Fix broken multimodal analysis flows (video, image, and URL)

[hrx01-dev]

Problem

Currently text analysis is the only analysis flow that reliably returns a result. Video analysis can remain stuck on Processing/Analyzing for a long time, while image, URL, and video-link flows may fail or provide poor feedback.

Objective

Make every advertised analysis mode reliably reach a result or a clear actionable error instead of hanging indefinitely.

Scope

• Audit and fix frontend requests for image, video, video URL, and URL analysis.
• Add request timeouts/abort handling so the UI cannot remain in a loading state indefinitely.
• Improve loading progress/status messaging for long-running video inference.
• Surface backend errors clearly to users.
• Validate file types and reasonable file sizes before upload.
• Verify the production backend routes and runtime configuration for all analysis modes.
• Ensure analysis results are normalized consistently for the existing result UI and history.
• Keep text analysis working unchanged.
• Avoid silently masking backend failures with fake/simulated results.

Acceptance criteria

• Text analysis continues working.
• Image analysis returns a result or actionable error.
• Video file analysis returns a result or actionable error without indefinite loading.
• Video URL analysis returns a result or actionable error.
• URL analysis returns a result or actionable error.
• Requests have bounded client-side waiting with cancellation/timeout handling.
• Loading states reset correctly after success, failure, or cancellation.
• Backend/runtime errors are visible to the user.
• No fake results are generated to hide failures.
• Existing authentication and history behavior remains intact.

[vercel]

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

Project	Deployment	Actions	Updated
veritas-ai	Error		Sep 4, 2026 3:23pm UTC