Summary

This project provides a local, first-class AI image generation and creative workspace. It integrates tldraw for an interactive canvas, Hono for a local API, SQLite for data persistence, and GPT Image 2 (or compatible models) for image synthesis. It targets users seeking a flexible, offline environment for prompt-to-image generation, reference-based image creation, and complex multi-step AI planning workflows.

How It Works

The architecture combines a Vite+React frontend leveraging tldraw for a dynamic canvas interface with a Hono-based backend API. Image generation requests are processed via configured providers (OpenAI-compatible endpoints), with results stored locally alongside project state and history in SQLite. A key feature is the Agent tab, enabling users to define multi-image tasks, which are then executed as a Directed Acyclic Graph (DAG), allowing for parallel processing and dependency management. Selected canvas images can serve as multimodal inputs for planning or as references for generation.

Quick Start & Requirements

• Primary Install/Run:
  1. pnpm install
  2. cp .env.example .env (or Copy-Item .env.example .env on Windows PowerShell)
  3. pnpm dev
• Prerequisites: Node.js 24.15.0, pnpm 9.14.2. An OpenAI API key (for gpt-image-2 or compatible) or Codex login is required for generation. Docker Desktop is optional for the Docker workflow.
• Access: The web app is available at http://localhost:5173.
• Links: No external documentation links provided beyond the README.

Highlighted Details

• Interactive Canvas: Generate and arrange AI images directly on a tldraw canvas, using existing images as references.
• Agent Planning: Define complex, multi-image generation workflows with DAG-based execution and retry capabilities.
• Local-First Data: All project state, generation history, and assets are stored locally in SQLite and the file system.
• Gallery View: Browse, rerun, locate, and manage locally generated assets.
• Optional Cloud Backup: Supports backing up generated images to Tencent Cloud COS.
• Flexible Provider Config: Configure image generation endpoints via .env files, an in-app dialog, or Codex login.

Maintenance & Community

No specific details regarding maintainers, community channels (e.g., Discord, Slack), or project roadmap are provided in the README.

Licensing & Compatibility

• License: MIT.
• Compatibility: The MIT license permits commercial use and integration into closed-source projects. The application is designed for local workstation use and requires additional security measures if exposed publicly.

Limitations & Caveats

The project is at version v0.2.0, indicating early development. Agent messages are not persisted in this version. Running pnpm dev and Docker concurrently against the same data directory is discouraged. Users must configure an image provider for generation capabilities. Potential SQLite version mismatches or corruption issues may require specific troubleshooting steps.