feat(insights): ship durable agent investigations

[izadoesdev]

What changed

• Replaced the parallel insight/digest pipeline with one durable investigation case per subject and a chronological observation/reply timeline.
• Let the agent investigate with live analytics, named goals and funnels, website context, and bound GitHub/deploy context.
• Exposed the same investigation lifecycle through the dashboard, MCP, and Slack; a Slack reply resumes the original case.
• Reduced organization configuration to Off, Daily, or Weekly plus an optional Slack channel.
• Removed duplicate evidence readers, formatter layers, prompt rails, feature gates, the synthetic eval package, and the manifest updater.
• Updated SPEC.md, ROADMAP.md, dashboard copy, and public docs around actionable investigations.
• Hardened production-shadow dry runs so anomaly detection is blocked and page scraping cannot write cache entries.
• Declared the docs tracker dependency so Turbo builds the tracker before docs copy tests.
• Made package-global AI mocks order-safe so CI cannot hide missing Redis or tracing exports.

The total branch diff is 257 files with 11,845 additions and 32,278 deletions.

Why

Databuddy should explain a concrete business or product problem, show the evidence, ask for missing context when it matters, and continue working toward a fix. It should not emit anonymous metric cards or narrate internal agent behavior.

Rollout

• The legacy generation-config columns are intentionally removed; the maintainer has already run the schema push for the current environment.
• Deploy API, Insights, Dashboard, and Slack from the same revision.
• The Slack manifest has already been deployed manually. No manifest updater script remains.

Validation

• bun install --frozen-lockfile --ignore-scripts
• bun run lint
• bun run check-types
• bun run test (including 3,512 AI tests)
• bun run build
• Insights integration: 28/28
• API insights integration: 5/5
• Dashboard Playwright PR suite: 7/7
• ClickHouse schema check
• Docs link check
• API, Insights, and Slack production Docker builds for linux/amd64
• Dry-run mutation/cache regression tests
• Randomized Bun mock-order regressions
• git diff --check

AI-assisted with Codex; the branch has been locally verified.

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
databuddy-status	Ready	Preview, Comment	Jul 20, 2026 2:16pm

2 Skipped Deployments

Project	Deployment	Actions	Updated (UTC)
dashboard	Skipped		Jul 20, 2026 2:16pm
documentation	Skipped		Jul 20, 2026 2:16pm

[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

Run ID: 45e1d6ca-eee1-4d08-8b05-1f0eb0e7928d

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 codex/insights-agent-simplification

---

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 20, 2026 2:17pm

[greptile-apps]

Greptile Summary

This PR replaces deterministic insight decisions with a bounded investigation agent. The main changes are:

• Adds model-selected evidence reads and decision submission.
• Keeps persistence, repair authorization, and lifecycle handling in the backend.
• Updates insight evaluation and production-shadow flows.
• Renames the dashboard copy section to “What changed.”

Confidence Score: 4/5

The production read-only investigation path can invoke live anomaly detection.

• Candidate selection and evidence citations are scoped to one signal.
• Invalid agent runs throw instead of persisting a fallback decision.
• The production reader must disable live anomaly detection before merging.

apps/insights/src/generation.ts

Security Review

The new evidence reader keeps website, date, and signal scope server-controlled. The production investigation path still permits live anomaly detection while running in dry-run mode.

Important Files Changed

Filename	Overview
apps/insights/src/agent.ts	Adds bounded candidate selection, evidence reads, and validated finding materialization.
apps/insights/src/generation.ts	Integrates the agent into generation, billing, persistence, and observation handling; the evidence-reader setup leaves live anomaly detection enabled.
packages/ai/src/ai/tools/insights-agent-tools.ts	Adds strict request schemas and signal-bound evidence reads.
packages/ai/src/ai/insights/ops-context.ts	Adds a live anomaly-detection switch that defaults to enabled.
packages/shared/src/insights.ts	Simplifies the needs-context decision shape.
packages/evals/src/insight-production-shadow.ts	Updates historical shadow evaluation to use the bounded agent and disables live anomaly detection there.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Detected regressions] --> B[Eligible candidate shortlist]
  B --> C[Bounded investigation agent]
  C --> D[Signal-scoped evidence reader]
  D --> E[Validated decision]
  E --> F[Persist insight and observation]
  F --> G[Resolve prior insights and effects]

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
flowchart TD
  A[Detected regressions] --> B[Eligible candidate shortlist]
  B --> C[Bounded investigation agent]
  C --> D[Signal-scoped evidence reader]
  D --> E[Validated decision]
  E --> F[Persist insight and observation]
  F --> G[Resolve prior insights and effects]

Loading

_{Reviews (1): Last reviewed commit: "feat(insights): restore bounded investig..." | Re-trigger Greptile}