Modify ExternalCursorAiUsageClient to handle Cursor's updated response shape (derive committer and usage date), send admin notifications when a date record contains no identifiable data, and add a migration to remove uniqueness of identity per provider. Result: more robust usage attribution and immediate alerts for unexpected empty Cursor responses.

Added a new API reference partial (_surveys.html.haml) that documents Surveys endpoints, parameters and responses. Change was released to production on Jan 16, 2026. Impact: provides authoritative, in-repo documentation to help developers integrate with and maintain the Surveys API.

Replaced the static hero with an aspect-ratio–controlled responsive image carousel and rebuilt the downloads page as a mobile-first two-column grid. Updated typography, spacing and breakpoints in app/assets/stylesheets/root/download_platform.scss and adjusted markup in app/views/root/download_platform.html.haml. Desktop download actions now route into the signup flow — reduces layout shift, improves mobile UX and visual stability, and boosts conversion potential.

Fixes bug in issue_browse where the 'All Time' group was rendered twice instead of 'All Time' followed by a 'Per Interval' grouping. Updated the view rendering and adjusted the integration test (test/integration/issue_browse_test.rb) to assert correct ordering and labels. Result: correct grouping in the UI and regression coverage to prevent recurrence.

Update phase1_analyzeQuery to detect and honor implicit or explicit user requests for the number of notes to return. Added test/search-agent/phase1-result-count.test.js with cases for parsing and enforcing requested result counts; plugin.js unchanged. Impact: search now respects user-specified result counts so result sets are correctly sized and more predictable.

Implements a retry-aware Phase 2 candidate collector (lib/functions/search/phase2-candidate-collection.js). Persists phase-2 state (rankedNotes, retryCount), batches keyword searches via MAX_SEARCH_CONCURRENCY, avoids reusing previously-rated notes, and tightens retry/sanity rules (skip at retry limit; retry when initial pass yields <50% of MIN_PHASE2_TARGET_CANDIDATES). Adds progress reporting and search-constant tweaks to reduce duplicates, improve candidate quality and observability.

Introduce pre-content heuristic scoring in lib/functions/search/search-candidate-note.js: add PRE_CONTENT_* constants, make originalContentLength null-safe, and add preContentMatchScore, scorePerKeyword and tagBoost. Implement ensureKeywordPreContentScores with helpers scoreFromNameMatch and scoreFromTagMatches to compute capped, weighted per-keyword scores (primary/secondary and tag boosts) and aggregate a preContentMatchScore to rank candidates by name/tags before fetching bodies, reducing ...

Fixes Copilot 403 token handling by adding retry windows, tolerating partial auth failures, and improving logging to surface 403 responses. Adds scaffolding for GitLab Issues (ExternalIssue::GitlabIssue and Gitlab_issue_tracker_integration) to seed future issue sync. Related to IN-497. Impact: reduces transient Copilot auth interruptions and provides a foundation for GitLab issue synchronization.

Update derive_series_segments in app/lib/presenters/concerns/ai_segment_concerns.rb to enforce an ascending sort of cohort series by usage (Least use → Most use). Multiple additions and small updates replace the prior ordering behavior so series are calculated and ordered deterministically. Impact: cohort charts now consistently present segments from least- to most-used, improving readability and analysis.

Update vendor/gems/gitclear_teams/app/models/team_repo.rb (ensure_cached_stat_rebuild_on_create_destroy) to enqueue a build of LLM usage metrics when a Team is created and may have LlmUsageMetrics. This schedules metric generation asynchronously on create (consistent with destroy handling), preventing missed initial LLM stats and improving accuracy of LLM usage reporting.

Return audit PRs based on the most-recent "merged_at" value when a closed-date filter is provided, and add new audit fields (business_minutes_awaited_first_review, business_minutes_cycle_time, business_minutes_lead_time, business_minutes_open_to_closed, first_commented_at, first_commit_authored_at, post_merge_delta, squash_commit_sha). Changes touch V1::AuditsController and PullRequestConcerns::PullRequestJsonConcerns so API audit responses and closed-date queries are more accurate for report...

Tidied version of Opus-4.5 suggestion

Adjust app/jobs/assign_pull_request_post_merge_commit_work.rb to search for and handle commit-message lines that indicate updates to squashed PR commits, with a small test tweak in test/unit/pull_request_post_merge_commit_test.rb. This fixes the failing test and ensures post-merge commit assignment correctly recognizes squashed-PR update lines so ownership is assigned reliably.

Return early from RepoImportProgress#calculate_percent_complete (app/models/repo_import_progress.rb) when the default branch has been processed as far back as the repository's processing epoch. This prevents unnecessary scanning/aggregation, fixes potential over-reporting, and reduces CPU/I/O during import progress calculation for more efficient, correct percent reporting.

Add an early-out to RepoImportProgress#calculate_percent_complete (app/models/repo_import_progress.rb) so that if the default branch has been processed as far back as the repository's provider_created_at the calculation returns immediately. This avoids scanning older history, prevents overcounting and wasted work, and improves import progress accuracy and performance.

Replace timeOffHtml with a timeOffArray: move derive_time_off_array out of vendor/gems/commit_activity_browser/app/lib/bubble_committer_concern.rb into app/javascript/bundles/cab-viewer/cab-committers-frame.js and update committer tooltip/content to consume the array. Prevents HTML-escaping problems in CAB and ensures correct time-off rendering.

Fixes a bug where the PR comment fallback time used 30.seconds without dividing by 1.minute, producing implausible minute estimates that propagated to PullRequestReviewer. Updated the calculation in code_comment_extra (with unit test adjustments) so minutes are computed correctly, restoring realistic comment-time estimates and preventing inaccurate reviewer timing metrics.

Add a sparkline to the committer avatar hover (updated in app/javascript/bundles/cab-viewer/cab-committers-frame.js). renderSparkline and tooltip markup were extended to render a 12‑month delta mini-chart inside the hovercard, giving reviewers a compact visual of recent committer activity for faster context.