fix(rag): content-hash chunk ids — dedup identical uploads per course

[Darkest-Teddy]

Problem

course_chunks ids were sha256(doc_id::index::text), so identical content uploaded by different students (or the same doc re-uploaded) stored a full copy of every chunk and its embedding per upload. A popular course with shared slides multiplies chunk storage by enrollment count.

Fix

• rag_service.chunk_id (new): ids are now sha256(course_id::chunk_text) — content-addressed, scoped per course. Identical text in the same course merges to one row regardless of document or uploader; the same text in different courses stays separate.
• doc_id / uploader_id / chunk_index become last-writer-wins metadata. Safe because nothing reads, filters, or deletes by them: the match_course_chunks RPC returns only course_id/chunk_text/similarity, and no code path deletes course_chunks by doc_id.
• In-batch dedup in index_document_chunks: a document repeating the same text (boilerplate headers) would otherwise put duplicate keys in one upsert payload, which Postgres rejects (ON CONFLICT DO UPDATE cannot affect row a second time).
• scripts/dedupe_course_chunks.py: one-time, idempotent, dry-run-by-default migration for existing rows — groups legacy rows by content-hash id, keeps a row that already has an embedding (no re-embedding cost), upserts winners before deleting stale ids so an interrupted run never drops a chunk. Catalog rows (category=catalog, own id scheme via ingest_catalog.py) are untouched.

Not covered

• Cross-course dedup (same textbook chapter in two courses) — intentional: rows carry per-course metadata, and retrieval filters by course.
• backfill_document_chunks.py's _already_indexed check keys on doc_id; under dedup a doc whose chunks were claimed by another doc's rows may re-index and flip doc_id — harmless churn, ids are stable.

Testing

• 3 new tests in test_rag_service.py (same text + course collides, different courses don't, in-batch dedup) and 4 in test_dedupe_course_chunks.py (collapse, embedding-preferring winner, idempotency, cross-course separation) — all written first and watched fail (TDD).
• Full backend suite: 963 passed, 1 skipped; the test_ocr_pipeline.py::test_save_to_db error is the pre-existing live-Supabase/event-loop artifact documented in ADR 0016 (reproduces identically with the change stashed).
• ruff check clean on all touched files.

Deploy note

After merge, run python scripts/dedupe_course_chunks.py (dry-run), review counts, then --apply — once per environment (dotenv -f .env.staging run -- ... for staging).

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Improved document indexing by identifying identical content within each course and avoiding duplicate entries.
  • Preserved separate entries for identical content across different courses.
  • Added a migration utility to consolidate existing duplicate document chunks, with a preview mode before changes are applied.

• Bug Fixes

  • Prevented repeated indexing of the same content from creating redundant records.
  • Preserved existing embeddings when consolidating duplicate chunks.

• Tests

  • Added coverage for deduplication, course scoping, migration behavior, and idempotent indexing.

[coderabbitai]

Warning

Review limit reached

@AndresL230, you've reached your PR review limit, so we couldn't start this review.

Next review available in:19 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: fdb9d637-c8a5-4a79-9e03-8a256bcda721

📥 Commits

Reviewing files that changed from the base of the PR and between dffac9b and b810f72.

📒 Files selected for processing (4)

• backend/scripts/dedupe_course_chunks.py
• backend/services/rag_service.py
• backend/tests/test_dedupe_course_chunks.py
• backend/tests/test_rag_service.py

📝 Walkthrough

Walkthrough

Course chunk indexing now uses course-scoped content hashes and deduplicates repeated content. A migration script plans and applies rewrites of legacy document rows, while tests cover deduplication, embedding preservation, idempotency, and course isolation.

Changes

Course chunk identifiers

Layer / File(s)	Summary
Content-hash indexing and validation backend/services/rag_service.py, backend/tests/test_rag_service.py	Adds course-scoped SHA-256 chunk IDs, deduplicates indexing payloads, and tests stable IDs, course isolation, and unique upsert counts.
Legacy row migration backend/scripts/dedupe_course_chunks.py, backend/tests/test_dedupe_course_chunks.py	Adds paginated document-row fetching, winner-based migration planning, dry-run and batched apply modes, and tests for duplicate collapse, embedding reuse, idempotency, and cross-course separation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
participant CLI as dedupe_course_chunks
participant Database as course_chunks
participant Planner as plan_migration
CLI->>Database: Fetch paginated document rows
CLI->>Planner: Group rows by course-scoped chunk ID
Planner-->>CLI: Return winner upserts and stale IDs
CLI->>Database: Upsert rewritten rows
CLI->>Database: Delete duplicate legacy rows

Loading

---

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.}

[cloudflare-workers-and-pages]

Deploying with Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	frontend-staging	b810f72	Commit Preview URL Branch Preview URL	Jul 29 2026, 09:25 AM

[Darkest-Teddy]

Problem

course_chunks ids were sha256(doc_id::index::text), so identical content uploaded by different students (or the same doc re-uploaded) stored a full copy of every chunk and its embedding per upload. A popular course with shared slides multiplies chunk storage by enrollment count.

Fix

• rag_service.chunk_id (new): ids are now sha256(course_id::chunk_text) — content-addressed, scoped per course. Identical text in the same course merges to one row regardless of document or uploader; the same text in different courses stays separate.
• doc_id / uploader_id / chunk_index become last-writer-wins metadata. Safe because nothing reads, filters, or deletes by them: the match_course_chunks RPC returns only course_id/chunk_text/similarity, and no code path deletes course_chunks by doc_id.
• In-batch dedup in index_document_chunks: a document repeating the same text (boilerplate headers) would otherwise put duplicate keys in one upsert payload, which Postgres rejects (ON CONFLICT DO UPDATE cannot affect row a second time).
• scripts/dedupe_course_chunks.py: one-time, idempotent, dry-run-by-default migration for existing rows — groups legacy rows by content-hash id, keeps a row that already has an embedding (no re-embedding cost), upserts winners before deleting stale ids so an interrupted run never drops a chunk. Catalog rows (category=catalog, own id scheme via ingest_catalog.py) are untouched.

Not covered

• Cross-course dedup (same textbook chapter in two courses) — intentional: rows carry per-course metadata, and retrieval filters by course.
• backfill_document_chunks.py's _already_indexed check keys on doc_id; under dedup a doc whose chunks were claimed by another doc's rows may re-index and flip doc_id — harmless churn, ids are stable.

Testing

• 3 new tests in test_rag_service.py (same text + course collides, different courses don't, in-batch dedup) and 4 in test_dedupe_course_chunks.py (collapse, embedding-preferring winner, idempotency, cross-course separation) — all written first and watched fail (TDD).
• Full backend suite: 963 passed, 1 skipped; the test_ocr_pipeline.py::test_save_to_db error is the pre-existing live-Supabase/event-loop artifact documented in ADR 0016 (reproduces identically with the change stashed).
• ruff check clean on all touched files.

Deploy note

After merge, run python scripts/dedupe_course_chunks.py (dry-run), review counts, then --apply — once per environment (dotenv -f .env.staging run -- ... for staging).

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Improved document indexing by identifying identical content within each course and avoiding duplicate entries.
  • Preserved separate entries for identical content across different courses.
  • Added a migration utility to consolidate existing duplicate document chunks, with a preview mode before changes are applied.

• Bug Fixes

  • Prevented repeated indexing of the same content from creating redundant records.
  • Preserved existing embeddings when consolidating duplicate chunks.

• Tests

  • Added coverage for deduplication, course scoping, migration behavior, and idempotent indexing.

[coderabbitai]

Warning

Review limit reached

@AndresL230, you've reached your PR review limit, so we couldn't start this review.

Next review available in:19 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: fdb9d637-c8a5-4a79-9e03-8a256bcda721

📥 Commits

Reviewing files that changed from the base of the PR and between dffac9b and b810f72.

📒 Files selected for processing (4)

• backend/scripts/dedupe_course_chunks.py
• backend/services/rag_service.py
• backend/tests/test_dedupe_course_chunks.py
• backend/tests/test_rag_service.py

📝 Walkthrough

Walkthrough

Course chunk indexing now uses course-scoped content hashes and deduplicates repeated content. A migration script plans and applies rewrites of legacy document rows, while tests cover deduplication, embedding preservation, idempotency, and course isolation.

Changes

Course chunk identifiers

Layer / File(s)	Summary
Content-hash indexing and validation backend/services/rag_service.py, backend/tests/test_rag_service.py	Adds course-scoped SHA-256 chunk IDs, deduplicates indexing payloads, and tests stable IDs, course isolation, and unique upsert counts.
Legacy row migration backend/scripts/dedupe_course_chunks.py, backend/tests/test_dedupe_course_chunks.py	Adds paginated document-row fetching, winner-based migration planning, dry-run and batched apply modes, and tests for duplicate collapse, embedding reuse, idempotency, and cross-course separation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
participant CLI as dedupe_course_chunks
participant Database as course_chunks
participant Planner as plan_migration
CLI->>Database: Fetch paginated document rows
CLI->>Planner: Group rows by course-scoped chunk ID
Planner-->>CLI: Return winner upserts and stale IDs
CLI->>Database: Upsert rewritten rows
CLI->>Database: Delete duplicate legacy rows

Loading

---

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.}

[cloudflare-workers-and-pages]

Deploying with Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	frontend-staging	b810f72	Commit Preview URL Branch Preview URL	Jul 29 2026, 09:25 AM

[Darkest-Teddy]

Problem

course_chunks ids were sha256(doc_id::index::text), so identical content uploaded by different students (or the same doc re-uploaded) stored a full copy of every chunk and its embedding per upload. A popular course with shared slides multiplies chunk storage by enrollment count.

Fix

• rag_service.chunk_id (new): ids are now sha256(course_id::chunk_text) — content-addressed, scoped per course. Identical text in the same course merges to one row regardless of document or uploader; the same text in different courses stays separate.
• doc_id / uploader_id / chunk_index become last-writer-wins metadata. Safe because nothing reads, filters, or deletes by them: the match_course_chunks RPC returns only course_id/chunk_text/similarity, and no code path deletes course_chunks by doc_id.
• In-batch dedup in index_document_chunks: a document repeating the same text (boilerplate headers) would otherwise put duplicate keys in one upsert payload, which Postgres rejects (ON CONFLICT DO UPDATE cannot affect row a second time).
• scripts/dedupe_course_chunks.py: one-time, idempotent, dry-run-by-default migration for existing rows — groups legacy rows by content-hash id, keeps a row that already has an embedding (no re-embedding cost), upserts winners before deleting stale ids so an interrupted run never drops a chunk. Catalog rows (category=catalog, own id scheme via ingest_catalog.py) are untouched.

Not covered

• Cross-course dedup (same textbook chapter in two courses) — intentional: rows carry per-course metadata, and retrieval filters by course.
• backfill_document_chunks.py's _already_indexed check keys on doc_id; under dedup a doc whose chunks were claimed by another doc's rows may re-index and flip doc_id — harmless churn, ids are stable.

Testing

• 3 new tests in test_rag_service.py (same text + course collides, different courses don't, in-batch dedup) and 4 in test_dedupe_course_chunks.py (collapse, embedding-preferring winner, idempotency, cross-course separation) — all written first and watched fail (TDD).
• Full backend suite: 963 passed, 1 skipped; the test_ocr_pipeline.py::test_save_to_db error is the pre-existing live-Supabase/event-loop artifact documented in ADR 0016 (reproduces identically with the change stashed).
• ruff check clean on all touched files.

Deploy note

After merge, run python scripts/dedupe_course_chunks.py (dry-run), review counts, then --apply — once per environment (dotenv -f .env.staging run -- ... for staging).

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Improved document indexing by identifying identical content within each course and avoiding duplicate entries.
  • Preserved separate entries for identical content across different courses.
  • Added a migration utility to consolidate existing duplicate document chunks, with a preview mode before changes are applied.

• Bug Fixes

  • Prevented repeated indexing of the same content from creating redundant records.
  • Preserved existing embeddings when consolidating duplicate chunks.

• Tests

  • Added coverage for deduplication, course scoping, migration behavior, and idempotent indexing.

[coderabbitai]

Warning

Review limit reached

@AndresL230, you've reached your PR review limit, so we couldn't start this review.

Next review available in:19 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: fdb9d637-c8a5-4a79-9e03-8a256bcda721

📥 Commits

Reviewing files that changed from the base of the PR and between dffac9b and b810f72.

📒 Files selected for processing (4)

• backend/scripts/dedupe_course_chunks.py
• backend/services/rag_service.py
• backend/tests/test_dedupe_course_chunks.py
• backend/tests/test_rag_service.py

📝 Walkthrough

Walkthrough

Course chunk indexing now uses course-scoped content hashes and deduplicates repeated content. A migration script plans and applies rewrites of legacy document rows, while tests cover deduplication, embedding preservation, idempotency, and course isolation.

Changes

Course chunk identifiers

Layer / File(s)	Summary
Content-hash indexing and validation backend/services/rag_service.py, backend/tests/test_rag_service.py	Adds course-scoped SHA-256 chunk IDs, deduplicates indexing payloads, and tests stable IDs, course isolation, and unique upsert counts.
Legacy row migration backend/scripts/dedupe_course_chunks.py, backend/tests/test_dedupe_course_chunks.py	Adds paginated document-row fetching, winner-based migration planning, dry-run and batched apply modes, and tests for duplicate collapse, embedding reuse, idempotency, and cross-course separation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
participant CLI as dedupe_course_chunks
participant Database as course_chunks
participant Planner as plan_migration
CLI->>Database: Fetch paginated document rows
CLI->>Planner: Group rows by course-scoped chunk ID
Planner-->>CLI: Return winner upserts and stale IDs
CLI->>Database: Upsert rewritten rows
CLI->>Database: Delete duplicate legacy rows

Loading

---

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.}

[cloudflare-workers-and-pages]

Deploying with Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	frontend-staging	b810f72	Commit Preview URL Branch Preview URL	Jul 29 2026, 09:25 AM

[Darkest-Teddy]

Problem

course_chunks ids were sha256(doc_id::index::text), so identical content uploaded by different students (or the same doc re-uploaded) stored a full copy of every chunk and its embedding per upload. A popular course with shared slides multiplies chunk storage by enrollment count.

Fix

• rag_service.chunk_id (new): ids are now sha256(course_id::chunk_text) — content-addressed, scoped per course. Identical text in the same course merges to one row regardless of document or uploader; the same text in different courses stays separate.
• doc_id / uploader_id / chunk_index become last-writer-wins metadata. Safe because nothing reads, filters, or deletes by them: the match_course_chunks RPC returns only course_id/chunk_text/similarity, and no code path deletes course_chunks by doc_id.
• In-batch dedup in index_document_chunks: a document repeating the same text (boilerplate headers) would otherwise put duplicate keys in one upsert payload, which Postgres rejects (ON CONFLICT DO UPDATE cannot affect row a second time).
• scripts/dedupe_course_chunks.py: one-time, idempotent, dry-run-by-default migration for existing rows — groups legacy rows by content-hash id, keeps a row that already has an embedding (no re-embedding cost), upserts winners before deleting stale ids so an interrupted run never drops a chunk. Catalog rows (category=catalog, own id scheme via ingest_catalog.py) are untouched.

Not covered

• Cross-course dedup (same textbook chapter in two courses) — intentional: rows carry per-course metadata, and retrieval filters by course.
• backfill_document_chunks.py's _already_indexed check keys on doc_id; under dedup a doc whose chunks were claimed by another doc's rows may re-index and flip doc_id — harmless churn, ids are stable.

Testing

• 3 new tests in test_rag_service.py (same text + course collides, different courses don't, in-batch dedup) and 4 in test_dedupe_course_chunks.py (collapse, embedding-preferring winner, idempotency, cross-course separation) — all written first and watched fail (TDD).
• Full backend suite: 963 passed, 1 skipped; the test_ocr_pipeline.py::test_save_to_db error is the pre-existing live-Supabase/event-loop artifact documented in ADR 0016 (reproduces identically with the change stashed).
• ruff check clean on all touched files.

Deploy note

After merge, run python scripts/dedupe_course_chunks.py (dry-run), review counts, then --apply — once per environment (dotenv -f .env.staging run -- ... for staging).

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Improved document indexing by identifying identical content within each course and avoiding duplicate entries.
  • Preserved separate entries for identical content across different courses.
  • Added a migration utility to consolidate existing duplicate document chunks, with a preview mode before changes are applied.

• Bug Fixes

  • Prevented repeated indexing of the same content from creating redundant records.
  • Preserved existing embeddings when consolidating duplicate chunks.

• Tests

  • Added coverage for deduplication, course scoping, migration behavior, and idempotent indexing.

[coderabbitai]

Warning

Review limit reached

@AndresL230, you've reached your PR review limit, so we couldn't start this review.

Next review available in:19 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: fdb9d637-c8a5-4a79-9e03-8a256bcda721

📥 Commits

Reviewing files that changed from the base of the PR and between dffac9b and b810f72.

📒 Files selected for processing (4)

• backend/scripts/dedupe_course_chunks.py
• backend/services/rag_service.py
• backend/tests/test_dedupe_course_chunks.py
• backend/tests/test_rag_service.py

📝 Walkthrough

Walkthrough

Course chunk indexing now uses course-scoped content hashes and deduplicates repeated content. A migration script plans and applies rewrites of legacy document rows, while tests cover deduplication, embedding preservation, idempotency, and course isolation.

Changes

Course chunk identifiers

Layer / File(s)	Summary
Content-hash indexing and validation backend/services/rag_service.py, backend/tests/test_rag_service.py	Adds course-scoped SHA-256 chunk IDs, deduplicates indexing payloads, and tests stable IDs, course isolation, and unique upsert counts.
Legacy row migration backend/scripts/dedupe_course_chunks.py, backend/tests/test_dedupe_course_chunks.py	Adds paginated document-row fetching, winner-based migration planning, dry-run and batched apply modes, and tests for duplicate collapse, embedding reuse, idempotency, and cross-course separation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
participant CLI as dedupe_course_chunks
participant Database as course_chunks
participant Planner as plan_migration
CLI->>Database: Fetch paginated document rows
CLI->>Planner: Group rows by course-scoped chunk ID
Planner-->>CLI: Return winner upserts and stale IDs
CLI->>Database: Upsert rewritten rows
CLI->>Database: Delete duplicate legacy rows

Loading

---

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.}

[cloudflare-workers-and-pages]

Deploying with Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	frontend-staging	b810f72	Commit Preview URL Branch Preview URL	Jul 29 2026, 09:25 AM

[Darkest-Teddy]

Problem

course_chunks ids were sha256(doc_id::index::text), so identical content uploaded by different students (or the same doc re-uploaded) stored a full copy of every chunk and its embedding per upload. A popular course with shared slides multiplies chunk storage by enrollment count.

Fix

• rag_service.chunk_id (new): ids are now sha256(course_id::chunk_text) — content-addressed, scoped per course. Identical text in the same course merges to one row regardless of document or uploader; the same text in different courses stays separate.
• doc_id / uploader_id / chunk_index become last-writer-wins metadata. Safe because nothing reads, filters, or deletes by them: the match_course_chunks RPC returns only course_id/chunk_text/similarity, and no code path deletes course_chunks by doc_id.
• In-batch dedup in index_document_chunks: a document repeating the same text (boilerplate headers) would otherwise put duplicate keys in one upsert payload, which Postgres rejects (ON CONFLICT DO UPDATE cannot affect row a second time).
• scripts/dedupe_course_chunks.py: one-time, idempotent, dry-run-by-default migration for existing rows — groups legacy rows by content-hash id, keeps a row that already has an embedding (no re-embedding cost), upserts winners before deleting stale ids so an interrupted run never drops a chunk. Catalog rows (category=catalog, own id scheme via ingest_catalog.py) are untouched.

Not covered

• Cross-course dedup (same textbook chapter in two courses) — intentional: rows carry per-course metadata, and retrieval filters by course.
• backfill_document_chunks.py's _already_indexed check keys on doc_id; under dedup a doc whose chunks were claimed by another doc's rows may re-index and flip doc_id — harmless churn, ids are stable.

Testing

• 3 new tests in test_rag_service.py (same text + course collides, different courses don't, in-batch dedup) and 4 in test_dedupe_course_chunks.py (collapse, embedding-preferring winner, idempotency, cross-course separation) — all written first and watched fail (TDD).
• Full backend suite: 963 passed, 1 skipped; the test_ocr_pipeline.py::test_save_to_db error is the pre-existing live-Supabase/event-loop artifact documented in ADR 0016 (reproduces identically with the change stashed).
• ruff check clean on all touched files.

Deploy note

After merge, run python scripts/dedupe_course_chunks.py (dry-run), review counts, then --apply — once per environment (dotenv -f .env.staging run -- ... for staging).

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Improved document indexing by identifying identical content within each course and avoiding duplicate entries.
  • Preserved separate entries for identical content across different courses.
  • Added a migration utility to consolidate existing duplicate document chunks, with a preview mode before changes are applied.

• Bug Fixes

  • Prevented repeated indexing of the same content from creating redundant records.
  • Preserved existing embeddings when consolidating duplicate chunks.

• Tests

  • Added coverage for deduplication, course scoping, migration behavior, and idempotent indexing.

[coderabbitai]

Warning

Review limit reached

@AndresL230, you've reached your PR review limit, so we couldn't start this review.

Next review available in:19 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: fdb9d637-c8a5-4a79-9e03-8a256bcda721

📥 Commits

Reviewing files that changed from the base of the PR and between dffac9b and b810f72.

📒 Files selected for processing (4)

• backend/scripts/dedupe_course_chunks.py
• backend/services/rag_service.py
• backend/tests/test_dedupe_course_chunks.py
• backend/tests/test_rag_service.py

📝 Walkthrough

Walkthrough

Course chunk indexing now uses course-scoped content hashes and deduplicates repeated content. A migration script plans and applies rewrites of legacy document rows, while tests cover deduplication, embedding preservation, idempotency, and course isolation.

Changes

Course chunk identifiers

Layer / File(s)	Summary
Content-hash indexing and validation backend/services/rag_service.py, backend/tests/test_rag_service.py	Adds course-scoped SHA-256 chunk IDs, deduplicates indexing payloads, and tests stable IDs, course isolation, and unique upsert counts.
Legacy row migration backend/scripts/dedupe_course_chunks.py, backend/tests/test_dedupe_course_chunks.py	Adds paginated document-row fetching, winner-based migration planning, dry-run and batched apply modes, and tests for duplicate collapse, embedding reuse, idempotency, and cross-course separation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
participant CLI as dedupe_course_chunks
participant Database as course_chunks
participant Planner as plan_migration
CLI->>Database: Fetch paginated document rows
CLI->>Planner: Group rows by course-scoped chunk ID
Planner-->>CLI: Return winner upserts and stale IDs
CLI->>Database: Upsert rewritten rows
CLI->>Database: Delete duplicate legacy rows

Loading

---

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.}

[cloudflare-workers-and-pages]

Deploying with Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	frontend-staging	b810f72	Commit Preview URL Branch Preview URL	Jul 29 2026, 09:25 AM

[Darkest-Teddy]

Problem

course_chunks ids were sha256(doc_id::index::text), so identical content uploaded by different students (or the same doc re-uploaded) stored a full copy of every chunk and its embedding per upload. A popular course with shared slides multiplies chunk storage by enrollment count.

Fix

• rag_service.chunk_id (new): ids are now sha256(course_id::chunk_text) — content-addressed, scoped per course. Identical text in the same course merges to one row regardless of document or uploader; the same text in different courses stays separate.
• doc_id / uploader_id / chunk_index become last-writer-wins metadata. Safe because nothing reads, filters, or deletes by them: the match_course_chunks RPC returns only course_id/chunk_text/similarity, and no code path deletes course_chunks by doc_id.
• In-batch dedup in index_document_chunks: a document repeating the same text (boilerplate headers) would otherwise put duplicate keys in one upsert payload, which Postgres rejects (ON CONFLICT DO UPDATE cannot affect row a second time).
• scripts/dedupe_course_chunks.py: one-time, idempotent, dry-run-by-default migration for existing rows — groups legacy rows by content-hash id, keeps a row that already has an embedding (no re-embedding cost), upserts winners before deleting stale ids so an interrupted run never drops a chunk. Catalog rows (category=catalog, own id scheme via ingest_catalog.py) are untouched.

Not covered

• Cross-course dedup (same textbook chapter in two courses) — intentional: rows carry per-course metadata, and retrieval filters by course.
• backfill_document_chunks.py's _already_indexed check keys on doc_id; under dedup a doc whose chunks were claimed by another doc's rows may re-index and flip doc_id — harmless churn, ids are stable.

Testing

• 3 new tests in test_rag_service.py (same text + course collides, different courses don't, in-batch dedup) and 4 in test_dedupe_course_chunks.py (collapse, embedding-preferring winner, idempotency, cross-course separation) — all written first and watched fail (TDD).
• Full backend suite: 963 passed, 1 skipped; the test_ocr_pipeline.py::test_save_to_db error is the pre-existing live-Supabase/event-loop artifact documented in ADR 0016 (reproduces identically with the change stashed).
• ruff check clean on all touched files.

Deploy note

After merge, run python scripts/dedupe_course_chunks.py (dry-run), review counts, then --apply — once per environment (dotenv -f .env.staging run -- ... for staging).

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Improved document indexing by identifying identical content within each course and avoiding duplicate entries.
  • Preserved separate entries for identical content across different courses.
  • Added a migration utility to consolidate existing duplicate document chunks, with a preview mode before changes are applied.

• Bug Fixes

  • Prevented repeated indexing of the same content from creating redundant records.
  • Preserved existing embeddings when consolidating duplicate chunks.

• Tests

  • Added coverage for deduplication, course scoping, migration behavior, and idempotent indexing.

[coderabbitai]

Warning

Review limit reached

@AndresL230, you've reached your PR review limit, so we couldn't start this review.

Next review available in:19 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: fdb9d637-c8a5-4a79-9e03-8a256bcda721

📥 Commits

Reviewing files that changed from the base of the PR and between dffac9b and b810f72.

📒 Files selected for processing (4)

• backend/scripts/dedupe_course_chunks.py
• backend/services/rag_service.py
• backend/tests/test_dedupe_course_chunks.py
• backend/tests/test_rag_service.py

📝 Walkthrough

Walkthrough

Course chunk indexing now uses course-scoped content hashes and deduplicates repeated content. A migration script plans and applies rewrites of legacy document rows, while tests cover deduplication, embedding preservation, idempotency, and course isolation.

Changes

Course chunk identifiers

Layer / File(s)	Summary
Content-hash indexing and validation backend/services/rag_service.py, backend/tests/test_rag_service.py	Adds course-scoped SHA-256 chunk IDs, deduplicates indexing payloads, and tests stable IDs, course isolation, and unique upsert counts.
Legacy row migration backend/scripts/dedupe_course_chunks.py, backend/tests/test_dedupe_course_chunks.py	Adds paginated document-row fetching, winner-based migration planning, dry-run and batched apply modes, and tests for duplicate collapse, embedding reuse, idempotency, and cross-course separation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
participant CLI as dedupe_course_chunks
participant Database as course_chunks
participant Planner as plan_migration
CLI->>Database: Fetch paginated document rows
CLI->>Planner: Group rows by course-scoped chunk ID
Planner-->>CLI: Return winner upserts and stale IDs
CLI->>Database: Upsert rewritten rows
CLI->>Database: Delete duplicate legacy rows

Loading

---

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.}

[cloudflare-workers-and-pages]

Deploying with Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	frontend-staging	b810f72	Commit Preview URL Branch Preview URL	Jul 29 2026, 09:25 AM

[Darkest-Teddy]

Problem

course_chunks ids were sha256(doc_id::index::text), so identical content uploaded by different students (or the same doc re-uploaded) stored a full copy of every chunk and its embedding per upload. A popular course with shared slides multiplies chunk storage by enrollment count.

Fix

• rag_service.chunk_id (new): ids are now sha256(course_id::chunk_text) — content-addressed, scoped per course. Identical text in the same course merges to one row regardless of document or uploader; the same text in different courses stays separate.
• doc_id / uploader_id / chunk_index become last-writer-wins metadata. Safe because nothing reads, filters, or deletes by them: the match_course_chunks RPC returns only course_id/chunk_text/similarity, and no code path deletes course_chunks by doc_id.
• In-batch dedup in index_document_chunks: a document repeating the same text (boilerplate headers) would otherwise put duplicate keys in one upsert payload, which Postgres rejects (ON CONFLICT DO UPDATE cannot affect row a second time).
• scripts/dedupe_course_chunks.py: one-time, idempotent, dry-run-by-default migration for existing rows — groups legacy rows by content-hash id, keeps a row that already has an embedding (no re-embedding cost), upserts winners before deleting stale ids so an interrupted run never drops a chunk. Catalog rows (category=catalog, own id scheme via ingest_catalog.py) are untouched.

Not covered

• Cross-course dedup (same textbook chapter in two courses) — intentional: rows carry per-course metadata, and retrieval filters by course.
• backfill_document_chunks.py's _already_indexed check keys on doc_id; under dedup a doc whose chunks were claimed by another doc's rows may re-index and flip doc_id — harmless churn, ids are stable.

Testing

• 3 new tests in test_rag_service.py (same text + course collides, different courses don't, in-batch dedup) and 4 in test_dedupe_course_chunks.py (collapse, embedding-preferring winner, idempotency, cross-course separation) — all written first and watched fail (TDD).
• Full backend suite: 963 passed, 1 skipped; the test_ocr_pipeline.py::test_save_to_db error is the pre-existing live-Supabase/event-loop artifact documented in ADR 0016 (reproduces identically with the change stashed).
• ruff check clean on all touched files.

Deploy note

After merge, run python scripts/dedupe_course_chunks.py (dry-run), review counts, then --apply — once per environment (dotenv -f .env.staging run -- ... for staging).

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Improved document indexing by identifying identical content within each course and avoiding duplicate entries.
  • Preserved separate entries for identical content across different courses.
  • Added a migration utility to consolidate existing duplicate document chunks, with a preview mode before changes are applied.

• Bug Fixes

  • Prevented repeated indexing of the same content from creating redundant records.
  • Preserved existing embeddings when consolidating duplicate chunks.

• Tests

  • Added coverage for deduplication, course scoping, migration behavior, and idempotent indexing.

[coderabbitai]

Warning

Review limit reached

@AndresL230, you've reached your PR review limit, so we couldn't start this review.

Next review available in:19 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: fdb9d637-c8a5-4a79-9e03-8a256bcda721

📥 Commits

Reviewing files that changed from the base of the PR and between dffac9b and b810f72.

📒 Files selected for processing (4)

• backend/scripts/dedupe_course_chunks.py
• backend/services/rag_service.py
• backend/tests/test_dedupe_course_chunks.py
• backend/tests/test_rag_service.py

📝 Walkthrough

Walkthrough

Course chunk indexing now uses course-scoped content hashes and deduplicates repeated content. A migration script plans and applies rewrites of legacy document rows, while tests cover deduplication, embedding preservation, idempotency, and course isolation.

Changes

Course chunk identifiers

Layer / File(s)	Summary
Content-hash indexing and validation backend/services/rag_service.py, backend/tests/test_rag_service.py	Adds course-scoped SHA-256 chunk IDs, deduplicates indexing payloads, and tests stable IDs, course isolation, and unique upsert counts.
Legacy row migration backend/scripts/dedupe_course_chunks.py, backend/tests/test_dedupe_course_chunks.py	Adds paginated document-row fetching, winner-based migration planning, dry-run and batched apply modes, and tests for duplicate collapse, embedding reuse, idempotency, and cross-course separation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
participant CLI as dedupe_course_chunks
participant Database as course_chunks
participant Planner as plan_migration
CLI->>Database: Fetch paginated document rows
CLI->>Planner: Group rows by course-scoped chunk ID
Planner-->>CLI: Return winner upserts and stale IDs
CLI->>Database: Upsert rewritten rows
CLI->>Database: Delete duplicate legacy rows

Loading

---

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.}

[cloudflare-workers-and-pages]

Deploying with Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	frontend-staging	b810f72	Commit Preview URL Branch Preview URL	Jul 29 2026, 09:25 AM

[Darkest-Teddy]

Problem

course_chunks ids were sha256(doc_id::index::text), so identical content uploaded by different students (or the same doc re-uploaded) stored a full copy of every chunk and its embedding per upload. A popular course with shared slides multiplies chunk storage by enrollment count.

Fix

• rag_service.chunk_id (new): ids are now sha256(course_id::chunk_text) — content-addressed, scoped per course. Identical text in the same course merges to one row regardless of document or uploader; the same text in different courses stays separate.
• doc_id / uploader_id / chunk_index become last-writer-wins metadata. Safe because nothing reads, filters, or deletes by them: the match_course_chunks RPC returns only course_id/chunk_text/similarity, and no code path deletes course_chunks by doc_id.
• In-batch dedup in index_document_chunks: a document repeating the same text (boilerplate headers) would otherwise put duplicate keys in one upsert payload, which Postgres rejects (ON CONFLICT DO UPDATE cannot affect row a second time).
• scripts/dedupe_course_chunks.py: one-time, idempotent, dry-run-by-default migration for existing rows — groups legacy rows by content-hash id, keeps a row that already has an embedding (no re-embedding cost), upserts winners before deleting stale ids so an interrupted run never drops a chunk. Catalog rows (category=catalog, own id scheme via ingest_catalog.py) are untouched.

Not covered

• Cross-course dedup (same textbook chapter in two courses) — intentional: rows carry per-course metadata, and retrieval filters by course.
• backfill_document_chunks.py's _already_indexed check keys on doc_id; under dedup a doc whose chunks were claimed by another doc's rows may re-index and flip doc_id — harmless churn, ids are stable.

Testing

• 3 new tests in test_rag_service.py (same text + course collides, different courses don't, in-batch dedup) and 4 in test_dedupe_course_chunks.py (collapse, embedding-preferring winner, idempotency, cross-course separation) — all written first and watched fail (TDD).
• Full backend suite: 963 passed, 1 skipped; the test_ocr_pipeline.py::test_save_to_db error is the pre-existing live-Supabase/event-loop artifact documented in ADR 0016 (reproduces identically with the change stashed).
• ruff check clean on all touched files.

Deploy note

After merge, run python scripts/dedupe_course_chunks.py (dry-run), review counts, then --apply — once per environment (dotenv -f .env.staging run -- ... for staging).

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Improved document indexing by identifying identical content within each course and avoiding duplicate entries.
  • Preserved separate entries for identical content across different courses.
  • Added a migration utility to consolidate existing duplicate document chunks, with a preview mode before changes are applied.

• Bug Fixes

  • Prevented repeated indexing of the same content from creating redundant records.
  • Preserved existing embeddings when consolidating duplicate chunks.

• Tests

  • Added coverage for deduplication, course scoping, migration behavior, and idempotent indexing.

[coderabbitai]

Warning

Review limit reached

@AndresL230, you've reached your PR review limit, so we couldn't start this review.

Next review available in:19 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: fdb9d637-c8a5-4a79-9e03-8a256bcda721

📥 Commits

Reviewing files that changed from the base of the PR and between dffac9b and b810f72.

📒 Files selected for processing (4)

• backend/scripts/dedupe_course_chunks.py
• backend/services/rag_service.py
• backend/tests/test_dedupe_course_chunks.py
• backend/tests/test_rag_service.py

📝 Walkthrough

Walkthrough

Course chunk indexing now uses course-scoped content hashes and deduplicates repeated content. A migration script plans and applies rewrites of legacy document rows, while tests cover deduplication, embedding preservation, idempotency, and course isolation.

Changes

Course chunk identifiers

Layer / File(s)	Summary
Content-hash indexing and validation backend/services/rag_service.py, backend/tests/test_rag_service.py	Adds course-scoped SHA-256 chunk IDs, deduplicates indexing payloads, and tests stable IDs, course isolation, and unique upsert counts.
Legacy row migration backend/scripts/dedupe_course_chunks.py, backend/tests/test_dedupe_course_chunks.py	Adds paginated document-row fetching, winner-based migration planning, dry-run and batched apply modes, and tests for duplicate collapse, embedding reuse, idempotency, and cross-course separation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
participant CLI as dedupe_course_chunks
participant Database as course_chunks
participant Planner as plan_migration
CLI->>Database: Fetch paginated document rows
CLI->>Planner: Group rows by course-scoped chunk ID
Planner-->>CLI: Return winner upserts and stale IDs
CLI->>Database: Upsert rewritten rows
CLI->>Database: Delete duplicate legacy rows

Loading

---

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.}

[cloudflare-workers-and-pages]

Deploying with Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	frontend-staging	b810f72	Commit Preview URL Branch Preview URL	Jul 29 2026, 09:25 AM