Pull Request Explorer

Summary

Fix Kafka partition key to spread traffic

Adjusts Kafka message key to include property key, distributing high‑volume team traffic across partitions and reducing pipeline bottlenecks.

Health Assessment

• • Rapid review and minimal changes indicate a straightforward, low‑risk fix that quickly resolved a production bottleneck.

Summary

Add auto-reconnect for notebook SSE and warning

Ensures real‑time collaboration remains active by reconnecting after server‑initiated stream rotations and displays a warning when the stream is down, improving user experience and reducing silent failures. This change prevents loss of live updates after the 5‑minute server timeout.

Health Assessment

• • The PR took over four days to merge, with a single round of rework after review, indicating moderate complexity but no major blockers.

Summary

Add blog post on human support

Adds a blog post discussing the future of customer support in an AI‑driven world, emphasizing the importance of human interaction and technical expertise.

Health Assessment

• • PR required several small edits after the initial review but was approved quickly and involved minimal code changes.

Summary

Align hosted editor translation copy with original-language model

Fixes copy inconsistency in hosted survey editor, aligning terminology with the base language model and removing redundant UI elements.

Health Assessment

• • The PR was reviewed and merged quickly after a single commit, but the overall cycle time was long due to queue delays.

Summary

Add tabbed expanded row on survey responses

Users can now view detailed survey responses with a tabbed layout, improving the inspection of underlying properties and flags.

Health Assessment

• • The PR was reviewed and merged quickly, with no significant rework or discussion, indicating a smooth development process.

Summary

Refactor taxonomic filter logic extraction

Extracts max AI context logic into a dedicated file, improving modularity and maintainability of the taxonomic filter component.

Health Assessment

• • Quick review cycle with minimal changes and no blockers, indicating a straightforward refactor with low risk to production.

Summary

Add voice-mode prompt for hands‑free TTS

Enables natural‑sounding TTS responses by instructing the LLM to spell out numbers, currencies, and percentages, and to avoid markdown when users speak via hands‑free mode.

Health Assessment

• • The PR received a quick review and approval, but the merge was delayed, resulting in a long overall cycle time.

Summary

Refactor taxonomic filter logic extraction

Extracts event metadata group definitions into a dedicated logic file, improving modularity and maintainability for the taxonomic filter component.

Health Assessment

• • Quick mechanical refactor with minimal code changes, fast review turnaround, and no blockers.

Summary

Guard HeatmapButton against non-string values

Prevents runtime crashes in Web Analytics heatmap rows by validating input types, ensuring stable UI rendering.

Health Assessment

• • Quick fix with minimal changes, resolved within hours, indicating efficient review and low risk.

Summary

Add dry run mode to GlobalRateLimiter

Enables visibility into token:distinct_id rate limiting behavior before shipping, allowing safer rollout of the new RFC-decided limiter.

Health Assessment

• • The PR was reviewed and merged within minutes, indicating low friction and high confidence in the change.

Summary

Swap status dot for state emoji in hands‑free surface

This cosmetic UI change improves visual alignment and clarity of the hands‑free status indicator, enhancing user experience in the Max feature.

Health Assessment

• • The PR had a long cycle time due to a delayed review, but the change was straightforward and had minimal risk.

Summary

Add lazy precomputation for web stats breakdowns

Enables precomputed web analytics tiles, reducing query load and improving dashboard responsiveness for users.

Health Assessment

• • The PR required extensive backend and frontend changes with several iterations after the initial review, indicating moderate complexity and a higher likelihood of integration challenges.

Summary

Add synthetic test interviewee link

Enables topic authors to test AI voice interviews without consuming real interviewee slots, improving QA and user experience.

Health Assessment

• • Significant rework with multiple AI‑driven refactors and security adjustments; long cycle time and many review rounds indicate high complexity and risk.

Summary

Clamp LLM evaluation sampling percentage to 2 decimal places

Fixes floating‑point drift in the LLM evaluation sampling controls, ensuring consistent numeric formatting across the UI and stored state.

Health Assessment

• • Fast review and minimal changes; AI‑assisted code generation reduced manual effort.

Summary

Refactor taxonomic filter logic extraction

Extracts HogQL expression group logic into a dedicated file, improving modularity and maintainability of the taxonomic filter component.

Health Assessment

• • The PR involved a mechanical refactor with minimal code changes and a quick review turnaround, indicating low risk and high maintainability.

Summary

Extend test coverage for regex flag validation

Adds comprehensive tests for feature flag regex validation, ensuring future refactors cannot bypass validation and maintain system integrity.

Health Assessment

• • Rapid review and merge with minimal changes indicate low complexity and risk.

Summary

Health Assessment

Summary

Add missing event filter option

Adds a new $logs_alert_errored event option to the product event filter list for log alert Hog Function configurations, enabling filtering on log alert error conditions.

Health Assessment

• • Quick, single commit, approved immediately, minimal changes, low risk.

Summary

Add blog post on migrating flag evaluation to Rust

Adds a blog post detailing the migration of local flag evaluation from Django to Rust, enhancing performance and maintainability.

Health Assessment

• • The PR had a long cycle time due to multiple merges and updates after the initial review, but the review process itself was quick.

Summary

Reject malformed tags query param with 400

Prevents silent failures by validating tags query param, returning 400 for malformed input, improving API reliability.

Health Assessment

• • Fast turnaround with a single review and minimal changes, ensuring reliable API behavior.

Summary

Extract tagged-item logging to registry

Simplifies activity logging by centralizing logic, enabling easier addition of new taggable models and reducing code duplication.

Health Assessment

• • Fast cycle time of 1.6 hours with a single commit and one approval indicates a smooth, low-risk refactor.

Summary

Fix XML parsing fallback in trace viewer

Ensures tool responses containing malformed XML are displayed as plain text, preventing error messages that hide content.

Health Assessment

• • The PR was reviewed immediately and required only a single commit after the first review, indicating a straightforward fix. The minimal code change and small scope reduce risk and impact on the codebase.

Summary

Add deprecation banner for revenue analytics

Announces removal of the revenue analytics dashboard, informing users of the upcoming change with a warning banner.

Health Assessment

• • Fast, AI‑generated UI copy update with minimal risk and quick review turnaround.

Summary

Restrict personal spend product to posthog_code

Enforces that the LLM spend API only accepts the supported product 'posthog_code', returning 400 for unsupported values, improving API contract clarity and preventing misleading data.

Health Assessment

• • The PR required rapid iteration with multiple commits and comments to resolve breaking API contract concerns, indicating significant review friction despite a short overall cycle time.

Summary

Fix tagger save failures on fractional percentages and blank tags

Ensures tagger configurations persist correctly, preventing silent save failures and improving user experience.

Health Assessment

• • Fast turnaround with minimal changes and comprehensive tests, indicating a straightforward bug fix with low risk.

Summary

Reduce width of unique visitors chart

Improves dashboard readability by adjusting chart width.

Health Assessment

• • Quick approval with minimal changes indicates low complexity and high confidence in UI adjustment.

Summary

Revert boolean filter normalization change

Restores original boolean property filtering logic, preventing regressions in insights queries.

Health Assessment

• • Revert performed quickly with minimal changes and a single AI-assisted review, indicating low complexity and risk.

Summary

Refactor taxonomic filter logic extraction

Extracts cohort taxonomic group definitions into a dedicated logic file, improving modularity and maintainability.

Health Assessment

• • Quick refactor with minimal changes, approved after brief review, indicating low complexity and high confidence.

Summary

Use path‑style object storage locally

Ensures local MinIO endpoints use path‑style addressing, preventing SDK URL resolution errors.

Health Assessment

• • Quick fix with a single line change, no review required, minimal risk to production.

Summary

Update LLM pricing models

Automated update of LLM cost data to keep pricing accurate for analytics.

Health Assessment

• • Fast, automated PR with minimal review and low risk of impact.

Summary

Deprecate top_traces, return empty

Deprecated the top_traces field in LLM analytics responses, always returning an empty list to eliminate non-actionable trace data and reduce UI clutter.

Health Assessment

• • Quick turnaround with immediate review and minimal rework indicates a straightforward bugfix with low complexity.

Summary

Fix proxy diagnose error handling

Returns actionable error when reverse proxy diagnose fails, improving user experience and debugging capabilities.

Health Assessment

• • The PR was reviewed and merged quickly, indicating a smooth development process. The changes are focused on improving error handling, which reduces the risk of introducing new bugs.

Summary

Move sender to send_document, use owner on create

Aligns PandaDoc API usage, ensuring correct sender identity and document ownership, improving compliance and user experience.

Health Assessment

• • Fast review and merge with minimal rework, indicating a straightforward bug fix.

Summary

Document dashboard sharing activity log events

Updates documentation to reflect new dashboard sharing activity log events added in PR #58964, helping users track audit logs for sharing actions.

Health Assessment

• • Long cycle time indicates low priority or slow review process; minimal code changes reduce risk.

Summary

Remove unused AI evals CI workflow

Eliminates an unused CI workflow and its Dockerfile, reducing maintenance overhead.

Health Assessment

• • Quick removal of unused CI workflow with minimal review and no tests required.

Summary

Remove sampling from docs nav

Users no longer see the sampling entry in the docs navigation, and existing links are redirected to avoid 404 errors.

Health Assessment

• • Quick fix with minimal changes, resolved within minutes, no rework needed.

Summary

Clean up hosted survey translations picker

Improves UI clarity and usability by removing low‑contrast label and preventing unintended auto‑focus in the translation picker, enhancing the survey editor experience.

Health Assessment

• • Quick cosmetic fix with minimal risk, approved immediately, no tests added.

Summary

Add date partition to S3 batch folder

Adds a date-based partition to S3 batch folders to enable automated cleanup of old batches, ensuring consumers have sufficient time to process data.

Health Assessment

• • Quick review but long cycle time due to delayed final commit, suggesting scheduling or testing delays.

Summary

Collapse unmatched path labels to unknown

Reduces Prometheus series cardinality by collapsing unmatched request paths into a single "unknown" label, improving query performance and cost.

Health Assessment

• • Fast turnaround with minimal changes and no review comments indicates low complexity.

Summary

Restore default port 8787 for local MCP dev

Reverts the default port to 8787 so existing local MCP setups and client configurations continue to work without manual changes.

Health Assessment

• • Fast turnaround with minimal changes; AI authored; no manual testing reported.

Summary

Add person_properties field to DataDeletionRequest

Enables operators to specify keys to drop from events.person_properties, expanding data deletion capabilities.

Health Assessment

• • The PR received a quick review but required many subsequent commits, reflecting active iteration before final merge after 88 hours.

Summary

Split caches and add session tracking

Reorganizes cache key scopes and introduces session ID tracking to improve data isolation and enable analytics.

Health Assessment

• • Fast review and minimal rework indicate a smooth, low-risk change.

Summary

Rename tombstone to free key on reuse

Fixes misleading errors when creating or renaming feature flags that collide with soft‑deleted flags, improving user experience and reducing support tickets.

Health Assessment

• • The PR required multiple commits to cover edge cases around soft‑deleted flags, but the final change was straightforward and did not introduce regressions.

Summary

Shorten session ID tooltip

Simplifies the session detail tooltip to a single line, improving readability and reducing UI clutter.

Health Assessment

• • Quick, single‑commit change with immediate approval and no downstream impact.

Summary

Fix MCP exec tool metrics in Grafana

Ensures Prometheus metrics report the actual tool invoked in single‑exec mode, improving monitoring accuracy for Grafana dashboards.

Health Assessment

• • Rapid 1.7‑hour cycle with a single AI‑assisted review and minimal rework indicates a smooth, low‑risk bugfix.

Summary

Deduplicate UUIDs in Experiment Operations

Fixes duplicate metric UUID errors that prevented users from editing or deleting experiments, restoring data integrity and user experience.

Health Assessment

• • Single review round with quick approval after three commits; moderate code changes but no major risk.

Summary

Show loading state while session tool calls reload

Fixes stale tool call display by rendering loading skeleton immediately when session changes, improving user experience during data fetches.

Health Assessment

• • Quick fix with minimal changes, approved immediately, low risk.

Summary

Clarify matured users label covers retention

The UI label now clearly indicates that the setting applies to both conversion and retention metrics, reducing user confusion and improving feature clarity.

Health Assessment

• • Quick single-line change with immediate approval, indicating minimal risk and high confidence in correctness.

Summary

Update mobile session replay copy to reflect GA status

Clarifies that mobile session replay is now generally available on Android, iOS, React Native, and Flutter, removing beta language from documentation to reduce confusion and improve user understanding.

Health Assessment

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

Summary

Remove experiments-ship-variant-release-mode flag

Eliminates an obsolete feature flag, simplifying experiment release logic and reducing maintenance overhead.

Health Assessment

• • Quick cleanup with minimal review, indicating low complexity and high confidence.