[ACTION] Extend Vertex AI MCP server with Veo3 support #17766
Conversation
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎ |
WalkthroughThis update introduces new actions for generating videos using Google Vertex AI's Veo models, including support for both text and image prompts. It adds a shared module for video generation logic, extends the app definition with new model options and API methods, and increments version numbers for several existing actions and the package. Changes
Sequence Diagram(s)sequenceDiagram
participant User
participant Action (Text/Image)
participant Common Generate Video
participant App (google_vertex_ai)
participant Vertex AI API
User->>Action (Text/Image): Provide prompt (and image)
Action (Text/Image)->>Common Generate Video: Prepare parameters
Common Generate Video->>App (google_vertex_ai): generateVideosLongRunning()
App (google_vertex_ai)->>Vertex AI API: POST /predict:predictLongRunning
Vertex AI API-->>App (google_vertex_ai): Return operation name
App (google_vertex_ai)-->>Common Generate Video: Operation started
loop Poll until done
Common Generate Video->>App (google_vertex_ai): fetchOperation()
App (google_vertex_ai)->>Vertex AI API: POST /operations/predict
Vertex AI API-->>App (google_vertex_ai): Operation status/result
App (google_vertex_ai)-->>Common Generate Video: Operation status/result
end
Common Generate Video-->>Action (Text/Image): Final result
Action (Text/Image)-->>User: Return summary and result
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~15–20 minutes Poem
Warning There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure. 🔧 ESLint
components/google_vertex_ai/actions/analyze-image-video/analyze-image-video.mjsOops! Something went wrong! :( ESLint: 8.57.1 Error [ERR_MODULE_NOT_FOUND]: Cannot find package 'jsonc-eslint-parser' imported from /eslint.config.mjs components/google_vertex_ai/actions/analyze-text-sentiment/analyze-text-sentiment.mjsOops! Something went wrong! :( ESLint: 8.57.1 Error [ERR_MODULE_NOT_FOUND]: Cannot find package 'jsonc-eslint-parser' imported from /eslint.config.mjs components/google_vertex_ai/actions/common/generate-video.mjsOops! Something went wrong! :( ESLint: 8.57.1 Error [ERR_MODULE_NOT_FOUND]: Cannot find package 'jsonc-eslint-parser' imported from /eslint.config.mjs
Note ⚡️ Unit Test Generation - BetaCodeRabbit's unit test generation is now available in Beta! Automatically generate comprehensive unit tests for your code changes, ensuring better test coverage and catching edge cases you might miss. Our AI analyzes your code structure and creates tests that follow best practices and your project's testing patterns. Learn more here, or just try it under ✨ Finishing Touches. ✨ Finishing Touches
🧪 Generate unit tests
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
CodeRabbit Configuration File (
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
components/google_vertex_ai/google_vertex_ai.app.mjs (1)
36-55: Consider using a stable model as the default.The model options are well-structured, but the default is set to a preview version (
veo-3.0-generate-preview). Consider using the stable version (veo-2.0-generate-001) as the default to ensure better reliability for users.- default: "veo-3.0-generate-preview", + default: "veo-2.0-generate-001",
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
⛔ Files ignored due to path filters (1)
pnpm-lock.yamlis excluded by!**/pnpm-lock.yaml
📒 Files selected for processing (8)
components/google_vertex_ai/actions/analyze-image-video/analyze-image-video.mjs(1 hunks)components/google_vertex_ai/actions/analyze-text-sentiment/analyze-text-sentiment.mjs(1 hunks)components/google_vertex_ai/actions/classify-text/classify-text.mjs(1 hunks)components/google_vertex_ai/actions/common/generate-video.mjs(1 hunks)components/google_vertex_ai/actions/generate-video-from-image/generate-video-from-image.mjs(1 hunks)components/google_vertex_ai/actions/generate-video-from-text/generate-video-from-text.mjs(1 hunks)components/google_vertex_ai/google_vertex_ai.app.mjs(2 hunks)components/google_vertex_ai/package.json(2 hunks)
🧰 Additional context used
🧠 Learnings (2)
components/google_vertex_ai/package.json (1)
Learnt from: jcortes
PR: #14935
File: components/sailpoint/package.json:15-18
Timestamp: 2024-12-12T19:23:09.039Z
Learning: When developing Pipedream components, do not add built-in Node.js modules like fs to package.json dependencies, as they are native modules provided by the Node.js runtime.
components/google_vertex_ai/actions/generate-video-from-image/generate-video-from-image.mjs (1)
Learnt from: js07
PR: #17375
File: components/tinypng/actions/compress-image/compress-image.mjs:18-23
Timestamp: 2025-07-01T17:01:46.327Z
Learning: In TinyPNG compress-image action (components/tinypng/actions/compress-image/compress-image.mjs), the syncDir property uses accessMode: "read" because this action only reads input files and returns API responses without writing files to /tmp, unlike other TinyPNG actions that save processed files to disk.
🧬 Code Graph Analysis (2)
components/google_vertex_ai/actions/generate-video-from-text/generate-video-from-text.mjs (1)
components/google_vertex_ai/actions/generate-video-from-image/generate-video-from-image.mjs (3)
operationResponse(41-67)completedOperation(74-79)videoCount(89-89)
components/google_vertex_ai/actions/generate-video-from-image/generate-video-from-image.mjs (1)
components/google_vertex_ai/actions/generate-video-from-text/generate-video-from-text.mjs (3)
operationResponse(26-48)completedOperation(55-60)videoCount(70-70)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (4)
- GitHub Check: Verify TypeScript components
- GitHub Check: Publish TypeScript components
- GitHub Check: pnpm publish
- GitHub Check: Lint Code Base
🔇 Additional comments (11)
components/google_vertex_ai/package.json (1)
3-3: LGTM! Version updates are appropriate for new feature additions.The minor version bump (0.1.0 → 0.2.0) correctly reflects the addition of new video generation capabilities, and the platform dependency update aligns with the enhanced functionality.
Also applies to: 16-16
components/google_vertex_ai/actions/classify-text/classify-text.mjs (1)
7-7: LGTM! Coordinated version update for consistency.The patch version bump maintains consistency across the Google Vertex AI component ecosystem as part of the release that introduces video generation capabilities.
components/google_vertex_ai/actions/analyze-image-video/analyze-image-video.mjs (1)
7-7: LGTM! Consistent version management across actions.The version update maintains consistency with other actions in the component without introducing unnecessary changes to existing functionality.
components/google_vertex_ai/actions/analyze-text-sentiment/analyze-text-sentiment.mjs (1)
7-7: LGTM! Appropriate version synchronization.The version update aligns with the coordinated release strategy for the Google Vertex AI component enhancements.
components/google_vertex_ai/actions/generate-video-from-image/generate-video-from-image.mjs (3)
1-2: LGTM! Imports are appropriate for the functionality.The imports correctly include the file handling utility and the shared video generation logic.
4-18: Well-structured action metadata.The action properly extends the common module and provides clear documentation. The image property description helpfully specifies supported formats and quality recommendations.
19-95: Solid implementation following established patterns.The run method correctly handles image file processing, converts to base64, and follows the same async operation pattern as the text-based video generation. Error handling is comprehensive and the summary message appropriately indicates the source type.
components/google_vertex_ai/actions/common/generate-video.mjs (3)
3-113: Comprehensive and well-documented props definition.The properties are thoughtfully defined with clear descriptions, appropriate defaults, and helpful model-specific notes (Veo 2 vs Veo 3). The validation constraints and optional flags are properly set.
115-147: Robust polling implementation with appropriate timeouts.The polling logic uses reasonable intervals (45 seconds) and timeout (6 attempts) for video generation operations. Error handling and progress logging are well-implemented.
148-159: Standard and correct stream-to-base64 implementation.The method properly handles stream events, efficiently concatenates chunks, and converts to base64. Error handling is appropriate.
components/google_vertex_ai/actions/generate-video-from-text/generate-video-from-text.mjs (1)
1-77: Clean implementation following established patterns.This text-based video generation action correctly implements the same async operation pattern as the image version, with appropriate simplifications (no file handling required). The error handling, polling logic, and response processing are consistent and robust.
WHY
Resolves #17363
Summary by CodeRabbit
New Features
Chores