Speak2Build for Visual Studio Code

Voice-first AI pair programmer for Python projects. Speak or type a request, review the generated patch, validate it, and approve changes from inside VS Code.

Features

• Push-to-talk and hands-free voice input.
• Local transcription with Whisper and Silero VAD.
• Project-aware workspace context.
• Configurable Gemini, Groq, Ollama, NVIDIA NIM, OpenAI, and custom providers.
• Unified diff review with voice or button-based apply and reject actions.
• Pytest validation and AI-assisted repair.

Screenshots

Product walkthrough

Click the thumbnail to watch the product walkthrough

Installation

From the repository:

npm install
npm run build

To package and install the VSIX:

npx @vscode/vsce package
code --install-extension speak2build-0.1.0.vsix

The extension automatically starts the local backend sidecar. For development, press F5 from the repository root to launch an Extension Development Host.

Settings

Configure these from Settings > Extensions > Speak2Build or the in-panel gear button:

• speak2build.backendUrl: Backend URL, default http://127.0.0.1:8000.
• speak2build.llmProvider: Provider identifier.
• speak2build.llmApiKey: Provider API key; not needed for local Ollama.
• speak2build.llmModel: Optional model override.

Keyboard shortcut

• Ctrl+Alt+Space (macOS: Cmd+Alt+Space): Toggle voice recording.