Repository navigation
Assessment: Support video attachments #1212
Description
Is your feature request related to a problem?
The current assessment batch only supports image and pdf attachments. With Google Gemini supporting video inputs, we need functionality for video attachments, including proper validation and handling to avoid errors with unsupported formats.
Describe the solution you'd like
- Add
videotoInputColumn.typeandAssessmentAttachment.type/type_value_map(url-format only). - Implement type detection for common video extensions (
.mp4,.mov,.webm, etc.). - Build
fileDatavideo part for Gemini (support for YouTube URLs and direct video URL mime types). - Configure
video_part_config(defaultvideoMetadata.fps) andmedia_resolutionthrough Google params mapper. - Ensure OpenAI/Anthropic builders skip unsupported video cells instead of throwing errors.
Original issue
Summary
Assessment BATCH input columns only supported image/pdf attachment types. Google Gemini supports video (including YouTube URLs), so we need a video input-column type end-to-end: schema validation, attachment routing/mime resolution, and Gemini request building — while OpenAI/Anthropic (which don't support video input) safely skip such cells instead of erroring.
Scope
- Add
videotoInputColumn.typeandAssessmentAttachment.type/type_value_map(url-format only, base64 rejected). - Extension→type detection for common video extensions (
.mp4,.mov,.webm, etc.). - Gemini
fileDatavideo part builder (YouTube URLs carry nomimeType; direct GCS/HTTPS URLs get a resolved video mime). video_part_config(defaultvideoMetadata.fps) andmedia_resolution(generationConfig.mediaResolution) defaults wired through the Google params mapper.- OpenAI/Anthropic attachment builders skip video cells (unsupported) rather than throwing.
Out of scope
- OpenAI/Anthropic native video support (not available upstream).
Activity
[-]feat(assessment): Support video attachments in batch processing and validation[/-][+]Assessment: Support video attachments[/+]github-actions commentedon Sep 29, 2026
🎉 This issue has been resolved in version 1.7.0-main.12 🎉
The release is available on GitHub release
Your semantic-release bot 📦🚀