Commits

Project activity for July 05, 2026:

agents.happitec.com

22 commits

• Implemented a new conceptual status diff system, introducing persistent, CloudKit-synchronized tier filters (e.g., Important, Review) and device-specific tracking of last-viewed states to enable on-demand, server-backed diffs instead of full syncs.
• Redesigned the Status UI to support three-tier rendering with per-tier filter pills, repo reference chips, and improved visual hierarchy (e.g., newspaper icon for Summary), while aligning behavior and labels (e.g., ‘Status’ instead of ‘What changed’) to legacy UX expectations.
• Fixed critical diff accuracy issues, including fractional-second since timestamps for precise matching, proper keying of diff groups by name (avoiding index instability), and eliminating manual animations in favor of consistent, single-parent collapse semantics and legacy-style row transitions.
• Refactored core logic—introducing StatusDiffSectionView, StatusTier with wire-fallback mapping, and fetchDiff with non-advancing markStatusViewed, alongside extracted shared components for needs-input lists and summary text to reduce duplication and improve maintainability.
• Enhanced data integrity & sync via CloudKit-backed filter persistence (SyncStore) and per-device last-viewed state, ensuring diffs are context-aware and resilient across devices and app restarts.

dash-agent

6 commits

• Consolidated workflow gates: Advisory review and administrivia checks were folded into the main trigger-feed/routing flow, creating a unified and more efficient feedback loop for message processing (#129).
• Improved end-of-sequence handling: Server-side detection of sequence completion—via the new Important notification tier—replaced earlier client-driven approaches; combined with a dedicated Stop hook in dash-agent-contained, this enables reliable, low-latency termination without relying on downstream timeouts or polling.
• Precision fix for status diffing: Resolved a subtle boundary issue where since was incorrectly interpreted as an exact key rather than a time-based range boundary (floorKey precision invariant remains correct), ensuring accurate and consistent delta computation in time-series status tracking (#128).

package-multitouch-recorder

11 commits

• Headless Capture Infrastructure: Implemented a full headless capture workflow with a dedicated HeadlessCaptureController that assembles sessions including orientation handling, lifecycle management via file-poll-based stop signals, and proper error reporting (e.g., surface write failures instead of silently ignoring them).
• Release & Build Rigor: Enhanced the recorder release pipeline on RUNNER_LINUX (requiring bash ≥4 for mapfile) to enforce v-prefix tags, disable xcodebuild, and automate building, verifying, and attaching the capture-inject dylib—improving reproducibility and compatibility (e.g., iOS-17 dylib floor, otool -L validation).
• Stability & Bootstrap Fixes: Corrected capture initialization to defer bootstrap via the main queue, ensuring start() reliably executes; ensured clean-slate captures with robust orientation snapshot at stop time.
• Dynamic Target Support: Introduced a multitouch-aware dynamic product and build script for simulator and device targets, alongside publicized snapshot helpers to support headless targeting.
• Documentation & Integration-Free Capture: Added DocC articles and README content documenting integration-free capture workflows, emphasizing ease of adoption without requiring Xcode integration.

package-multitouch-tester

17 commits

• Implemented an integration-free gesture capture pipeline with a new simgesture command that orchestrates device interaction via simctl, including a CaptureCoordinator to manage recorder injection and timing.
• Strengthened reliability and validation: enforced a minimum --duration of 1.0s (with CLI tests and non-TTY safeguards), added SHA-256 verification against pinned release assets for recorder v0.7.0, and fixed incorrect injection logic (SIMCTL_CHILD_DYLD_INSERT_LIBRARIES instead of phantom --inject).
• Improved developer experience and documentation: added Capturability docs (including a DocC topic article), updated implementation plans to reflect integration-free design decisions, and captured requirements for timestamp fidelity, orientation, and stop-signal semantics.
• Hardened the tooling stack: corrected fallback paths for --build-dylib, resolved dimension mismatch bugs in CaptureCoordinator, and pinned/updated recorder dependencies consistently across CI and dev workflows.