improvement(media-blocks): new versions of image and video gen with latest models + fixes

[icecrasher321]

Summary

New versions of image and video gen models

Support for new providers incl. Google Nano Banana, OpenAI Sora, Seeddance via Fal

Updated docs

Type of Change

• Other: Integration update

Testing

Tested manually

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		May 19, 2026 11:36pm

[cursor]

PR Summary

High Risk
Adds new server-side generation paths that call external AI providers and stores generated media, plus expands Fal.ai model/parameter validation; mistakes here can impact cost, latency/timeouts, and file handling/security around remote URL fetching.

Overview
Adds new generation-capable block versions and tools. Introduces image_generator_v2 (OpenAI GPT Image, Google Gemini/Nano Banana, Fal.ai multi-model) backed by a new image_generate tool and a new POST /api/tools/image route that generates, downloads/validates, and stores images (execution upload when available; otherwise storage fallback).

Updates video generation to a new video_generator_v3. Expands Fal.ai model catalog/params (durations, aspect ratios, resolutions, optional generateAudio) with centralized validation, improves MiniMax defaults and adds an endpoint option affecting resolution, and relaxes aspect-ratio validation for Fal.ai’s model-specific constraints.

Aligns product/docs plumbing. Updates docs pages, integration registry/icon mappings, integrations JSON (including docs URL change for video), built-in tool type lists, and the docs generation script to emit separate icon mappings for docs (including hidden versioned blocks) vs landing (visible only).

^{Reviewed by Cursor Bugbot for commit 71b4080. Configure here.}

[greptile-apps]

Greptile Summary

This PR adds multi-provider image generation (OpenAI GPT Image, Google Nano Banana via Gemini, Fal.ai) through a new image_generator_v2 block and POST /api/tools/image endpoint, and expands the Fal.ai video generation backend with a dozen new models (Veo 3.1 Fast, Sora 2/Pro, Seedance 2.0, Kling v3/o3, LTX 2.3, WAN 2.2) via a config-driven FALAI_MODEL_CONFIGS map.

• Image route: replaces the old per-provider ad-hoc dispatch with structured generateWithOpenAI, generateWithGemini, and generateWithFalAI helpers; SSRF protection via validateUrlWithDNS/secureFetchWithPinnedIP is correctly applied in bufferFromImageUrl; CANCELLED status is handled in the Fal.ai polling loop.
• Video route: migrates model-to-endpoint mapping to FALAI_MODEL_CONFIGS, fixes the minimax-hailuo-2.3 endpoint (was incorrectly pointing at hailuo-02), adds generateAudio support, and fixes CANCELLED polling exit.
• Blocks/tools: adds ImageGeneratorV2Block, imageGenerateTool, and updates falaiVideoTool/minimaxVideoTool descriptions to reflect new model availability.

Confidence Score: 5/5

Safe to merge; all major backend changes are additive, SSRF guards are intact, and the previous polling/status bugs are fixed.

The image and video routes are well-structured with proper input validation, SSRF protection, and terminal-status handling. The config-driven model dispatch replaces fragile hardcoded maps. The two non-blocking nits (duplicate Gemini/Fal.ai model labels, missing duration validation for two legacy Kling configs) have no correctness impact at runtime.

apps/sim/app/api/tools/video/route.ts — the two legacy Kling configs (kling-2.5-turbo-pro, kling-2.1-pro) lack duration option lists; apps/sim/blocks/blocks/image_generator.ts — Gemini model labels duplicate Fal.ai labels.

Important Files Changed

Filename	Overview
apps/sim/app/api/tools/image/route.ts	Adds a new POST endpoint for multi-provider image generation (OpenAI, Gemini, Fal.ai); includes proper SSRF guards via validateUrlWithDNS/secureFetchWithPinnedIP in bufferFromImageUrl, CANCELLED status handling, and config-driven model validation.
apps/sim/app/api/tools/video/route.ts	Refactors Fal.ai video generation to a config-driven FALAI_MODEL_CONFIGS map; adds new models (Veo 3.1 Fast, Sora 2 Pro, Seedance 2.0, Kling v3/o3, LTX 2.3, WAN 2.2); fixes MiniMax hailuo-2.3 endpoint; now handles CANCELLED status. Two legacy Kling model configs lack durationOptions validation.
apps/sim/blocks/blocks/image_generator.ts	Adds ImageGeneratorV2Block with multi-provider support; Gemini model labels (Nano Banana 2/Pro/Nano Banana) duplicate Fal.ai model labels, which could confuse users switching providers.
apps/sim/blocks/blocks/video_generator.ts	Adds comprehensive sub-block UI for the many new Fal.ai video models with model-specific duration, aspect ratio, resolution, and audio dropdowns; introduces FALAI_LATEST_MODEL_OPTIONS alongside FALAI_PREVIOUS_MODEL_OPTIONS for versioned model selection.
apps/sim/tools/image/generate.ts	New tool definition for image_generate; parameters use correct user-only/user-or-llm visibility; clean transform response with graceful error handling.
apps/sim/tools/video/falai.ts	Updated model descriptions and added generateAudio param to tool config; matches the new route capabilities cleanly.
apps/sim/tools/video/minimax.ts	Updated default model to hailuo-2.3 and added endpoint parameter with standard/pro description; matches route changes.

Sequence Diagram

sequenceDiagram
    participant Client
    participant ImageRoute as POST /api/tools/image
    participant OpenAI as OpenAI API
    participant Gemini as Gemini API
    participant FalAI as Fal.ai Queue

    Client->>ImageRoute: "{provider, apiKey, model, prompt, ...}"
    ImageRoute->>ImageRoute: parseRequest + checkInternalAuth
    alt "provider = openai"
        ImageRoute->>OpenAI: POST /v1/images/generations
        OpenAI-->>ImageRoute: b64_json or url
        ImageRoute->>ImageRoute: bufferFromImageUrl (SSRF-safe)
    else "provider = gemini"
        ImageRoute->>Gemini: POST generateContent
        Gemini-->>ImageRoute: inlineData (base64)
    else "provider = falai"
        ImageRoute->>FalAI: "POST queue.fal.run/{endpoint}"
        FalAI-->>ImageRoute: "{request_id, status_url, response_url}"
        loop Poll every 3s
            ImageRoute->>FalAI: GET status_url
            FalAI-->>ImageRoute: "{status: QUEUED|IN_PROGRESS|COMPLETED|FAILED|CANCELLED}"
        end
        ImageRoute->>FalAI: GET response_url
        FalAI-->>ImageRoute: "{images: [{url, content_type}]}"
        ImageRoute->>ImageRoute: bufferFromImageUrl (SSRF-safe)
    end
    ImageRoute->>ImageRoute: storeGeneratedImage (execution or copilot)
    ImageRoute-->>Client: "{imageUrl, imageFile, metadata}"

Loading

_{Reviews (3): Last reviewed commit: "address api mismatches" | Re-trigger Greptile}

[icecrasher321]

@greptile

[icecrasher321]

bugbot run

[icecrasher321]

bugbot run

[icecrasher321]

@greptile

[icecrasher321]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 54d2e1f. Configure here.}

[icecrasher321]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 71b4080. Configure here.}