Use public Blob URLs for large source ingestion

[suguanYang]

Summary

• switch large source staging uploads back to the current public Vercel Blob store so the browser multipart upload matches the deployed Blob token/store access
• keep staged Blob paths unguessable with per-upload UUID paths plus Vercel's random suffix, and avoid original file names in the Blob pathname
• send the public Blob URL to Knowhere as a URL parse job instead of downloading/re-uploading through the Next.js server
• persist staged Blob metadata on parsing sources, then delete and clear it once reconciliation sees the Knowhere job complete or fail
• add immediate best-effort Blob cleanup when URL job creation or metadata handoff fails

Verification

• pnpm test
• pnpm exec tsc --noEmit
• pnpm lint
• pnpm build
• git diff --check

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
knowhere-notebook-staging	Ready	Preview, Comment	May 10, 2026 11:23am