Pull Request Explorer

Summary

Add file type icons to space feed chips

Enhances space feed cards by displaying file type icons and aligning artifact rules, improving user navigation and analytics.

Health Assessment

• • Fast cycle time of 1 hour and a single review round indicate minimal friction and a straightforward change.

Summary

Add person display name lookup for flag calls

Optimizes person lookup for flag‑call lists, cutting query cost for organizations using the flag‑evaluations‑only mode and improving performance.

Health Assessment

• • Fast cycle time (7.7h) and quick first review (0.1h) indicate a smooth process; AI assistance helped reduce rework and keep the change focused.

Summary

Skip event writes for flag evaluation only orgs

Prevents unnecessary event writes for organizations in FLAG_EVALUATIONS_ONLY mode, reducing storage and processing overhead while ensuring accurate flag evaluation data.

Health Assessment

• • Rapid iteration with AI assistance led to a concise merge within 7.6 hours, but multiple post-review commits and extensive comments suggest significant rework and review friction.

Summary

Add data helpers for artifact comments

Provides pure helper functions for handling artifact comments in Desktop, enabling future UI logic and improving collaboration on artifacts.

Health Assessment

• • Rapid approval with minimal review, indicating low complexity and high confidence. The change is behind a feature flag and adds only helper logic, so risk to production is minimal.

Summary

Add canvas timeline rail UI

Improves canvas timeline readability by arranging versions on a vertical rail, reducing clutter and enabling quick navigation.

Health Assessment

• • Fast approval with no back‑and‑forth, indicating low complexity.

Summary

Open timeline from version button

Replaces dropdown menus with buttons that open the timeline tab, improving visibility of version details for users.

Health Assessment

• • Quick approval with no rework indicates a straightforward UI change that was well understood by reviewers.

Summary

Update scanner goal prompt copy

Improves user guidance by clarifying the scanner goal prompt, making it easier for users to specify sessions and objectives.

Health Assessment

• • Quick review and merge with minimal changes; no blockers.

Summary

Fix Replay Vision scanner large masked media handling

Improves Replay Vision by correctly interpreting large masked media, preventing false load failure reports and improving accuracy of media playback analysis.

Health Assessment

• • Merged within 27.5 hours with minimal changes and a single review round, indicating a straightforward bug fix with low risk.

Summary

Clear alert history before lifecycle tests

Ensures test isolation by truncating alert history, preventing flaky test failures.

Health Assessment

• • Quick fix with minimal changes, no production impact, low risk.

Summary

Calibrate Replay Vision Watchable Threshold

Adjusts the watchable feed threshold and adds telemetry to improve friction detection accuracy, enhancing user experience by reducing false positives.

Health Assessment

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

Summary

Add full-screen mode for canvases on web

Enables users to view canvases in full-screen, improving focus and usability.

Health Assessment

• • The PR was reviewed and approved within minutes, with no significant discussion, indicating a straightforward change. The quick cycle time and minimal iterations suggest low risk and high confidence in the implementation.

Summary

Fix bulk right-click and escape behavior

Ensures bulk actions appear on right-click of selected rows and that the Escape key does not prematurely clear selections, improving user workflow.

Health Assessment

• • The PR was merged within 0.8 hours with a single approval and no rework, indicating a low-friction, low-risk change.

Summary

Add flag evaluation sync source setting

Adds a configuration to select the source table for flag evaluation sync, improving accuracy of stale flag detection.

Health Assessment

• • PR was AI-assisted, reviewed quickly with minimal comments, and merged within 9 hours, indicating a smooth and efficient process.

Summary

Move custom instructions to server behind flag

Ensures user‑defined instructions are stored once on the server, preventing duplication across Desktop, web, and Slack, and improving consistency for all task runs.

Health Assessment

• • Fast turnaround with a single review round and minimal rework; feature gated behind a flag reduces production risk.

Summary

Enable disabling materialization for ineligible endpoints

Provides users the ability to turn off materialization on endpoints that are no longer eligible, improving control and ensuring accurate history logs.

Health Assessment

• • Fast cycle time and minimal rework indicate a straightforward bug fix with clear business impact.

Summary

Add canvases tab and feed links to spaces

Enables users to view and manage canvases directly from space pages and feed, improving workflow and reducing friction.

Health Assessment

• • Fast 7‑hour cycle with AI‑assisted code; multiple review comments but quick approvals; moderate risk due to frontend changes but well‑tested.

Summary

Add flag call reading on activity page

Enables accurate flag call tracking by reading from flag_evaluations table, improving analytics for organizations on FLAG_EVALUATIONS_ONLY mode.

Health Assessment

• • Quick review and merge with minimal rework indicates high confidence in changes.

Summary

Embed flags, cohorts, and documents in tasks

Enhances task pages by embedding feature flags and cohorts directly, and improves file listing for updated documents, reducing navigation friction for users.

Health Assessment

• • Merged within 6.5 hours with minimal review rounds; front‑end only change behind a feature flag, indicating low risk and quick turnaround.

Summary

Fix backfill disk wait logic for ClickHouse

Improves reliability of flag evaluation backfill by pausing while ClickHouse moves parts, preventing shard stops and ensuring smoother production runs.

Health Assessment

• • Rapid review and merge with minimal iterations, indicating low complexity and high confidence in changes.

Summary

Fix distinct ID reset for deleted persons

Prevents ghost records and 404 errors by ensuring resets only occur when a live person exists, improving data integrity and user experience.

Health Assessment

• • Quick resolution with minimal rework; AI-assisted code improved correctness.

Summary

Record safe undeclared argument names in MCP

Enhances analytics by capturing guessed argument names, improving error visibility without exposing sensitive data.

Health Assessment

• • Fast review cycle with minimal rework; AI-assisted code contributed; low risk to production.

Summary

Add per-project flag evaluation counts

Provides accurate 7‑day per‑project flag evaluation metrics for organizations using the FLAG_EVALUATIONS_ONLY mode, ensuring correct analytics in the Projects tab.

Health Assessment

• • Fast cycle time and review, minimal rework, AI-assisted development contributed to a smooth, low‑friction merge.

Summary

Remove dead code and correct stale MCP docs

The MCP server codebase is cleaned of unused components, and documentation is updated to reflect current functionality. This reduces maintenance effort and improves developer clarity.

Health Assessment

• • Fast cycle time and minimal review iterations indicate efficient cleanup.

Summary

Gate execute-sql to query:read only

Aligns the MCP catalog with backend permissions, ensuring only tokens with query:read scope can access the execute-sql tool and preventing unauthorized SQL execution.

Health Assessment

• • Fast, single-commit fix with AI assistance, minimal code churn, and no significant review friction.

Summary

Add inbox ranking sort and time window

Staff can order inbox reports by model probability and filter by creation time, improving prioritization.

Health Assessment

• • Rapid approval with minimal review, but large code changes across backend and frontend.

Summary

Fix hidden row key selection in marketing analytics compare

Ensures marketing analytics compare queries work when users hide row key columns, preventing query failures and improving reliability.

Health Assessment

• • Quick fix with minimal changes, resolved in under two hours with no review friction.

Summary

Refactor replay to exclude idle from chapters

Improves summary clarity by removing idle periods and adding inactive period tracking, enhancing user experience and reducing noise in replay summaries.

Health Assessment

• • Fast cycle time (1.3h) and zero review lag indicate smooth process; AI assistance likely contributed to rapid, low‑friction changes.

Summary

Split Shadow Calls Between Managed Models

Enables controlled shadow testing of multiple model variants, improving experimentation accuracy without affecting production.

Health Assessment

• • Rapid iteration with a single review and merge within 5 hours indicates an efficient process; AI-assisted review helped streamline the changes.

Summary

Add custom image selection for cloud environments

Allows users to pick custom images for cloud environments, enhancing deployment flexibility and visual consistency.

Health Assessment

• • Fast review and merge with minimal changes, indicating low risk and high confidence in implementation.

Summary

Add custom sandbox image management UI

Enables users to view, create, and manage custom sandbox images via the web UI, improving developer workflow and reducing setup time.

Health Assessment

• • Rapid approval with minimal iterations; front‑end only change behind a feature flag.

Summary

Add marketing analytics cross-sell for paid traffic

Adds a marketing analytics card that helps users connect ad sources and analyze paid traffic, improving attribution and revenue insights.

Health Assessment

• • Fast cycle time of 3.2h with a single review, minimal rework, and a feature gated behind a flag. Despite a large code addition (~1160 lines), extensive tests and CI checks mitigate risk.

Summary

Remove unused watched column from replay collections

Eliminates an unused database column, reducing storage and simplifying the UI for replay collections.

Health Assessment

• • Fast review and merge cycle (under 2 hours) with minimal code changes, indicating low complexity and risk.

Summary

Align new view menu icons and clarify copy

Improves UI alignment and provides clearer descriptions for new view types, enhancing user experience.

Health Assessment

• • Quick fix with minimal changes, AI-assisted, merged within 45 minutes.

Summary

Fix Signals metric window timing

Ensures metric verdicts reflect full post‑resolution windows, improving data accuracy for users.

Health Assessment

• • Rapid iteration with multiple commits and reviews, but quick resolution indicates efficient collaboration.

Summary

Split NodeJS into import-derived merge queue lanes

Improves CI merge queue efficiency by partitioning NodeJS code into separate lanes, reducing unrelated PR interference.

Health Assessment

• • Fast cycle time and minimal review iterations indicate a smooth process; the large code change is confined to CI logic and poses low risk.

Summary

Add agent settings search by desktop names

Enables desktop users to find settings by familiar names, improving usability and reducing confusion.

Health Assessment

• • Fast approval cycle with minimal rework; feature behind a flag, low risk to production.

Summary

Add Slack DM toggle for task comments in web

Users can now enable or disable Slack DM notifications for task comments directly from the web settings, improving control and reducing unwanted alerts.

Health Assessment

• • Rapid approval with minimal review, indicating a straightforward change with low risk.

Summary

Add cloud environment management UI

Provides a web interface for users to list, create, edit, and delete sandbox environments, eliminating the need for Desktop-only access.

Health Assessment

• • Fast cycle time and immediate approvals suggest a smooth integration with minimal friction.

Summary

Add usage page for agent spend

Adds a new settings page that displays PostHog agent spend metrics for today, month, and 30 days, enabling users to monitor costs directly in the web UI.

Health Assessment

• • Fast review and merge with minimal rework, straightforward frontend change with tests.

Summary

Rename agent settings and add visibility tags

Users can now see who each setting applies to, improving clarity and reducing confusion.

Health Assessment

• • Fast review cycle, minimal changes, low risk

Summary

Fix quill chat grouping and row styling

Aligns the PostHog AI chat UI with desktop behavior, improving consistency and user experience.

Health Assessment

• • The PR required multiple iterations after a quick review, indicating complex UI changes; the large number of lines and commits suggests high effort but the rapid review cycle mitigates risk somewhat.

Summary

Limit token team override to team routes

Ensures project tokens do not incorrectly affect organization-level API routes, preventing permission mismatches and potential data leakage.

Health Assessment

• • Quick, single-commit fix with no rework; fast review and merge cycle indicates low complexity and minimal risk.

Summary

Remove PostHog AI deep links from Replay Vision scanner prompts

Updates blog content to remove broken AI links, improving user guidance.

Health Assessment

• • Quick documentation edit with minimal review and no rework; fast cycle time indicates low impact on product.

Summary

Improve engineering analytics performance by bounding shell filter

This change reduces engineering analytics page load times by optimizing SQL queries, improving developer productivity and reducing server load.

Health Assessment

• • The PR was reviewed and merged within 2 hours, indicating a smooth review process.
• • The performance improvements are internal and low risk, with comprehensive tests covering the changes.

Summary

Remove legacy person deletion order

Eliminates dead code for person deletion, simplifying logic and removing obsolete metrics, with no impact on user-facing behavior.

Health Assessment

• • Fast cycle time and minimal review comments indicate smooth integration, but the large code removal introduces high risk due to scope.

Summary

Retire precomputed attribution reader

Remove obsolete cached attribution reader, simplifying marketing attribution queries and reducing maintenance overhead.

Health Assessment

• • Fast turnaround with minimal review iterations indicates a straightforward maintenance change.