GeezerVison

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

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview, Comment	May 4, 2026 8:51pm

PR Summary

Medium Risk
Moderate risk because it expands the image generation parameter surface and request shaping for a new OpenAI model, and flips a block’s authType from none to api-key, which could affect access/config in production.

Overview
Adds gpt-image-2 support to the Image Generator block and openai_image tool, including new UI options and request params (quality, background, outputFormat, moderation) plus model-specific validation/defaulting for size and other fields.

Updates docs and integration metadata to reflect the new image model capabilities, refreshes Knowledge tool docs with new reranker inputs, updates Mem0 docs to match new async/paginated response fields, and adds a new table trigger docs page and entry in triggers meta.json. Also changes one integration’s authType in integrations.json from none to api-key.

^{Reviewed by Cursor Bugbot for commit 4073f0d. Configure here.}

Greptile Summary

Adds gpt-image-2 as a third model option in the Image Generator block, wiring up its dedicated size options (up to 3840×2160), quality levels, background (opaque/auto only), output format, and moderation parameters. The implementation correctly validates size, quality, and background via ALLOWED_* allowlists in the block params function, maps camelCase outputFormat → snake_case output_format in the tool, and shares the quality/outputFormat/moderation path with gpt-image-1. The PR also includes several unrelated documentation changes (knowledge, mem0, table trigger docs) and a authType: "none" → "api-key" fix for the knowledge integration entry in integrations.json.

Confidence Score: 5/5

Safe to merge — no P0 or P1 issues found in the feature implementation.

The core feature code is correct: new params are validated against allowlists, the camelCase → snake_case mapping is handled in the right layer, the response handler already supports b64_json for gpt-image-2, and dall-e-3 is not affected. The one open comment is a P2 documentation nit on an unrelated table trigger doc file.

No files require special attention; the documentation in apps/docs/content/docs/en/triggers/table.mdx has a minor duplicate description worth cleaning up.

Important Files Changed

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Image Generator Block] --> B{Model Selection}
    B -->|dall-e-3| C[Size: 1024x1024 / 1024x1792 / 1792x1024\nQuality: standard / hd\nStyle: vivid / natural]
    B -->|gpt-image-1| D[Size: auto / 1024x1024 / 1536x1024 / 1024x1536\nQuality: auto / low / medium / high\nBackground: auto / transparent / opaque\nOutputFormat: png / jpeg / webp\nModeration: auto / low]
    B -->|gpt-image-2 NEW| E[Size: auto / 1024x1024 / 1536x1024 / 1024x1536 / 2560x1440 / 3840x2160\nQuality: auto / low / medium / high\nBackground: auto / opaque\nOutputFormat: png / jpeg / webp\nModeration: auto / low]
    C --> F[ALLOWED_* Validation in block params]
    D --> F
    E --> F
    F --> G[imageTool POST /v1/images/generations]
    G -->|dall-e-3| H[url response → proxy fetch → base64]
    G -->|gpt-image-1 / gpt-image-2| I[b64_json response → base64]
    H --> J[Output: image base64 + metadata]
    I --> J

_{Reviews (4): Last reviewed commit: "docs" | Re-trigger Greptile}

@greptile

@cursor review

@greptile

@cursor review

@greptileai please re-review — all prior comments resolved.

@cursor review