Release: actionable intelligence, DQL, and platform cleanup

[izadoesdev]

What

• Ship tenant-isolated DQL through authenticated API/RPC paths and retire the legacy custom-query and chart stack.
• Split platform intelligence into readable Insights and durable Investigations, with release/PR evidence, exact recommendations, and native goal edit/delete actions.
• Expose published insights to the dashboard agent, Slack, and MCP while keeping one canonical investigation timeline and reply path.
• Add intelligence plans/credits and align the dashboard, docs, and product contract.
• Remove retired packages, duplicate schemas/helpers/tests, obsolete recommendation compatibility, and special-case UI state.

Why

Databuddy should tell teams what changed, why it matters, and what they can do—not collapse every useful finding into a case or return vague analytics prose. This release makes quiet insights readable, keeps material work durable, and lets exact goal recommendations open the existing editor prefilled for explicit review.

Product impact

• Improvements, recoveries, and useful findings remain visible without manufacturing incidents.
• Actionable cases retain evidence, human context, rechecks, and resolution history.
• Goal cleanup recommendations name the exact proposed value and lead directly to the existing edit/delete flow.
• Slack and MCP read the same stored intelligence instead of inventing parallel advice.
• The final release is smaller overall: 181 files, 6,114 additions, 7,492 deletions (net -1,378 LOC).

Safety and compatibility

• Historical stored observations still parse; only the never-persisted recommendation-without-changes shape was removed (live count: 0).
• DQL remains read-only and tenant scoped.
• The recommendation contract is one strict discriminated union shared by generation, storage, dashboard, Slack, and MCP.
• Invalid, stale, or already-applied goal proposals do not open an unchanged editor.

Validation

• bun run lint
• bun run check-types — 34 tasks across all 31 packages
• Full pre-push bun run test — 25/25 workspace test tasks successful
• Focused contracts — 99 tests passed across Shared, Insights, and AI
• Controlled Terra replay — schema/provider accepted; exact self-contained goal edit produced; estimated cost $0.013681
• Two independent adversarial cleanup reviews — clean

The seven local DB-backed API integration cases loaded successfully but skipped because the local test database was unavailable; CI provides the service-backed run.

---

Summary by cubic

Ships tenant-scoped, read-only DQL with hardened Bun-only provisioning, and a rebuilt intelligence model: readable Insights and durable Investigations with executable goal recommendations. Unifies delivery across the dashboard, Slack, and MCP, adds investigation credits and new plans, and removes the legacy custom-query/chart stack.

• New Features

  • Authenticated DQL execute/schema endpoints with tenant isolation and hard limits; enforces a 512 KB JSON body cap and returns 413 for oversized or streamed requests; requires CLICKHOUSE_DQL_URL and isolates Bun-only provisioning.
  • Split intelligence into readable Insights and durable Investigations; recommendations open goal edit/delete prefilled for review; dashboard shows an Insights feed and updated nav.
  • Investigation credits and new plan family in billing; dashboard and docs updated; Slack/MCP read the same stored intelligence and share one reply path.

• Migration

  • Set CLICKHOUSE_DQL_URL to a separately provisioned, least‑privileged ClickHouse user for the Bun DQL service; DQL remains read‑only and rejects bodies over 512 KB.
  • The legacy /v1/query/custom endpoint is removed. Use /v1/query/compile and /v1/query instead; docs and rate limits updated.
  • Retired the dashboard custom‑query UI and chart image code; removed @databuddy/charts and related tests. Existing stored observations remain compatible.

^{Written for commit 60d56b2. Summary will update on new commits.}

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
dashboard (staging)	Ready	Preview, Comment	Jul 25, 2026 1:10pm
databuddy-status	Ready	Preview, Comment	Jul 25, 2026 1:10pm
documentation (staging)	Ready	Preview, Comment	Jul 25, 2026 1:10pm

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 843386c5-de95-42ec-a91c-53ffbec69fb1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch staging

---

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.}

[unkey-deploy]

The latest updates on your projects. Learn more about Unkey Deploy

Name	Status	Preview	Inspect	Updated (UTC)
api (preview)	Ready	Visit Preview	Inspect	Jul 25, 2026 1:09pm

[greptile-apps]

Too many files changed for review. (181 files found, 100 file limit)

Bypass the limit by tagging @greptile-apps to review.