My Hands-Free AI Streaming Setup (CodeRabbit + Claude Code)

Project Overview and Objectives

• The goal is to create a hands-free AI streaming setup that allows seamless switching between devices, including a DJX Spark, Mac Mini, and MacBook, using voice control and Twitch chat commands, with the help of Code Rabbit and Claude Code 10s.
• The setup aims to utilize OBS web socket and integrate with Twitch for features like text-to-speech (TTS) and voice control, enabling commands such as "Switch to the DJX Spark" or "Go to camera one" 2m6s.
• To achieve this, a GitHub workflow will be used, with Code Rabbit acting as a quality gate to review code in between phases, and Claude Code 47 will be leveraged for coding, due to the unavailability of Codex tokens 4m47s.
• A Product Requirements Document (PRD) has been created to outline the project's objectives, and a cloud MD file will be used to track the setup and progress, with a focus on creating an agentic engineering loop 6m6s.

Development Workflow and Tools

• The project will be completed in phases, with each phase involving planning, implementation, running the Code Rabbit agent, and opening a pull request (PR) for review, using Code Rabbit's CLI and dashboard for monitoring and feedback 8m6s.
• Code Rabbit is an AI-powered platform for code review, planning, and development workflows, allowing for real-time feedback and integration with GitHub, Jira, and Slack, which will be utilized throughout the project 10m6s.
• The process begins with GitHub authentication and installing Code Rabbit, which is done by heading to github.com/apps and selecting the desired repository, in this case, "AI Stream", to grant Code Rabbit access and enable features like agentic PR checks 10s.

Setup and Initial Integration

• After installing Code Rabbit, the first phase of the setup involves the director core, HTTP API, and OBS integration, and this is considered the preferred branch, allowing for testing and automatic code review with Code Rabbit after each phase 2m6s.
• The tests are run, and upon passing, CodeRabbit is used to review the work done so far, utilizing the CLI agent to connect to the review service, prepare a sandbox, and analyze the first phase 4m30s.
• CodeRabbit's review findings are reported back to CloudCode, which then autonomously fixes the errors found, creating a genetic loop where the process is mostly autonomous, requiring only occasional checks 6m40s.

Code Review and Autonomous Fixes

• The loop continues until no more findings are reported, and the process moves on to the next phase, involving committing, pushing, and opening a pull request, which triggers Code Rabbit's pull request review 8m50s.
• The pull request review is showcased, with Code Rabbit starting to check the pull request, running various checks, and reporting the results, demonstrating the integration of Code Rabbit and CloudCode in the development workflow 11m20s.
• The process involves polling every minute to check for comments and merge a pull request (PR) into the project, with the goal of setting up an autonomous workflow 10s.

Pull Request Review and Merging

• Code Rabbit completes the PR review, and upon refreshing, there are items to change, including config.ts, with four actionables and three nitpicks 42s.
• The autonomous loop involves pulling, reviewing, fixing, and testing until everything is complete, allowing for a continuous workflow, especially for more serious projects 2m6s.
• The tests are run, with 42 of 42 tests passing, and the loop is repeated until everything looks good, at which point the PR can be merged 4m30s.
• The merged PR is then reviewed on GitHub, and the first fully reviewed and tested phase of the project is complete, with the ability to continue to the next phase 6m15s.

Monitoring and Dashboard Insights

• The Code Rabbit dashboard provides information on active repositories, acceptance rates, and other details, allowing for monitoring of the project's progress 8m20s.
• The setup involves running an agent, PR review, and merging, with everything tested and ready for the next phase, and the goal is to use voice, agent, chat, and other controls to manage the stream 10m30s.

Stream Setup and Hands-Free Functionality

• The stream setup includes FFmpeg listeners, a voice model like Parakeet on the DGX Spark, and connections to Twitch chat and OBS, allowing for a hands-free streaming experience 12m40s.
• The hands-free AI streaming setup is being tested, and voice commands are used to switch between different devices, such as DGX, MacBook screen, and MacBook camera, with some lag but overall successful functionality 10s.
• The setup allows for dictation in text fields, including a "push to talk" feature, and the ability to start a Twitch stream using voice commands, with a link to the Twitch channel provided in the description 42s.
• The stream is live, and voice commands can be used to change the camera and end the stream, with some minor issues with lag, but overall the setup is working as intended 2m6s.

Testing and Future Improvements

• The pipeline setup is working out of the box, and the user is happy with the results, with plans to test the setup further on Twitch over the weekend 2m6s.
• Additional features, such as text-to-speech (TTS), may be added in the future, and the user is considering using Claude code to control the stream in a future project 4m30s.

Closing Remarks and Sponsorship

• The user recommends checking out Code Rabbit, a tool for managing code bases and performing Git reviews, which offers a free version with limited requests and reviews, and is sponsoring the video 6m0s.
• The user thanks Code Rabbit for sponsoring the video and invites viewers to check out their Twitch stream over the weekend for AI automation and to have a great weekend 8m0s.