Codoro

Next-gen developer-first live streaming platform featuring sub-second streaming via AWS IVS, real-time Socket.io chat, and AI code assistants powered by Bedrock.

Comment

💡 Inspiration As developers, we love watching live coding streams to learn new frameworks, tools, and paradigms. However, generic streaming platforms like YouTube and Twitch are built for gaming and entertainment, not syntax. The sub-second interactions required to debug code in real time are ruined by 10-second stream latency, generic chat inputs don't support code blocks, and viewers are left to manually take notes on what was built.

We were inspired to build Codoro—a developer-first live streaming ecosystem designed to make live coding highly interactive, educational, and collaborative.

🚀 What it does Codoro is a next-generation interactive live coding platform that brings broadcasters and viewers closer together:

Ultra-Low Latency Streaming: Streams deliver at sub-second speeds, allowing creators to ask questions and debug with viewers in absolute real time. Developer Chat & Presence: A custom chat system equipped with live viewer count presences so developers can collaborate without delay. AI Coding Assistant: Viewers can paste code blocks directly into the sidebar to query an inline AI assistant trained to run code reviews, find bugs, and suggest optimizations. AI Session Summaries: Once a broadcast ends, our system automatically synthesizes a structured JSON summary card containing key topics, highlights, and difficulty tags so viewers can easily review what was built. ⚙️ How we built it We architected Codoro as a distributed serverless and realtime ecosystem:

Frontend / Serverless Engine: Built using React 19, Next.js 16 (App Router), TypeScript, and Tailwind CSS hosted on Vercel for lightning-fast page delivery. Video Delivery Pipeline: Integrated Amazon Interactive Video Service (IVS) to handle video ingestion (both OBS and direct browser streaming) and globally distribute HLS playbacks. Realtime Communication: Decoupled persistent WebSocket chat handlers into a standalone Socket.IO server hosted on Railway to prevent serverless function execution timeouts. Database: Configured Amazon DynamoDB as a single-digit millisecond NoSQL database to store stream metadata, communities, and real-time chat histories. AI Processing: Configured Amazon Bedrock (Nova Lite) to run conversational Q&A threads and parse completed stream transcripts into structured JSON summaries. Security & Auth: Secured routes and serverless actions on the Edge using Clerk Authentication. 🚧 Challenges we ran into Multi-Platform Lockfile Mismatches: Installing native bindings like @emnapi created build discrepancies between our macOS development laptops and Railway's Linux environments, which we resolved by optimizing lockfile packages resolution and forcing standard npm install runs. Serverless WebSocket Limitations: Managing WebSocket states inside Vercel's serverless handlers proved impossible due to connection timeouts. We resolved this by decoupling the websocket logic into a dedicated node microservice on Railway. Build-Time Prerendering Failures: Next.js attempted to compile our AI pages statically at build-time on Vercel without access to live AWS credentials. We resolved this by forcing dynamic runtime rendering and introducing try-catch safety fallbacks. 🏆 Accomplishments that we're proud of Successfully integrating Amazon IVS to deliver sub-second streaming direct from standard web browsers. Generating fully structured JSON summary outputs from Bedrock to render customized topic tags (React, Go, Python, AWS) and highlight checkboxes dynamically. Building a glassmorphic dashboard UI that feels responsive and alive. 🧠 What we learned We gained hands-on experience orchestrating multi-cloud architectures. We learned how to handle WebSocket clustering, how to structure prompt guidelines for strict JSON returns in Bedrock, and how to safely write serverless actions close to the Edge while keeping database latency minimal.

🔮 What's next for Codoro We plan to introduce a collaborative online IDE sidebar allowing viewers to edit code files alongside the stream, implement subscriber paywalls for gated premium courses, and use Amazon Transcribe to feed speech-to-text audio directly into Bedrock for even more accurate automatic code summary clippings.

Built With

Share this project:

Updates