feat: integrate Google Drive ingestion, selection and Source history (MEM-76)

[nhuxuanviet]

Scope

Integrate the previously committed Google Drive work and Source follow-ups into current main, preserving the logical commit history. Includes reusable Tenant-owned OAuth credentials, immutable General/Specific scopes, independent schedules, native Sheets/Docs and bounded binary/table extraction, linked-document discovery/explicit approvals, durable asynchronous selection validation, exact acquisition-to-indexing run history, paged selection/Files/history, shared pagination and stale-Source action guards. Preserve the existing Docling settings/15-minute processing-budget commit.

Main Chat/JIT and Search PRs #83/#87 are already integrated through 3f236d5; Source conflict resolution preserves the feature HEAD. Generated OpenAPI and browser clients are reconciled. Integration-only Search retry and Chat HTTP/SSE fixture corrections preserve production behavior and assertions.

Related: MEM-76, scoped implementation under MEM-9/MEM-10/MEM-60/MEM-63. This is not closure of broader Google ACL/reader, live-provider or customer-data acceptance.

Local verification

• Full uncached, serial clean check :api:bootJar :worker:bootJar --continue: BUILD SUCCESSFUL in 10m18s. 495 cases: 491 passed, 4 conditional skips, zero failures/errors. Skips: one optional live Chat model and three real Docling cases. Build outputs isolated from running review artifacts.
• pnpm check: 93 unit tests in 18 files plus generated-contract/route drift, CI consistency, lint/format/TypeScript/build/fonts.
• Full Playwright: 54/54 passed, one worker, no retries, 4.3m.
• Rendered read-only desktop/mobile Source, Chat and latest Search smoke: exact paging/keyboard behavior, setup navigation, literal highlights and generation-pinned preview. Synthetic HTTP proof, not live Google/model acceptance.
• OpenSearch infrastructure: seven passed, two Linux-only cases skipped locally on Windows; Linux CI remains required.
• Latest commit only records user authorization/status; it does not alter verified code.

Canonical verification · Active plan

Migration and operational boundaries

• Main V1–V20 remains byte-identical. Twelve unpublished Drive migrations move V18–V29 → V21–V32 with unchanged SQL bodies.
• Never apply this layout to existing memoryos_main_review or memoryos_drive_review histories. No Flyway repair/history rewrite/reset, database cutover or deployment is included.
• No changes to retained Sources, credentials, roots, approvals, schedules or Google content; no reindex/extraction retry. Google remains RESTRICTED and does not inherit FILE PUBLIC read eligibility.
• Known pre-existing pnpm audit: five high/five moderate development-tool transitive advisories (fast-uri, js-yaml); separate default-branch Dependabot report: nine high/two moderate. Existing Chat bundle-size warning retained; no security-clean or IDE-clean claim.

Merge gate

User authorized merge-when-green. Require latest-head CI, one complete CodeRabbit review pass with actionable findings resolved, fresh base and exact-head guard. Merge commit only; do not squash/rebase. No staging deployment. Preserve broader provider/ACL/customer-data acceptance as active.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Advanced

Run ID: 20a41e3d-8630-4c5b-aac6-8489003b0d6f

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[nhuxuanviet]

@coderabbitai full review

[coderabbitai]

⚠️ Action not completed

Review skipped: 286 files exceed the limit of 100.

[nhuxuanviet]

Single CodeRabbit pass captured at 48fc186: bot explicitly reported 286 eligible files against a 100-file quota and insufficient usage credits. Both bot comments, submitted reviews, paginated inline comments and all review threads were collected once: no actionable findings, no submitted review, zero threads. This is missing review coverage, not CodeRabbit approval. No second request or evidence polling will be made. The repository PR-loop quota fallback remains conditional on all latest-head CI checks green, a fresh base and exact-head guarded merge under user authorization. Fix round 1 (e46caa2) only adds the exact historical MEM-66 table-prose fingerprint to existing .gitleaksignore and records evidence. Full remote-history Gitleaks 8.30.1 changed from one reproduced false positive to zero; no broad exclusion or workflow weakening. Waiting for fresh CI before any merge.

[nhuxuanviet]

Merged with exact-head guard: e46caa2 -> merge commit 274bc78. Git ancestry and identical merge/feature trees verified. Latest-head CI: https://github.com/kl3inIT/MemoryOS/actions/runs/34378139200 . Exact merge-SHA main CI: https://github.com/kl3inIT/MemoryOS/actions/runs/34378905200 (all seven jobs success, including image publication). Deploy staging run https://github.com/kl3inIT/MemoryOS/actions/runs/34379631781 was skipped. No runtime cutover. MEM-76 evidence updated; broader live-provider/ACL acceptance remains active. Review-runtime worktree/build artifacts intentionally retained; owned temporary scanner removed. CodeRabbit quota fallback evidence is recorded above, not claimed as reviewer approval.