Support v2 page-memory chunks

[suguanYang]

Summary

• create Knowhere clients with API v2 so notebook uploads use the page-memory ingestion API when the SDK supports it
• normalize page chunks with contentSource, readable page-summary content, metadata.entities, and page-number ranges
• update parsed chunk cards and chunks panel display so page chunks render as page summaries and page assetUrl values are not treated as images
• add tests covering normalization, model labels/ranges, card rendering, and client construction

Notes

• Runtime v2 constructor support depends on the Node SDK PR Add v2 page-memory schema support knowhere-node-sdk#108 being released as a patch version.
• The API-side schema work is in Add v2 page-memory ingestion API knowhere#196.

Verification

• pnpm install --frozen-lockfile --ignore-scripts
• pnpm run typecheck
• pnpm run lint
• pnpm test (588 passed, 14 skipped; existing jsdom navigation warning still appears)
• pnpm run build (passed; existing Next lockfile SWC patch and QStash token warnings still appear, generated lockfile patch not committed)
• 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	Jul 2, 2026 7:19am

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
knowhere-notebook	Ignored	Preview	Jul 2, 2026 7:19am

[suguanYang]

Follow-up update for the v2-only direction:

• Notebook now depends on @ontos-ai/knowhere-sdk@^2.0.0.
• Removed the apiVersion: "v2" constructor shim; Notebook relies on the SDK v2-only client behavior.

Local validation:

• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/eslint
• ./node_modules/.bin/vitest run (588 passed, 14 skipped)
• ./node_modules/.bin/next build
• git diff --check

Remote status on commit 7fe9026: CodeQL is green, but Lint, Typecheck, Test, Build and the staging Vercel deployment fail at dependency install because package.json now requires @ontos-ai/knowhere-sdk@^2.0.0 while the committed lockfile still resolves ^0.10.0. I did not refresh the lockfile because @ontos-ai/knowhere-sdk@2.0.0 is not published yet; after the SDK PR publishes 2.0.0, the lockfiles need to be refreshed and this PR re-run.