agent-usage-tracker 0.1.0__tar.gz → 0.1.1__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (259) hide show
  1. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/PKG-INFO +2 -2
  2. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/README.md +1 -1
  3. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/config.example.toml +2 -2
  4. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/improvement.md +61 -9
  5. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/local-labels.md +7 -2
  6. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/reference.md +3 -3
  7. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/__init__.py +1 -1
  8. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/cli.py +8 -7
  9. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/config.py +47 -1
  10. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/git_days.py +3 -14
  11. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/history_steers.py +27 -8
  12. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/cli.py +7 -0
  13. agent_usage_tracker-0.1.1/src/agent_usage_tracker/improve/groups.py +260 -0
  14. agent_usage_tracker-0.1.1/src/agent_usage_tracker/improve/inbox.py +70 -0
  15. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/judge.py +29 -1
  16. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/rubric.py +64 -26
  17. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/schedule.py +16 -28
  18. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/backends.py +15 -5
  19. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/bakeoff.py +45 -34
  20. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/core.py +14 -8
  21. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/evalset.py +26 -23
  22. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/gold.py +14 -10
  23. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/replay.py +21 -5
  24. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/run.py +23 -19
  25. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/tasks.py +14 -12
  26. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/teach.py +6 -4
  27. agent_usage_tracker-0.1.1/tests/improve/test_groups.py +275 -0
  28. agent_usage_tracker-0.1.1/tests/improve/test_inbox.py +76 -0
  29. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_rubric.py +68 -3
  30. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_schedule.py +59 -0
  31. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/label_helpers.py +18 -5
  32. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_history_steers.py +28 -2
  33. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label.py +5 -5
  34. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_bakeoff.py +63 -25
  35. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_context.py +5 -5
  36. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_core.py +25 -4
  37. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_evalset.py +49 -27
  38. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_gap.py +9 -8
  39. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_gold.py +8 -8
  40. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_replay.py +22 -1
  41. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_tasks.py +32 -12
  42. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_teach.py +1 -1
  43. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_writing.py +3 -3
  44. agent_usage_tracker-0.1.1/tests/pipeline/test_ledger_zone.py +51 -0
  45. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_sink_sql.py +22 -0
  46. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/.gitignore +0 -0
  47. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/CONTRACT.md +0 -0
  48. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/DEPLOY.md +0 -0
  49. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/LICENSE +0 -0
  50. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/collectors-to-add.md +0 -0
  51. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/harness-coverage.md +0 -0
  52. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/human-html/2026-11-08-understanding-implicit-state-machines.html +0 -0
  53. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/human-html/2026-11-09-status-state-reliability-fixes.html +0 -0
  54. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/human-html/docs-index.json +0 -0
  55. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/human-html/index.html +0 -0
  56. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/improvement-studies.md +0 -0
  57. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/new-laptop.md +0 -0
  58. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/observations.md +0 -0
  59. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/research/codexbar-learnings.md +0 -0
  60. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/research/otel-lane.md +0 -0
  61. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/research/sdk-usage-capture.md +0 -0
  62. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/superpowers/plans/2026-07-22-aut-showcase-metrics.md +0 -0
  63. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/docs/superpowers/specs/2026-07-22-aut-showcase-metrics-design.md +0 -0
  64. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/README.md +0 -0
  65. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/claude/commands/outcome.md +0 -0
  66. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/launchd/aut-scan.sh +0 -0
  67. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/launchd/dev.aut.backup.plist +0 -0
  68. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/launchd/dev.aut.improve-audit.plist +0 -0
  69. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/launchd/dev.aut.retry.plist +0 -0
  70. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/launchd/dev.aut.scan.plist +0 -0
  71. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/observations/Dockerfile +0 -0
  72. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/observations/Dockerfile.dockerignore +0 -0
  73. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/observations/README.md +0 -0
  74. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/observations/aut-observations.service +0 -0
  75. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/sql/turn-snapshot-grant.sql +0 -0
  76. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/systemd/agent-usage-tracker.service +0 -0
  77. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/systemd/agent-usage-tracker.timer +0 -0
  78. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/packaging/windows/register-task.ps1 +0 -0
  79. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/pyproject.toml +0 -0
  80. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/backup.py +0 -0
  81. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/__init__.py +0 -0
  82. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/_util.py +0 -0
  83. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/activity_common.py +0 -0
  84. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/activitywatch.py +0 -0
  85. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/claude_code.py +0 -0
  86. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/claude_code_activity.py +0 -0
  87. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/codex.py +0 -0
  88. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/commits.py +0 -0
  89. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/cursor.py +0 -0
  90. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/cursor_api.py +0 -0
  91. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/cursor_cli.py +0 -0
  92. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/cursor_sdk.py +0 -0
  93. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/generic_sdk.py +0 -0
  94. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/pi.py +0 -0
  95. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/session_digest.py +0 -0
  96. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/collectors/t3.py +0 -0
  97. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/coverage.py +0 -0
  98. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/coverage_catalog.json +0 -0
  99. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/cursor_run.py +0 -0
  100. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/diagnostics.py +0 -0
  101. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/episode_store.py +0 -0
  102. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/event.py +0 -0
  103. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/git_day_cache.py +0 -0
  104. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/git_safe.py +0 -0
  105. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/harness_inventory.py +0 -0
  106. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/__init__.py +0 -0
  107. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/case_view.py +0 -0
  108. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/desktop.py +0 -0
  109. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/evidence.py +0 -0
  110. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/proposals.py +0 -0
  111. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/rubrics/avoidable_recovery.md +0 -0
  112. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/rubrics/code_quality.md +0 -0
  113. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/rubrics/completion_supported.md +0 -0
  114. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/rubrics/redundant_tests.md +0 -0
  115. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/rubrics/verbosity.md +0 -0
  116. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/runtime.py +0 -0
  117. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/search_check.py +0 -0
  118. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/store.py +0 -0
  119. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/studies.py +0 -0
  120. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/study_cli.py +0 -0
  121. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/study_decision.py +0 -0
  122. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/study_followup.py +0 -0
  123. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/t3.py +0 -0
  124. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/trial.py +0 -0
  125. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/improve/validation.py +0 -0
  126. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/inspector.py +0 -0
  127. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/__init__.py +0 -0
  128. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/baseline.py +0 -0
  129. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/clef.py +0 -0
  130. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/context.py +0 -0
  131. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/covered_run.py +0 -0
  132. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/local.py +0 -0
  133. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/local_parallel.py +0 -0
  134. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/merge.py +0 -0
  135. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/label/outcome_coverage.py +0 -0
  136. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/local_page.py +0 -0
  137. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/manual_time.py +0 -0
  138. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/net.py +0 -0
  139. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/observation_cli.py +0 -0
  140. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/observation_report.py +0 -0
  141. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/observation_server.py +0 -0
  142. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/observations.py +0 -0
  143. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/outcome_git.py +0 -0
  144. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/outcomes.py +0 -0
  145. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/pricing.py +0 -0
  146. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/push.py +0 -0
  147. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/redact.py +0 -0
  148. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/repo_keys.py +0 -0
  149. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/review.py +0 -0
  150. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/scan.py +0 -0
  151. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/scheduled.py +0 -0
  152. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sdk.py +0 -0
  153. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink.py +0 -0
  154. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/README.md +0 -0
  155. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/apply.sh +0 -0
  156. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/compose.yaml +0 -0
  157. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/env.example +0 -0
  158. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/grafana/dashboards/agents-effort.json +0 -0
  159. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/grafana/dashboards/agents-usage.json +0 -0
  160. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/grafana/provisioning/dashboards/aut.yaml +0 -0
  161. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/grafana/provisioning/datasources/agents.yaml +0 -0
  162. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/sql/01_schema.sql +0 -0
  163. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/sql/02_price.sql +0 -0
  164. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/sql/03_effort.sql +0 -0
  165. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/sql/04_grants.sql +0 -0
  166. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sink_stack/sql/refresh.sql +0 -0
  167. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/sources.py +0 -0
  168. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/state.py +0 -0
  169. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/src/agent_usage_tracker/telemetry.py +0 -0
  170. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/codex_pi_helpers.py +0 -0
  171. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_activity.py +0 -0
  172. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_activitywatch.py +0 -0
  173. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_activitywatch_macos.py +0 -0
  174. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_claude_code.py +0 -0
  175. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_claude_profiles.py +0 -0
  176. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_codex.py +0 -0
  177. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_codex_pi.py +0 -0
  178. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_collectors.py +0 -0
  179. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_commits.py +0 -0
  180. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_cursor.py +0 -0
  181. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_cursor_api.py +0 -0
  182. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_cursor_cli.py +0 -0
  183. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_cursor_cli_cache.py +0 -0
  184. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_cursor_sdk.py +0 -0
  185. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_generic_sdk.py +0 -0
  186. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_links.py +0 -0
  187. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_links_claude.py +0 -0
  188. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_pi.py +0 -0
  189. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_read_horizon.py +0 -0
  190. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_session_digest.py +0 -0
  191. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_source_registry.py +0 -0
  192. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/collectors/test_t3.py +0 -0
  193. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/conftest.py +0 -0
  194. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/__init__.py +0 -0
  195. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/conftest.py +0 -0
  196. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_case_view.py +0 -0
  197. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_cli.py +0 -0
  198. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_desktop.py +0 -0
  199. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_evidence.py +0 -0
  200. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_evidence_citations.py +0 -0
  201. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_judge.py +0 -0
  202. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_judge_choices.py +0 -0
  203. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_judge_deadline.py +0 -0
  204. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_judge_native_deadline.py +0 -0
  205. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_judge_sampling.py +0 -0
  206. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_judge_stall.py +0 -0
  207. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_proposals.py +0 -0
  208. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_runtime.py +0 -0
  209. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_search_check.py +0 -0
  210. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_store.py +0 -0
  211. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_studies.py +0 -0
  212. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_studies_decision.py +0 -0
  213. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_t3.py +0 -0
  214. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_trial.py +0 -0
  215. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/improve/test_validation.py +0 -0
  216. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_context_paths.py +0 -0
  217. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_covered_run.py +0 -0
  218. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_cursor_origin.py +0 -0
  219. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_baseline.py +0 -0
  220. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_local.py +0 -0
  221. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_models.py +0 -0
  222. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_parallel.py +0 -0
  223. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_label_reliability.py +0 -0
  224. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/label/test_outcome_coverage.py +0 -0
  225. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/outcomes/test_outcome_git.py +0 -0
  226. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/outcomes/test_outcome_reliability.py +0 -0
  227. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/outcomes/test_outcomes.py +0 -0
  228. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/fake_psycopg.py +0 -0
  229. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_backup.py +0 -0
  230. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_cli.py +0 -0
  231. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_coverage.py +0 -0
  232. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_coverage_diagnostics.py +0 -0
  233. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_coverage_scan.py +0 -0
  234. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_cursor_run.py +0 -0
  235. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_delivery_cli.py +0 -0
  236. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_git_day_cache.py +0 -0
  237. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_git_days.py +0 -0
  238. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_harness_inventory.py +0 -0
  239. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_heartbeat.py +0 -0
  240. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_inspect.py +0 -0
  241. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_local_page.py +0 -0
  242. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_net.py +0 -0
  243. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_pricing.py +0 -0
  244. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_push.py +0 -0
  245. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_push_outbox.py +0 -0
  246. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_push_snapshots.py +0 -0
  247. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_redact.py +0 -0
  248. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_repo_keys.py +0 -0
  249. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_review.py +0 -0
  250. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_scan_wrapper.py +0 -0
  251. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_sessions.py +0 -0
  252. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_sink.py +0 -0
  253. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/pipeline/test_state.py +0 -0
  254. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/test_observation_cli.py +0 -0
  255. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/test_observation_report.py +0 -0
  256. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/test_observation_server.py +0 -0
  257. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/test_observations.py +0 -0
  258. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/test_observe_entrypoint.py +0 -0
  259. {agent_usage_tracker-0.1.0 → agent_usage_tracker-0.1.1}/tests/test_telemetry.py +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.5
2
2
  Name: agent-usage-tracker
3
- Version: 0.1.0
3
+ Version: 0.1.1
4
4
  Summary: Passive, subscription-safe tracker of coding-agent work (Claude Code, Codex, Cursor, Pi): usage, sessions, activity, active time, outcomes and local labels. Reads local logs, normalizes, pushes to a self-hosted sink.
5
5
  Project-URL: Homepage, https://github.com/rhnfzl/agent-usage-tracker
6
6
  Project-URL: Issues, https://github.com/rhnfzl/agent-usage-tracker/issues
@@ -193,7 +193,7 @@ They can also classify your instructions and infer whether a piece of work is re
193
193
 
194
194
  Local estimates read Claude Code, Codex, Cursor CLI, and Pi inputs.
195
195
  Prompts and final replies stay inside the episode's time bounds.
196
- Daily work also stops at local midnight, so shared sessions cannot mix separate work days.
196
+ Daily work also stops at midnight in `ledger_timezone`, so shared sessions cannot mix separate work days.
197
197
  Cursor replies are omitted because their stores lack reliable per-reply timestamps.
198
198
  Cases without sufficient evidence stay unclassified; uncertain model decisions stay unknown.
199
199
 
@@ -161,7 +161,7 @@ They can also classify your instructions and infer whether a piece of work is re
161
161
 
162
162
  Local estimates read Claude Code, Codex, Cursor CLI, and Pi inputs.
163
163
  Prompts and final replies stay inside the episode's time bounds.
164
- Daily work also stops at local midnight, so shared sessions cannot mix separate work days.
164
+ Daily work also stops at midnight in `ledger_timezone`, so shared sessions cannot mix separate work days.
165
165
  Cursor replies are omitted because their stores lack reliable per-reply timestamps.
166
166
  Cases without sufficient evidence stay unclassified; uncertain model decisions stay unknown.
167
167
 
@@ -10,8 +10,8 @@
10
10
  machine = "laptop"
11
11
 
12
12
  # What a "day" is in the sink's views: the IANA zone the sink was installed with (its AUT_TZ).
13
- # Default: Europe/Amsterdam. git-days counts days in it; history-steers uses it to find days
14
- # already counted. Both stop when it differs from the zone a sink_stack sink records.
13
+ # Default: Europe/Amsterdam. git-days, history-steers and the label commands count days in it.
14
+ # git-days and history-steers stop when it differs from the zone a sink_stack sink records.
15
15
  # ledger_timezone = "Europe/Berlin"
16
16
 
17
17
  # Which agents to collect. Omit to collect every collector except the opt-in activitywatch.
@@ -154,7 +154,8 @@ Invalid reviews retain their bounded, redacted responses in private `review_atte
154
154
  One schema repair is allowed. Both attempts count against the model-call and time budgets.
155
155
  They are not human labels or calibrated accuracy measurements.
156
156
  Ask a human only when evidence cannot settle a required fact, authority, or preference.
157
- The private `inbox.html` shows escaped review text and case identifiers.
157
+ The private `inbox.html` shows escaped review text and case identifiers, and the top 10
158
+ failure groups (next section).
158
159
 
159
160
  ## Rubrics
160
161
 
@@ -221,17 +222,67 @@ counts, even without a version bump. Each judgment records `rubric_version`, `ru
221
222
  `rubric_snapshot`. Judgments made before markdown rubrics have none of these. Their diagnosis
222
223
  policy holds the old rubric text, and `rubric.legacy_version` maps it to version 1 when it equals
223
224
  the text the v1 files were moved from, else 0. Validation of an older diagnosis uses its old
224
- identity, so its result does not change. The first night after the update re-votes the pending
225
- audit's cases once, because the rubric identity changed.
225
+ identity, so its result does not change. Because the rubric identity changed, the first night
226
+ or two after the update re-vote the pending audit's cases once: under a tight call budget the
227
+ blocking re-votes alone can fill the first night, and the next night finishes from the saved
228
+ votes. Older judgments outside the pending audit are re-voted once, when a later audit picks
229
+ their cases up again.
226
230
 
227
231
  The diagnosis judges rubric by rubric: blocking rubrics first, each group in priority order, so
228
232
  the others get only what is left. Before markdown rubrics it went case by case; after a cut-off
229
233
  night, the first rubric may now have more cases judged than the second. `complete` counts
230
234
  blocking work only, so with no blocking rubric enabled it is complete at once. `finished` and
231
235
  `total` count blocking judgments; `optional_finished` and `optional_total` count the rest;
232
- `verdicts` counts both. `reason` is `budget` when any work, blocking or not, was left. The scheduled worker releases its pending audit at the first
236
+ `verdicts` counts both. `reason` is `budget` when any work, blocking or not, was left.
237
+ An error during non-blocking work never fails the diagnosis. An invalid model vote skips that
238
+ case only; an error outside a vote, such as low memory, stops the remaining non-blocking
239
+ rubrics. `reason` is then `optional_error` (or `budget` when blocking work was also left) and
240
+ `optional_skipped` lists each rubric and error type. A skipped case is tried again next night.
241
+ An error during blocking work still fails the run.
242
+ The scheduled worker releases its pending audit at the first
233
243
  checkpoint where blocking work is complete, so a slow non-blocking rubric never holds the next
234
- audit, even if the run is cut off later.
244
+ audit, even if the run is cut off later. When the run ends it writes the released state once
245
+ more, so an `audit --cycle` that started during the non-blocking work returns `superseded`.
246
+
247
+ ## Group recurring failures
248
+
249
+ ```bash
250
+ aut improve groups --days 14 --threshold 3
251
+ ```
252
+
253
+ This reads the stored judgments, runs no model, and saves a new report. It groups judgments whose case falls in the
254
+ last `--days` local work days (the date of the case's first timestamped event) and that are
255
+ `fail`, or `unknown` with at least one model voting `fail` at the rubric threshold (a split
256
+ vote). Only the newest judgment per case and rubric counts: the highest `rubric_version`, then
257
+ the one the latest diagnosis listed. A diagnosis that was killed journals no list, so a tie
258
+ among its judgments falls back to a fixed but arbitrary order. Cases dated after today are
259
+ left out.
260
+
261
+ The group key is a digest of five signals:
262
+
263
+ | Signal | From |
264
+ |---|---|
265
+ | `rubric` | the judgment |
266
+ | `cause` | the evidence review's cause, else `unreviewed` |
267
+ | `tool` | the call answered by the anchor result: the first error result the review cites, else the first result it cites, else the case's last error result. MCP tools show as `mcp:server` |
268
+ | `error_class` | the anchor's text, lower-cased, with paths, long hex runs, numbers and quoted strings removed, matched against a fixed table. A JSON-shaped result (Codex command and MCP results, T3 dict outputs) is matched on its error fields (`error`, `stderr`, `message`), else its output fields (`output`, `content`, ...), plus a non-zero `exit_code`, not on the raw JSON. A successful anchor has an empty error class (`command_not_found`, `no_such_file`, `missing_module`, `permission_denied`, `timeout`, `unknown_tool`, `user_declined`, `shell_glob`, `push_rejected`, `conflict`, `syntax_error`, `test_failure`, `nonzero_exit`). The whole text is matched, because Claude Code and Codex print `Exit code 1` first. No match gives `other:` and 8 hex of a digest of the first line |
269
+ | `command_family` | the anchor call's command, as for rubric `requires` (for example `git push`, `python -m pytest`, `pytest`) |
270
+
271
+ The project is left out of the key, so the same failure in three repositories is one group;
272
+ `projects` counts its tasks per repository. `distinct_tasks` counts distinct sessions (the
273
+ same identity studies use), so one long session counts once. A group is `recurring` at
274
+ `--threshold` tasks or more. `audit_incidents` counts the newest scheduled audit's error results
275
+ with the same tool, error class and command family, including results no case sampled.
276
+ Groups sort by recurring, then tasks, then last work day.
277
+
278
+ Each run saves a private `group_reports` record and journals `groups`; the command prints the
279
+ report as JSON. A record holds ids, signal names and counts only, never evidence text. A tool or
280
+ command name that is long or has unusual characters becomes `other`. Records that cannot be
281
+ read are counted as `skipped`.
282
+
283
+ `aut improve run` and `aut improve diagnose` refresh the groups after the worker exits, outside the judge's budget and
284
+ on every outcome. A grouping failure journals `groups_error` with the error type and never
285
+ fails the cycle or the inbox.
235
286
 
236
287
  ## Prepare a scoped change
237
288
 
@@ -382,7 +433,7 @@ Matching worker checkpoints remain available when the outer deadline kills the w
382
433
  Results from an earlier worker or another policy cannot count as the current result.
383
434
  The command returns their identifiers without printing captured trace text.
384
435
  An interrupted or budget-limited cycle resumes its frozen audit before collecting another batch.
385
- The process writes `last_cycle.json` and refreshes `inbox.html`.
436
+ The process writes `last_cycle.json`, refreshes the failure groups and then `inbox.html`.
386
437
  It does not apply proposals, execute trial commands, or upload evidence.
387
438
 
388
439
  A pending frozen audit is reused while its own `--days` and `--limit` match the run.
@@ -395,7 +446,8 @@ aut improve audit --cycle --days 7 --limit 30 --budget-s 1500
395
446
  ```
396
447
 
397
448
  It returns `frozen`, `kept` (a matching audit is already pending), `busy`, `budget`, or
398
- `superseded` (a cycle finished while it built, so it queues nothing).
449
+ `superseded` (a cycle finished while it built, so it queues nothing; this includes a cycle
450
+ that released its audit early and was still judging non-blocking rubrics when the build began).
399
451
  Only `budget` exits 1, and it leaves the pending audit unchanged.
400
452
  `--cycle` takes no `--project`. Plain `aut improve audit` is unchanged and writes no pending audit.
401
453
  `packaging/launchd/dev.aut.improve-audit.plist` is a template that runs this step at 03:50.
@@ -412,12 +464,12 @@ Evidence remains until deliberately removed. Configure backups and retention sep
412
464
 
413
465
  ### Install a stable runtime
414
466
 
415
- Build a wheel, then install it outside the development checkout using cached dependencies:
467
+ Build a wheel, then install the newest one outside the development checkout using cached dependencies:
416
468
 
417
469
  ```bash
418
470
  uv build --wheel
419
471
  python -m agent_usage_tracker.improve.runtime \
420
- --wheel dist/agent_usage_tracker-0.1.0-py3-none-any.whl \
472
+ --wheel "$(ls -t dist/agent_usage_tracker-*-py3-none-any.whl | head -n 1)" \
421
473
  --store "$HOME/.local/share/aut/improve" \
422
474
  --runtime-root "$HOME/.local/share/aut/improve-runtime"
423
475
  ```
@@ -180,7 +180,8 @@ Outcome inputs stay inside their own episode's time bounds. Local estimates skip
180
180
  prompts or a missing attributable final reply, reporting these as unclassified rather than
181
181
  using a reply to an earlier input. Replay previews report `incomplete_outcome_context` separately.
182
182
  Work-category prompts and activity also stay inside their own episode. Final replies stop at
183
- the earlier of the episode end or local midnight, so shared sessions cannot mix separate work days.
183
+ the earlier of the episode end or midnight in `ledger_timezone`, so shared sessions cannot mix
184
+ separate work days.
184
185
  Native Codex prompt order resolves tied timestamps within one session. Conflicting replies
185
186
  from simultaneous sessions do not supply an attributable final reply.
186
187
  The draft `writing_task()` and `bounded_outcome_task()` are not registered. They define explicit
@@ -213,7 +214,8 @@ For scheduled estimates, set `AUT_LABEL=1` and `AUT_LABEL_MODE=local` in `scan.e
213
214
  ### Historical local replay
214
215
 
215
216
  `aut label replay` prepares historical labels without resetting daily collection. First choose
216
- one task and explicit local dates. The end date is exclusive and cannot include an unfinished day:
217
+ one task and explicit dates in `ledger_timezone`. The end date is exclusive and cannot include an
218
+ unfinished day:
217
219
 
218
220
  ```bash
219
221
  aut label replay --task steer_intent --since 2026-10-01 --until 2026-10-08
@@ -226,6 +228,9 @@ excluded because this machine lacks their text. It runs no model and sends no ro
226
228
  same job with `--apply`; progress advances only after each batch is delivered. The default budget
227
229
  is 600 seconds, with a maximum of 3600. An incomplete result reports `remaining` and
228
230
  `complete: false`; repeat the saved job to continue. Nightly watermarks are untouched.
231
+ A job records its `ledger_timezone`. It refuses to continue under another zone; prepare a new job.
232
+ A job saved before the zone was recorded goes on only when the machine's zone has the ledger
233
+ zone's UTC offsets over its range, a day either side.
229
234
  The budget stops new scoring work. An active model call or sink delivery can finish after it.
230
235
  `partial_local_text` counts eligible episode units with some missing steer text. Counts cover
231
236
  supported local source units, not all historical work. Replies and context can also be absent.
@@ -30,9 +30,9 @@ other command refuses it.
30
30
  | `aut label run --local-estimates [--task T]` | classify eligible local inputs with a saved provisional recipe | `label` extra, `pg_dsn`, `label-local.json`, cached model weights |
31
31
  | `aut label import-baseline FILE [--apply]` | preview or append a validated provisional historical baseline; no model calls | `pg_dsn`, private baseline bundle |
32
32
  | `aut cursor-run --model MODEL --workspace DIR < prompt.txt` | register an automated chat before its first scripted Cursor prompt; preserves JSON output | installed `cursor-agent`, signed-in account; may spend account credits |
33
- | `aut label replay --task T --since D --until D` | freeze redacted inputs for an explicit local date range and preview counts; no scoring or sink writes | saved local recipe, logs, `pg_dsn` |
33
+ | `aut label replay --task T --since D --until D` | freeze redacted inputs for an explicit ledger date range (`ledger_timezone`) and preview counts; no scoring or sink writes. A saved job refuses another zone | saved local recipe, logs, `pg_dsn` |
34
34
  | `aut label replay --job ID [--apply] [--budget-s N]` | preview or resume the private job; `--apply` scores local models and delivers batches, with separate progress | same, plus local model runtime |
35
- | `aut history-steers [--dry-run]` | historical daily steer estimates from CLI history files (rows of kind `steer_estimate`, a count and no text) | `pg_dsn` |
35
+ | `aut history-steers [--dry-run]` | historical daily steer estimates from CLI history files (rows of kind `steer_estimate`, a count and no text, at noon in `ledger_timezone`). It stops, and prints the DELETE that removes them, while estimate rows written under another zone remain | `pg_dsn` |
36
36
  | `aut git-days [--dry-run] [--local-only] [--allow-lower]` | commits per ledger day (`ledger_timezone`) from `since` through yesterday. Read-only local and SSH scans, with no fetch. A private salted identity cache combines available sources without duplicate commits. Partial reads preserve earlier counts and omit unknown zeros. Same-count coverage upgrades publish. `--local-only` skips SSH; the global `--git-defer-remotes-until YYYY-MM-DD` flag skips SSH before that date for `scan` or `git-days`. Only fresh complete scans of every source permit `--allow-lower` and reset the cache. Dry runs print year totals without writes. Enabled scans run daily, with incomplete-source retries after 6 h | `[git_days]`, `pg_dsn`, updated completeness column/view in [DEPLOY.md](../DEPLOY.md) |
37
37
  | `aut label backfill --task T --days N` | label one task over the last N days; the nightly watermark is not moved back. `soft_outcome` always covers the last 14 days, whatever `--days` or `--since` say (it notes this on stderr) | same |
38
38
  | `aut label teach --task T --n N [--budget-usd X] [--days D]` | teacher (silver) labels for a stratified sample of up to N units (default spend cap 5 USD, window 90 days) | same, plus `claude` on `PATH` |
@@ -86,7 +86,7 @@ matter:
86
86
  | `cursor_sdk_manifests` | unset | glob of Cursor SDK run `manifest.json` files |
87
87
  | `cursor_db_path` | found per OS | path to Cursor's `state.vscdb` |
88
88
  | `home` | the running user's home | the home directory the collectors read |
89
- | `ledger_timezone` | `Europe/Amsterdam` | the sink's `AUT_TZ`. `git-days` counts days in it; `history-steers` uses it to find days already counted. Both stop when it differs from the zone a `sink_stack` sink records |
89
+ | `ledger_timezone` | `Europe/Amsterdam` | the sink's `AUT_TZ`, and what a day is: `git-days`, `history-steers` and the `label` commands count days in it. `git-days` and `history-steers` stop when it differs from the zone a `sink_stack` sink records. Label gold, silver, eval records, encoders and choices record it; one from another zone is left out, and one with no zone counts only when the machine's zone has the same UTC offsets since 2024 |
90
90
  | `activitywatch_*` | see the example | URL, title retention (`0` keeps titles), titles on or off, deny lists, work title patterns, categories |
91
91
  | `outcome_repo_roots` | `["~/projects"]` | where `aut outcomes` looks for clones (6 levels deep) |
92
92
  | `outcome_fetch` | `false` | fetch and ask `gh` before scoring. Turn it on where the nightly job runs, or squash-merged PRs count as abandoned |
@@ -5,4 +5,4 @@ session, activity and active-time rows, and pushes them to a self-hosted sink. C
5
5
  local-model labels are separate pipelines over the same sink.
6
6
  """
7
7
 
8
- __version__ = "0.1.0"
8
+ __version__ = "0.1.1"
@@ -59,13 +59,13 @@ import argparse
59
59
  import json
60
60
  import os
61
61
  import sys
62
- from datetime import date, timedelta
62
+ from datetime import date, datetime, timedelta
63
63
  from typing import Sequence
64
64
 
65
65
  from . import backup, cursor_run, git_days, inspector, manual_time, outcomes, pricing, review, scan
66
66
  from .collectors import REGISTRY
67
67
  from .config import (ALL_AGENTS, LABEL_DAYS, LABEL_DECISION_TASKS, LABEL_TASK_NAMES, AutError,
68
- Config, load_config, config_path)
68
+ Config, ledger_zone, load_config, config_path)
69
69
  from .push import is_operational_failure, outbox_status, redacted_error, replay_outbox, webhook_url_allowed
70
70
  from .state import collector_wm, load_state
71
71
 
@@ -192,13 +192,14 @@ def _label_run(cfg: Config, args: argparse.Namespace) -> None:
192
192
  """run | backfill: label new units, or one task over the last N days."""
193
193
  from .label import run as label_run
194
194
  from .label.core import local_midnight
195
+ tz = ledger_zone(cfg)
195
196
  tasks: list[str] | None
196
197
  if args.label_cmd == "backfill":
197
- since = date.today() - timedelta(days=args.days)
198
+ since = datetime.now(tz).date() - timedelta(days=args.days)
198
199
  tasks, dry = [args.task], False
199
200
  else:
200
201
  since, tasks, dry = args.since, [args.task] if args.task else None, args.dry_run
201
- report = label_run.run_label(cfg, tasks, local_midnight(since) if since else None,
202
+ report = label_run.run_label(cfg, tasks, local_midnight(since, tz) if since else None,
202
203
  dry_run=dry, local_estimates=getattr(args, "local_estimates", False))
203
204
  label_run.print_report(report, dry_run=dry)
204
205
  failed = [name for name, c in report.items()
@@ -385,7 +386,7 @@ def _build_parser() -> argparse.ArgumentParser:
385
386
  p_lrun = lab_sub.add_parser("run", help="label new units since the watermark")
386
387
  p_lrun.add_argument("--task", choices=LABEL_TASK_NAMES, help="one task (default: label_tasks)")
387
388
  p_lrun.add_argument("--since", type=date.fromisoformat,
388
- help="local YYYY-MM-DD to start from instead of the watermark")
389
+ help="ledger YYYY-MM-DD to start from instead of the watermark")
389
390
  p_lrun.add_argument("--dry-run", action="store_true", help="score and count, write nothing")
390
391
  p_lrun.add_argument("--local-estimates", action="store_true",
391
392
  help="apply saved local decision recipes as separate provisional estimates")
@@ -394,8 +395,8 @@ def _build_parser() -> argparse.ArgumentParser:
394
395
  p_lback.add_argument("--days", type=int, required=True)
395
396
  p_replay = lab_sub.add_parser("replay", help="preview or resume a private historical local job")
396
397
  p_replay.add_argument("--task", choices=("steer_intent", "work_category", "soft_outcome", "writing_activity"))
397
- p_replay.add_argument("--since", type=date.fromisoformat, help="first local date, inclusive")
398
- p_replay.add_argument("--until", type=date.fromisoformat, help="last local date, exclusive")
398
+ p_replay.add_argument("--since", type=date.fromisoformat, help="first ledger date, inclusive")
399
+ p_replay.add_argument("--until", type=date.fromisoformat, help="last ledger date, exclusive")
399
400
  p_replay.add_argument("--job", help="saved private job ID instead of a new range")
400
401
  p_replay.add_argument("--apply", action="store_true", help="score and deliver the saved job")
401
402
  p_replay.add_argument("--budget-s", type=int, default=600, help="1 to 3600 seconds per resume")
@@ -6,14 +6,16 @@ written as "env:VAR" is resolved from the environment at run time.
6
6
  from __future__ import annotations
7
7
 
8
8
  from typing import Any
9
+ import functools
9
10
  import os
10
11
  import re
11
12
  import socket
12
13
  import sys
13
14
  import tomllib
14
15
  from dataclasses import dataclass, field
15
- from datetime import date, datetime
16
+ from datetime import date, datetime, timedelta, timezone, tzinfo
16
17
  from pathlib import Path
18
+ from zoneinfo import ZoneInfo, ZoneInfoNotFoundError
17
19
 
18
20
  from .state import warn_if_group_readable
19
21
 
@@ -44,6 +46,50 @@ class AutError(Exception):
44
46
  """An error for the operator: main prints the message and exits 1, with no traceback."""
45
47
 
46
48
 
49
+ def ledger_zone(cfg: Config) -> ZoneInfo:
50
+ """The zone of a ledger day, as in the sink's views (ledger_timezone)."""
51
+ return _zone(cfg.ledger_timezone)
52
+
53
+
54
+ def _zone(name: str) -> ZoneInfo:
55
+ try:
56
+ return ZoneInfo(name)
57
+ except (ZoneInfoNotFoundError, ValueError):
58
+ raise AutError(f"ledger_timezone {name!r} is unknown here: check "
59
+ "the name, or install the time zone database (the system tzdata package, "
60
+ "or pip install tzdata)") from None
61
+
62
+
63
+ def _machine_offset(t: float) -> timedelta | None:
64
+ return datetime.fromtimestamp(t, timezone.utc).astimezone().utcoffset()
65
+
66
+
67
+ def machine_zone_agrees(tz: tzinfo, start: datetime, end: datetime) -> bool:
68
+ """The machine's zone has tz's UTC offset at every hour from start to end, so a day or an hour
69
+ the machine counted there is the same in tz. Offsets, not midnights: a midnight can be missing
70
+ (a DST jump at 00:00) in both zones, or agree while the hours around it do not."""
71
+ t, stop = start.timestamp() // 3600 * 3600, end.timestamp()
72
+ while t <= stop:
73
+ if _machine_offset(t) != datetime.fromtimestamp(t, tz).utcoffset():
74
+ return False
75
+ t += 3600
76
+ return True
77
+
78
+
79
+ @functools.cache
80
+ def _machine_agrees_since_2024(zone: str) -> bool:
81
+ return machine_zone_agrees(_zone(zone), datetime(2024, 1, 1, tzinfo=timezone.utc),
82
+ datetime.now(timezone.utc))
83
+
84
+
85
+ def record_zone_ok(rec: dict[str, Any], zone: str) -> bool:
86
+ """A saved label record (gold, silver, eval, encoder, choice) holds under ledger zone `zone`:
87
+ it was made in it, or it has no zone (saved before the zone was) and the machine zone, which
88
+ its days were counted in, agrees with `zone` since 2024."""
89
+ saved = rec.get("ledger_timezone")
90
+ return saved == zone if saved is not None else _machine_agrees_since_2024(zone)
91
+
92
+
47
93
  def config_path() -> Path:
48
94
  if os.environ.get("AUT_CONFIG"):
49
95
  return Path(os.environ["AUT_CONFIG"]).expanduser()
@@ -29,9 +29,8 @@ from contextlib import nullcontext
29
29
  from datetime import date, datetime, timedelta, timezone, tzinfo
30
30
  from pathlib import Path, PurePosixPath
31
31
  from typing import Any, Callable, Iterable
32
- from zoneinfo import ZoneInfo, ZoneInfoNotFoundError
33
32
 
34
- from .config import AutError, Config
33
+ from .config import AutError, Config, ledger_zone
35
34
  from .git_safe import HARDENED_ENV, git_argv
36
35
  from .git_day_cache import Snapshot, Status, load_cache, merge_sources
37
36
  from .push import JobBeat, job_heartbeat, redacted_error
@@ -327,16 +326,6 @@ def load_stored(dsn: str) -> dict[date, tuple[int, bool]]:
327
326
  conn.execute("SELECT day, commits, complete FROM agents.git_commits_daily").fetchall()}
328
327
 
329
328
 
330
- def _ledger_zone(cfg: Config) -> tzinfo:
331
- """The ledger's day, as in the sink's views."""
332
- try:
333
- return ZoneInfo(cfg.ledger_timezone)
334
- except (ZoneInfoNotFoundError, ValueError):
335
- raise AutError(f"git-days: ledger_timezone {cfg.ledger_timezone!r} is unknown here: check "
336
- "the name, or install the time zone database (the system tzdata package, "
337
- "or pip install tzdata)") from None
338
-
339
-
340
329
  def run(cfg: Config, *, dry_run: bool, allow_lower: bool = False, now: datetime | None = None,
341
330
  tz: tzinfo | None = None, local_only: bool = False,
342
331
  defer_remotes_until: date | None = None) -> dict[str, Any]:
@@ -348,7 +337,7 @@ def run(cfg: Config, *, dry_run: bool, allow_lower: bool = False, now: datetime
348
337
  dsn = None if dry_run else cfg.require_pg("git-days")
349
338
  if dsn:
350
339
  check_ledger_zone(dsn, cfg.ledger_timezone)
351
- tz = tz or _ledger_zone(cfg)
340
+ tz = tz or ledger_zone(cfg)
352
341
  deferred = local_only or (defer_remotes_until is not None
353
342
  and now.astimezone(tz).date() < defer_remotes_until)
354
343
  cache_path = state_path().with_name('git-days-sources.json')
@@ -481,7 +470,7 @@ def run_if_due(cfg: Config, now: datetime | None = None, *,
481
470
  state = load_state()
482
471
  wm = collector_wm(state, NAME)
483
472
  deferred = (defer_remotes_until is not None
484
- and now.astimezone(_ledger_zone(cfg)).date() < defer_remotes_until)
473
+ and now.astimezone(ledger_zone(cfg)).date() < defer_remotes_until)
485
474
  new_remote_reader = (wm.get("partial_pending") is True and bool(cfg.git_days_remotes)
486
475
  and not deferred
487
476
  and wm.get("remote_scan_version") != _REMOTE_SCAN_VERSION)
@@ -14,12 +14,12 @@ from __future__ import annotations
14
14
 
15
15
  import json
16
16
  from collections import Counter
17
- from datetime import date, datetime
17
+ from datetime import date, datetime, tzinfo
18
18
  from pathlib import Path
19
19
 
20
20
  from .collectors.claude_code import _is_injected
21
21
  from .collectors.codex import is_scheduled_automation
22
- from .config import Config
22
+ from .config import AutError, Config, ledger_zone
23
23
  from .event import ActivityEvent, make_event_id
24
24
  from .push import send_activity_rows
25
25
  from .sink import check_ledger_zone, connect
@@ -39,8 +39,8 @@ def is_steer(text: object) -> bool:
39
39
 
40
40
 
41
41
  def daily_counts(path: Path, text_key: str, time_key: str, per_second: int, first: date,
42
- last: date) -> Counter[date]:
43
- """Steers per local day in the window. A missing file is no rows; a bad line is skipped."""
42
+ last: date, tz: tzinfo) -> Counter[date]:
43
+ """Steers per ledger day (tz) in the window. A missing file is no rows; a bad line is skipped."""
44
44
  days: Counter[date] = Counter()
45
45
  if not path.is_file():
46
46
  return days
@@ -48,7 +48,7 @@ def daily_counts(path: Path, text_key: str, time_key: str, per_second: int, firs
48
48
  for line in f:
49
49
  try:
50
50
  o = json.loads(line)
51
- day = datetime.fromtimestamp(o[time_key] / per_second).date()
51
+ day = datetime.fromtimestamp(o[time_key] / per_second, tz).date()
52
52
  except (ValueError, KeyError, TypeError, OSError):
53
53
  continue
54
54
  if first <= day <= last and is_steer(o.get(text_key)):
@@ -64,16 +64,30 @@ def counted_days(dsn: str, zone: str) -> set[tuple[str, date]]:
64
64
  return {(a, d) for a, d in rows}
65
65
 
66
66
 
67
+ def foreign_estimates(dsn: str, zone: str) -> int:
68
+ """Estimate rows not at noon in `zone`: written under another zone. Their event ids name the
69
+ same days, so a rerun would add rows beside them rather than replace them."""
70
+ with connect(dsn) as conn:
71
+ row = conn.execute("SELECT count(*) FROM agents.activity_event WHERE kind = 'steer_estimate' "
72
+ "AND (ts AT TIME ZONE %s)::time <> '12:00'", (zone,)).fetchone()
73
+ return int(row[0]) if row else 0
74
+
75
+
76
+ def drop_foreign_sql(zone: str) -> str:
77
+ return ("DELETE FROM agents.activity_event WHERE kind = 'steer_estimate' "
78
+ f"AND (ts AT TIME ZONE '{zone}')::time <> '12:00';")
79
+
80
+
67
81
  def estimates(cfg: Config, counted: set[tuple[str, date]]) -> list[ActivityEvent]:
68
- rows = []
82
+ rows, tz = [], ledger_zone(cfg)
69
83
  for agent, (rel, text_key, time_key, per_second, first, last) in SOURCES.items():
70
- counts = daily_counts(cfg.home / rel, text_key, time_key, per_second, first, last)
84
+ counts = daily_counts(cfg.home / rel, text_key, time_key, per_second, first, last, tz)
71
85
  for day, n in sorted(counts.items()):
72
86
  if (agent, day) in counted:
73
87
  continue
74
88
  rows.append(ActivityEvent(
75
89
  event_id=make_event_id(agent, "steer_estimate", day), agent=agent,
76
- ts=datetime(day.year, day.month, day.day, 12).astimezone().isoformat(),
90
+ ts=datetime(day.year, day.month, day.day, 12, tzinfo=tz).isoformat(),
77
91
  kind="steer_estimate", session_id=None, turn_id=None,
78
92
  attrs={"count": n, "source": "history"}, machine=cfg.machine))
79
93
  return rows
@@ -82,6 +96,11 @@ def estimates(cfg: Config, counted: set[tuple[str, date]]) -> list[ActivityEvent
82
96
  def run(cfg: Config, dry_run: bool) -> Counter[str]:
83
97
  dsn = cfg.require_pg("history-steers")
84
98
  check_ledger_zone(dsn, cfg.ledger_timezone)
99
+ if n := foreign_estimates(dsn, cfg.ledger_timezone):
100
+ raise AutError(f"history-steers: {n} steer_estimate rows were written under another zone "
101
+ f"(not noon in ledger_timezone {cfg.ledger_timezone}); a rerun would count "
102
+ f"their days twice. Remove them, then run again: "
103
+ f"{drop_foreign_sql(cfg.ledger_timezone)}")
85
104
  rows = estimates(cfg, counted_days(dsn, cfg.ledger_timezone))
86
105
  if not dry_run:
87
106
  send_activity_rows(dsn, rows)
@@ -68,6 +68,10 @@ def register(sub: Any) -> None:
68
68
  desktop.add_argument("--trace", type=Path, required=True, help="raw trace file or segment directory")
69
69
  desktop.add_argument("--responses", type=Path, help="local response companion JSONL")
70
70
  desktop.add_argument("--mapping", type=Path, required=True, help="explicit task/session/parent JSON")
71
+ grouping = actions.add_parser("groups", help="group failed judgments by deterministic signals")
72
+ grouping.add_argument("--days", type=int, default=14)
73
+ grouping.add_argument("--threshold", type=int, default=3,
74
+ help="distinct tasks that make a group recurring")
71
75
  study_cli.register(actions)
72
76
  actions.add_parser("status", help="show local progress counts without trace text")
73
77
 
@@ -97,6 +101,9 @@ def run(args: argparse.Namespace) -> None:
97
101
  case_limit=args.limit, budget_s=args.budget_s, max_calls=args.max_calls,
98
102
  explain_model=args.explain_model)
99
103
  result = report.get("diagnosis", report)
104
+ elif args.improve_cmd == "groups":
105
+ from .groups import refresh
106
+ result = refresh(store, window_days=args.days, threshold=args.threshold)
100
107
  elif args.improve_cmd == "inspect":
101
108
  from .case_view import write_case_view
102
109
  result = write_case_view(store, args.case)