GPEX-Cloud
GPEX-Cloud
Launch OpenGPEX

OpenGPEX 1.x documentation (legacy editor).

v2 Docs

LLM Integration

OpenGPEX integrates with Large Language Model (LLM) services through the AI Bridge — a unified interface for generating and editing images, describing images, and chatting with AI agents.

Overview

The AI Bridge drawer provides three operation modes:

Mode API Endpoint Description
Generate /v1/images/generations Create new images from text prompts
Edit /v1/images/edits Modify an existing image with text instructions
Describe /v1/chat/completions Get a text description of the current image

The Bridge also supports Agents: named endpoint + model configurations for multi-turn chat conversations with the model (see Settings).

Setting Up an Endpoint

The AI Bridge ships with built-in provider presets — OpenAI, Anthropic, Google Gemini, xAI Grok, Qwen (Aliyun DashScope), Ollama (self-hosted), and LocalAI (self-hosted) — and works with any OpenAI-compatible API:

  1. Open the AI Bridge drawer (right sidebar)
  2. Open Settings to configure endpoints
  3. Add an endpoint from a provider preset, or edit the default one:
    • Name: Display name for this endpoint
    • Base URL: The API base (e.g., https://api.openai.com)
    • API Key: Your authentication key
  4. Click Fetch Models to retrieve available models from the provider
  5. Select your preferred model

Multiple Endpoints

You can configure multiple endpoints and switch between them — useful for comparing results or using different providers for different tasks.

URL Validation

The system auto-corrects common URL mistakes:

  • Strips /v1/images/generations or /v1/images/edits suffixes
  • Removes trailing /v1 (you only need the base URL)
  • Validates URL format

Generate Mode

Create new images entirely from text descriptions:

  1. Switch to Generate mode
  2. Enter your prompt (what you want to see)
  3. Optionally enter a negative prompt (what to avoid — depends on provider support)
  4. Choose size (preset dropdown with 1024×1024, or type any WxH value or auto)
  5. Set seed (-1 for random, or a specific number for reproducibility)
  6. Click Generate

The generated image is added to the canvas as a new trunk creation.

Edit Mode

Modify your existing canvas content using text instructions:

  1. Switch to Edit mode
  2. Choose the input source: the active layer, or the merged (flattened) frame
  3. Enter a prompt describing the desired edit
  4. Click Generate

Describe Mode

Get a text description of your image:

  1. Switch to Describe mode
  2. Choose the input source: the active layer, or the merged frame
  3. Enter an instruction (e.g., "Describe this image")
  4. The model's text reply is shown in the panel — no image is generated

Generation Parameters

Parameter Description
Prompt Text description of desired output
Negative Prompt Elements to avoid in generation (hidden when the provider doesn't support it)
Size Output dimensions as WxH (default 1024×1024; free-form input, or auto)
Seed Random seed (-1 = random each time; fixed value = reproducible results)
Model Which AI model to use (fetched from the provider's model list)

Not all providers support every parameter — controls are hidden automatically when the selected provider declares no support.

Generation History

Every generation is recorded with full metadata:

  • Timestamp, endpoint, model, mode
  • Prompt and negative prompt
  • Seed and size settings
  • Success/failure status and duration

View past generations in the History panel to reference or re-use settings.

Privacy & Data Flow

  • The AI Bridge sends your image data only to the providers you configure
  • No built-in analytics or telemetry on your prompts
  • API keys are stored locally in your browser (pluginConfig / localStorage)
  • You control which server receives your data

Next Steps

Last updated: 2026-07-30