Patchbay

Patchbay is a local, privacy-first control plane for manually routing and patching context between AI web sessions and official APIs. It enables hybrid workflows that combine UI-only capabilities (e.g., Deep Research, Canvas) with fast, stable API-based execution—without running a centralized backend.

🚀 Key Features

• Hybrid Provider Model: Choose between Web Mode (Browser Automation) and API Mode (Direct REST) for each AI model independently.
• Context Piping: Easily move data between models. Use Gemini Web for research and pipe the result directly to GPT-4 API for code generation.
• Privacy-First: All API keys are stored locally in your browser (localStorage). No data ever leaves your machine to an intermediate server.
• Zero-Backend (Production): The final application is a static web client that communicates directly with AI providers through a Chrome Extension bridge.
• Orchestra View: Compare responses from multiple AI models simultaneously.

🛠 Project Structure

• index.html: The main "Mission Control" dashboard.
• src/: Core application logic, including the Orchestration Engine and UI managers.
• docs/: Detailed design and architecture documentation.
• proxy.server.js: A temporary development proxy to handle CORS before the Chrome Extension is fully implemented.

🏁 Getting Started

Prerequisites

• Node.js (v18+)
• NPM

Installation

1. Clone the repository:

   git clone https://github.com/kiryltech/patchbay.git
   cd patchbay

2. Install dependencies:

   npm install

Running the Development Environment

During the initial development phases (Phase 1 & 2), a local proxy is required to handle CORS requests to AI providers.

Start the dashboard and proxy simultaneously:

npm start

The dashboard will be available at http://localhost:5173 (or the port specified by Vite).

🗺 Implementation Roadmap

• Phase 1 (Completed): API-First Core. Basic dashboard, API adapters, and secure key storage.
• Phase 2 (Completed): Advanced Orchestration. Multi-model support and context foundation.
• Phase 3: Collaborative UX. Group chat interface, @Agent targeting, unified context history, markdown rendering, and parallel streaming.
• Phase 4: Analytics & Optimization. Usage tracking (tokens/cost), response time monitoring, and prompt optimization tools.
• Phase 5 (Future): The Manual Bridge. "Delta Sync" workflow for manual state transfer to external web AIs (e.g., Deep Research).