Pull Request Explorer

Summary

Fix MSSQL credentials error messaging

Corrects misleading error message for SQL Server authentication failures, aligning with other SQL sources and improving user clarity.

Health Assessment

• • Quick fix with minimal changes and fast review, indicating low complexity and high confidence in the change.

Summary

Add precompute health trend charts

Provides staff with time‑series visualizations of precompute health, enabling quicker detection of regressions and more efficient incident response.

Health Assessment

• • The PR was reviewed and merged within 4 hours, with no post‑review commits, indicating a smooth, low‑friction process.

Summary

Defer Snowflake Connector Import to Reduce Startup Cost

Lazily imports the heavy Snowflake SDK to cut backend and test startup time, improving worker performance.

Health Assessment

• • Quick, low‑impact change with fast review and merge, minimal risk to production.

Summary

Map Supabase pooler errors to friendly messages

Adds specific error mappings for Supabase pooler failures, improving user feedback and reducing noise.

Health Assessment

• • Quick approval with minimal changes indicates a straightforward bug fix that was easy to review and merge.

Summary

Update LLM costs

Automated update of LLM pricing models to keep cost calculations accurate.

Health Assessment

• • Automated PR with minimal changes, fast merge, no review needed.

Summary

Expose SignalTeamConfig in Django admin

Adds an admin interface for SignalTeamConfig, enabling support to set a team's default Slack channel without shell access, improving self‑service and reducing support overhead.

Health Assessment

• • Fast 2.3‑hour cycle with minimal changes and no tests, indicating low risk and smooth review process.

Summary

Move prompt archive and duplicate to overflow menu

Reorganizes the prompt detail UI to reduce destructive actions and streamline duplicate functionality, improving user experience and reducing accidental data loss.

Health Assessment

• • Quick turnaround with single approval and minimal rework, indicating a straightforward UI refactor with low risk.

Summary

Enable link navigation in markdown notebook editor

Adds ability to open links while editing markdown notebooks, improving user experience and reducing friction.

Health Assessment

• • Quick fix with single commit and rapid review, indicating low complexity and high confidence in changes.

Summary

Improve error visibility for data-warehouse view creation

Adds logging of underlying exceptions for data-warehouse view creation failures, enabling faster diagnosis without changing user-facing behavior.

Health Assessment

• • Quick fix with minimal changes, low review friction, fast cycle time.

Summary

Add drag‑to‑zoom to lifecycle and funnel charts

Enables users to zoom into lifecycle and funnel trend charts, improving data exploration and insight accuracy.

Health Assessment

• • Fast review and merge with minimal rework; AI assistance likely accelerated implementation and testing.

Summary

Record deprecated flag-config usage on experiment update

Adds telemetry to track deprecated flag-config usage during experiment updates, improving visibility into legacy traffic and supporting the deprecation strategy.

Health Assessment

• • Fast cycle time (12.2h) and single review round indicate smooth integration; minimal code changes and comprehensive tests suggest low risk.

Summary

Add ingestion-warnings scout to Signals

Automates detection and resolution of ingestion issues, reducing manual effort and improving data reliability.

Health Assessment

• • Quick, AI-assisted PR merged within 1.5 hours with minimal review, indicating high confidence and low risk.

Summary

Update experiment skill docs to new feature_flag shape

Docs-only rewrite to eliminate deprecated experiment flag parameters, reducing telemetry noise and aligning agent training with current API.

Health Assessment

• • Docs-only PR with quick turnaround, single review, minimal changes, no code impact.

Summary

Document missing OAuth scopes for Plain source

Adds required user:read and label:read scopes to documentation and error messages, preventing 403 errors during Plain data imports.

Health Assessment

• • Quick fix with minimal changes, approved within minutes, indicating low complexity and high confidence in the change.

Summary

Clarify error message for duplicate data-warehouse source

Improves user onboarding by providing clearer guidance when adding a second data source of the same type, reducing confusion and retry errors. This change does not alter functionality, only updates error text.

Health Assessment

• • Quick approval with minimal changes indicates low complexity and high confidence in the copy update.

Summary

Fix GitHub 404 handling for org‑scoped endpoints

Ensures GitHub data sync gracefully skips missing org teams, preventing false failures and improving reliability.

Health Assessment

• • Quick single‑commit fix with no rework, cycle time under 8 hours and immediate approval indicates low friction and minimal risk.

Summary

Fix Postgres pooler checkout timeout

Makes Postgres pooler checkout timeout retryable, preventing sync failures due to transient pooler saturation.

Health Assessment

• • Fast review and merge, low complexity, and strong author familiarity indicate a healthy PR process.

Summary

Retry server-cursor read on connection drop

Fixes transient connection drops during Postgres data import, improving reliability of full-table scans. Ensures data syncs resume without data loss or duplication.

Health Assessment

• • Quick fix with minimal rework, approved within minutes, demonstrating efficient issue triage and testing.

Summary

Remove deprecated split_percent from MCP experiments

Eliminates dead code and simplifies experiment data handling, reducing maintenance overhead and potential confusion for developers.

Health Assessment

• • Fast 12‑hour cycle with minimal rework; AI‑assisted removal of a deprecated field received quick approval and no significant risk.

Summary

Add personal and shared MCP installation scope

Enables team-wide MCP connections, improving collaboration and automation across projects by allowing shared installations while preserving personal defaults.

Health Assessment

• • Fast cycle time and single review round indicate a well‑defined change; large code churn suggests significant impact but low friction in review process.

Summary

Clean up experiment service to reduce repetition

Refactors duplicate analytics reporting and validation logic, improving maintainability without changing behavior.

Health Assessment

• • Quick approval with no rework, behavior-preserving refactor, minimal risk.

Summary

Add macOS support to React Native plugin

Enables native error tracking on macOS for the PostHog React Native plugin, expanding platform coverage and ensuring crash reporting works on macOS.

Health Assessment

• • The PR had a moderate cycle time with a single review and several follow-up commits to address review nits, indicating a focused but iterative process.

Summary

Fix MathJax dynamic font load failure

Ensures LaTeX blocks render gracefully by awaiting dynamic font loads, preventing unhandled errors and improving user experience.

Health Assessment

• • Rapid iteration with minimal review; the fix was merged within 1.3 hours, indicating low complexity and high confidence.

Summary

Scaffold 66 new data warehouse source stubs

Adds 66 new source stubs for data warehouse integration, enabling future vendor support without impacting existing functionality.

Health Assessment

• • Rapid iteration with minimal review time indicates high confidence in changes; large scope but well‑structured scaffolding reduces downstream impact.

Summary

Fix editor infinite update loop on graph rebuilds

Resolved a critical React error that caused workflow editor crashes by stabilizing graph array identities, improving reliability for users.

Health Assessment

• • Fast resolution with minimal rework; AI-assisted code improved stability.

Summary

Remove feature flag gates for workflows

Eliminates gating logic for wait‑until‑event and recurring‑schedules, simplifying UI and ensuring features are always available.

Health Assessment

• • Merged within an hour with minimal review, indicating straightforward removal of obsolete feature flag logic.

Summary

Document email deduplication for batch workflows

Adds documentation explaining automatic email deduplication behavior for batch workflows, improving clarity for users.

Health Assessment

• • Quickly merged after minimal review, indicating straightforward documentation update.

Summary

Backfill default email channel on Zendesk re-import

Previously imported Zendesk tickets lacking email channels were left incomplete, causing notification failures. This change automatically assigns the default email channel to those tickets during re-import, eliminating manual intervention.

Health Assessment

• • Quick turnaround with a single commit and no review iterations indicates a straightforward, low-risk change.

Summary

Fix group-by picker attribute key fetching

Ensures metrics viewer offers attribute key suggestions, improving user experience and data exploration.

Health Assessment

• • Quick fix with minimal rework, fully tested, merged within a day.

Summary

Guard BatchingPipeline against empty feeds

Prevents ingestion pipeline from hanging and leaking capacity when processing empty data, improving reliability and resource usage.

Health Assessment

• • Fast approval with no rework; the change addresses a critical edge case with minimal code churn.

Summary

Redirect /desktop to /code

Adds a redirect to maintain consistency in product naming until official rename.

Health Assessment

• • Quick turnaround with minimal changes, low risk

Summary

Fix exec tool size limit for MCP clients

Ensures the MCP exec tool payload stays within client size caps, preventing silent tool drops for Claude web/desktop and improving reliability for users.

Health Assessment

• • Fast review and merge with minimal rework; comprehensive tests guard against future regressions.

Summary

Fix experiment clone rollout percentage inheritance

Ensures duplicated experiments retain original rollout percentage, preventing unintended 100% rollout and maintaining consistent experiment behavior.

Health Assessment

• • Quick fix with a single regression test, minimal code changes, fast review cycle, and low risk to production.

Summary

Remove CDP self-loop guard mode flag

Simplifies CDP configuration by eliminating a dead flag, ensuring the self-loop guard always enforces and reducing risk of accidental disablement.

Health Assessment

• • Fast cycle time (≈21h) with a single review round and minimal rework; internal config cleanup with no runtime behavior change.

Summary

Remove broken span message search from filter bar

Fixes a bug in the tracing filter bar that injected an invalid synthetic search option for spans, preventing errors and improving user experience.

Health Assessment

• • Quick fix with minimal changes, approved within 0.3h, no backend changes, low risk.

Summary

Fix CI product test sharding with optimal_chunks

Improves CI reliability by preventing empty shards and balancing test distribution, reducing flaky failures for product jobs.

Health Assessment

• • Rapid fix with minimal rework; AI-assisted coding and review accelerated the process.

Summary

Fix test function references

Corrects test references to use the proper textCardConverter methods, eliminating errors in the TextCard markdown tests. This restores frontend type checking and prevents Jest failures on the main branch.

Health Assessment

• • Fast merge with minimal changes and quick review, indicating low complexity and risk.

Summary

Make state claim path the only v3 queue reader

This PR removes the legacy claim path, improving queue performance and simplifying the codebase, reducing backlog and eliminating liveness issues.

Health Assessment

• • Fast review and single commit indicate low friction; the change removes legacy code and improves performance, with minimal risk.

Summary

Update Codex reasoning levels for GPT-5.6

Documentation updated to reflect new GPT-5.6 model support, ensuring accurate guidance for developers.

Health Assessment

• • Merged quickly after a single review with minimal changes, indicating low complexity and risk.

Summary

Improve experiment precompute by spilling to disk

Reduces memory errors in high-volume experiment precomputes, ensuring reliable analytics.

Health Assessment

• • Fast turnaround with minimal rework; the change is a small, low‑risk fix that prevents OOM failures during experiment precomputation.

Summary

Fix CI: Handle Pytest Exit Code 5

Prevents product test shards from failing when pytest-split assigns 0 tests to a shard, ensuring job success even when all actual tests pass in other shards.

Health Assessment

• • Quick review and merge indicate a straightforward fix for a specific issue, with low risk due to the small scope and clear problem description.

Summary

Surface PR merged state in installation progress card

Enables real‑time UI updates when a PR is merged, improving user awareness during onboarding.

Health Assessment

• • Fast cycle time and minimal review friction indicate a well‑understood change with clear scope.

Summary

Add legacy GeoIP transformation migration command

Adds a management command to migrate legacy GeoIP transformations to the modern template, preserving user settings and enabling automated worker reloads.

Health Assessment

• • The PR was reviewed quickly and required minimal rework, indicating a smooth process.

Summary

Bump tiptap to fix text card corruption

Fixes a serialization bug in the dashboard text editor that caused corrupted text cards when mixing formatting with inline code, improving reliability for users. The update ensures consistent markdown rendering and prevents data loss during editing.

Health Assessment

• • Fast review and merge with a single commit, minimal rework, but a large lockfile diff; overall low risk to functionality.

Summary

Use ingestion-warnings-list in health-checks scout

The update replaces manual HogQL queries with the ingestion-warnings-list MCP tool, streamlining the detection of ingestion warnings. This improves accuracy and reduces maintenance overhead for monitoring teams.

Health Assessment

• • Quick review and merge, minimal changes, low risk.

Summary

Keep freeze exposure dialog open with loading state

Improves user experience by providing visual feedback during long freeze/unfreeze operations, reducing confusion and perceived latency.

Health Assessment

• • Quick turnaround with minimal rework; user-facing UX improvement delivered efficiently.

Summary

Update MCPcat attribution to AgentCat

Updates attribution to reflect MCPcat rebranding to AgentCat, ensuring compliance with MIT license requirements.

Health Assessment

• • Quick review and approval process, indicating a straightforward and low-risk change.

Summary

Add wizard PR-ready email notification

Adds a transactional email to notify users when an onboarding wizard PR is ready, gated by a feature flag and ensuring idempotent sends with proper access checks.

Health Assessment

• • Fast review and merge with minimal iterations, but the PR introduces a large number of changes across backend logic and email templates, indicating a significant scope that could impact stability if not thoroughly tested.

Summary

Instrument metrics scene usage with tracked events and data-attrs

Adds comprehensive analytics instrumentation to the metrics product, capturing 15 usage events while ensuring privacy by excluding customer data. This enhances visibility into user interactions and supports data‑driven product decisions.

Health Assessment

• • Fast cycle time (7.1h) and a single review round indicate smooth integration and low risk.

Summary

Improve web analytics path precompute performance

Reduces latency for user-facing analytics by pruning read scans and enforcing budget limits, improving response times and resource usage.

Health Assessment

• • The PR underwent several merge conflicts and required multiple commits after the initial review, indicating moderate iteration intensity, but the overall cycle time remained under a day.