fix(cloudstore): invalidate dashboard cache after every InsertMutationBatch commit

[danielgap]

🔗 Linked Issue

Closes #251

---

🏷️ PR Type

• type:bug — Bug fix

---

📝 Summary

• InsertMutationBatch now unconditionally invalidates the dashboard read model cache after every successful commit, aligning its behavior with WriteChunk.
• Before this fix, only batches that produced materialized chunks (session/observation/prompt entities) triggered cache invalidation. Batches containing only non-materializable entities (e.g. relation) left the cache stale, causing 404s on the dashboard until a manual restart.

📂 Changes

File	Change
internal/cloud/cloudstore/cloudstore.go	Moved invalidateDashboardReadModel() outside the if len(chunks) > 0 conditional so it always runs after commit
internal/cloud/cloudstore/cloudstore_test.go	Added TestInsertMutationBatchInvalidatesDashboardReadModel regression test using a relation-only batch (zero chunks)

🧪 Test Plan

• Unit tests pass locally: go test ./internal/cloud/cloudstore/ — all pass (47 tests, 0 failures)
• E2E tests pass locally: go test -tags e2e ./internal/server/... — skipped (no local Postgres)
• Manually verified that InsertMutationBatch with a relation-only batch triggers cache invalidation

---

🤖 Automated Checks

These run automatically and all must pass before merge:

Check	What it verifies	Status
Check Issue Reference	PR body contains Closes #N / Fixes #N / Resolves #N	⏳
Check Issue Has status:approved	Linked issue has status:approved label	⏳
Check PR Has type:* Label	PR has exactly one type:* label	⏳
Unit Tests	go test ./... passes	⏳
E2E Tests	go test -tags e2e ./internal/server/... passes	⏳

---

✅ Contributor Checklist

• I linked an approved issue above (Closes #251)
• I added exactly one type:* label to this PR (type:bug)
• I ran unit tests locally: go test ./internal/cloud/cloudstore/
• I ran e2e tests locally: no local Postgres available
• Docs updated (if behavior changed) — no docs change needed, internal cache behavior
• Commits follow conventional commits format
• No Co-Authored-By trailers in commits

---

💬 Notes for Reviewers

The root cause is that InsertMutationBatch guarded the cache invalidation behind if len(chunks) > 0, but WriteChunk (the other write path) invalidates unconditionally. This meant that mutations arriving via autosync that contained only non-materializable entities (like relation) never invalidated the cache, making the dashboard stale until a server restart. The fix is a one-line change that matches the proven pattern from WriteChunk.

[Copilot]

Pull request overview

This PR fixes dashboard staleness in the cloudstore write path by ensuring InsertMutationBatch invalidates the dashboard read-model cache after every successful commit (including batches that produce no materialized chunks, e.g., relation-only batches), and adds a regression test to prevent reintroducing the bug.

Changes:

• Make InsertMutationBatch always call invalidateDashboardReadModel() after a successful transaction commit.
• Add a regression test proving cache invalidation happens even when the mutation batch yields zero materialized chunks.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
internal/cloud/cloudstore/cloudstore.go	Unconditionally invalidate the dashboard read-model cache after committing InsertMutationBatch.
internal/cloud/cloudstore/cloudstore_test.go	Add regression coverage for cache invalidation on a relation-only mutation batch (zero chunks).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.