector-based semantic search

Accomplishments that we're proud of

1.Built a fully unified workspace that no existing tool currently offers 2.Achieved accurate code summaries and commit insights 3.Automated meeting transcription + task extraction 4.Integrated three major AI providers seamlessly 5.Designed a clean, fast, modern developer experience 6.Built an early prototype capable of answering deep technical questions about a project

What we learned

1.How to architect systems that combine multiple AI modalities 2.Efficient repository parsing and vector database optimization 3.The importance of UI/UX in productivity tools 4.Handling real-world constraints like API limits, latency, and accuracy 5.That AI can dramatically reshape how teams understand software

What's next for DevSync

1.Add real-time collaborator chat with AI memory 2.Build AI-driven pull request reviewers 3.Expand to multi-project knowledge graphs 4.Offer enterprise-level on-prem vector DB deployment 5.Create smart architecture diagrams generated directly from code 6.Launch a VS Code extension for inline summaries + Q&A 7.Implement fully autonomous “AI Project Manager” workflows

Built With

• actions)
• assets)-authentication-clerk-(user-management-+-secure-auth)-ai-&-ml-integrations-openai-api-(code-analysis
• billing)
• ci/cd
• code
• commit-data
• containerization)
• deployment
• deployment)
• devops
• docker
• documentation)-google-gemini-api-(insights
• eslint
• fly.io
• frontend-next.js-15-react-19-rc-typescript-tailwind-css-radix-ui-backend-node.js-trpc-next.js-api-routes-prisma-orm-database-&-storage-postgresql-(primary-relational-db)-pgvector-(vector-embeddings)-firebase-storage-(files
• github
• global
• hosting
• management)
• other
• package
• payments
• pnpm
• prettier
• q&a
• quality)
• reasoning)-assemblyai-(meeting-transcription-&-audio-intelligence)-developer-integrations-github-api-(repository-analysis
• schema
• stripe
• subscriptions
• tools
• transcripts
• webhooks)
• zod