feat: Add ModelsLab image generation provider

[adhikjoshi]

Summary

Adds ModelsLab as a new image generation provider to sim's Image Generator block, alongside the existing OpenAI providers (DALL-E 3, GPT Image).

What this PR adds

New Provider: ModelsLab

Five popular models available out of the box:

• Flux (Schnell) — fast, high-quality default
• Juggernaut XL — photorealistic portraits
• RealVisXL v5 — realistic photography style
• DreamShaper XL — artistic/creative
• Stable Diffusion XL — general purpose

How it works

The Image Generator block now has a Provider dropdown (OpenAI | ModelsLab). Selecting ModelsLab shows:

• Model picker (Flux, Juggernaut XL, etc.)
• Size presets (512×512 to 1344×768)
• Optional negative prompt
• ModelsLab API key field

Files changed (6)

File	Change
tools/image/modelslab.ts	Client-side tool definition
tools/image/types.ts	Shared request/response types
tools/image/index.ts	Tool exports
app/api/tools/image/generate/route.ts	Server-side handler with async polling
blocks/blocks/image_generator.ts	Added provider selector + ModelsLab options
tools/registry.ts	Registered image_modelslab tool

Architecture follows existing patterns

• Server-side route mirrors the /api/tools/video pattern for long-running generation
• Polls /api/v6/images/fetch/{id} every 5s until completion (max 200s timeout)
• Image URL fetched and base64-encoded via the existing /api/tools/image proxy

Usage example

Provider:         ModelsLab
Model:            Flux (Schnell)
Prompt:           a futuristic cityscape at sunset, ultra detailed
Size:             1024x1024
Negative Prompt:  blurry, low quality, watermark
API Key:          [ModelsLab API key from modelslab.com]

ModelsLab

• Website: https://modelslab.com
• API Docs: https://docs.modelslab.com
• Get API key: https://modelslab.com/dashboard (free tier available)

Checklist

• TypeScript compiles without errors
• Biome lint passes
• Follows existing video/TTS block provider-selector pattern
• Server-side route handles both sync and async ModelsLab responses
• Backward compatible — existing OpenAI blocks unchanged (provider defaults to openai)

[vercel]

@adhikjoshi is attempting to deploy a commit to the Sim Team on Vercel.

A member of the Team first needs to authorize it.

[greptile-apps]

Greptile Summary

Adds ModelsLab as a new image generation provider alongside OpenAI, enabling access to Flux, Juggernaut XL, RealVisXL, DreamShaper, and Stable Diffusion XL models. The implementation follows established patterns from the video generation feature, including a new API route with async polling, provider-based conditional UI in the Image Generator block, and proper separation of tool definitions and types.

Major changes:

• New image_modelslab tool with correct user-only visibility for API keys
• Provider selector dropdown in Image Generator block with conditional field rendering
• Server-side /api/tools/image/generate route that polls ModelsLab API until completion (max 200s)
• Size parsing logic converts dropdown values (e.g. "1024x768") to width/height parameters
• Backward compatible - existing OpenAI blocks default to openai provider

Confidence Score: 4/5

• This PR is safe to merge with minor considerations
• The implementation follows established patterns from video generation, correctly uses user-only visibility for API keys per custom instructions, and maintains backward compatibility. The code is well-structured with proper error handling and validation. Score reduced to 4 due to the extensive conditional logic in the block config which increases complexity.
• Pay close attention to apps/sim/blocks/blocks/image_generator.ts due to complex nested conditional logic

Important Files Changed

Filename	Overview
apps/sim/tools/image/modelslab.ts	ModelsLab tool definition with correct visibility settings and comprehensive parameter descriptions
apps/sim/blocks/blocks/image_generator.ts	Provider selector and ModelsLab integration with conditional UI - extensive changes to support multi-provider workflow
apps/sim/app/api/tools/image/generate/route.ts	New API route for ModelsLab image generation with async polling - mirrors video generation pattern

Sequence Diagram

sequenceDiagram
    participant User
    participant Block as Image Generator Block
    participant Tool as image_modelslab Tool
    participant Route as /api/tools/image/generate
    participant ML as ModelsLab API
    participant Proxy as /api/tools/image

    User->>Block: Select ModelsLab provider
    Block->>Block: Show ModelsLab options (model, size, negativePrompt)
    User->>Block: Configure & execute
    Block->>Tool: Call with params (provider, apiKey, model, prompt, width, height)
    Tool->>Route: POST request body
    Route->>Route: Validate params & auth
    Route->>ML: POST /api/v6/images/text2img
    ML-->>Route: status: "processing", id: jobId
    
    loop Poll every 5s (max 40 attempts)
        Route->>ML: POST /api/v6/images/fetch/{jobId}
        ML-->>Route: status: "processing"
    end
    
    ML-->>Route: status: "success", output: [imageUrl]
    Route->>Proxy: GET with imageUrl param
    Proxy-->>Route: image binary
    Route->>Route: Convert to base64
    Route-->>Tool: { imageUrl, imageFile, model, provider }
    Tool-->>Block: Success response
    Block-->>User: Display generated image

Loading

_{Last reviewed commit: 9645e69}

[greptile-apps]

_{6 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}