agent-usage-tracker 0.2.0__tar.gz → 0.4.0__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.
- agent_usage_tracker-0.4.0/PKG-INFO +121 -0
- agent_usage_tracker-0.4.0/README.md +84 -0
- agent_usage_tracker-0.4.0/docs/assets/aut-local-setup.png +0 -0
- agent_usage_tracker-0.4.0/docs/assets/aut-workflow.png +0 -0
- agent_usage_tracker-0.4.0/docs/development.md +69 -0
- agent_usage_tracker-0.4.0/docs/evidence.md +281 -0
- agent_usage_tracker-0.4.0/docs/first-run.md +145 -0
- agent_usage_tracker-0.4.0/docs/improvement-studies.md +359 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/improvement.md +47 -9
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/reference.md +18 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/pyproject.toml +8 -3
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/__init__.py +1 -1
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/backup.py +8 -1
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/cli.py +6 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/codex.py +2 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/cursor_cli.py +7 -5
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/access.py +569 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/access_cli.py +263 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/access_mcp.py +101 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/case_view.py +1 -1
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/cli.py +5 -3
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/cursor_evidence.py +164 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/decide.py +23 -6
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/draft.py +24 -7
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/draft_adapters.py +44 -18
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/draft_gguf.py +80 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/evidence.py +616 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/file_reads.py +93 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/investigation.py +256 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/judge.py +67 -29
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/outputs.py +20 -13
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/proposals.py +41 -11
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/ranked_search.py +113 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/report_card.py +3 -2
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/runtime.py +12 -6
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/session_observability.py +55 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/setup_git.py +79 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/skill_proposals.py +83 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/store.py +86 -22
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/studies.py +89 -7
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/study_candidates.py +125 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/study_cli.py +135 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/study_decision.py +8 -3
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/study_followup.py +6 -3
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/study_independence.py +58 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/study_model_calls.py +97 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/study_observe.py +450 -0
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/study_reviews.py +74 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/t3.py +176 -10
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/trial.py +55 -18
- agent_usage_tracker-0.4.0/src/agent_usage_tracker/improve/turn_revisions.py +222 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/validation.py +40 -1
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve_publish.py +2 -1
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/local_page.py +2 -1
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/net.py +11 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/redact.py +21 -1
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_activitywatch.py +2 -1
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_cursor_api.py +5 -3
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_source_registry.py +3 -1
- agent_usage_tracker-0.4.0/tests/improve/test_count_evidence.py +171 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_decide.py +27 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_draft_adapters.py +84 -0
- agent_usage_tracker-0.4.0/tests/improve/test_evidence_access.py +377 -0
- agent_usage_tracker-0.4.0/tests/improve/test_evidence_mcp.py +153 -0
- agent_usage_tracker-0.4.0/tests/improve/test_file_reads.py +48 -0
- agent_usage_tracker-0.4.0/tests/improve/test_investigation.py +147 -0
- agent_usage_tracker-0.4.0/tests/improve/test_native_evidence.py +187 -0
- agent_usage_tracker-0.4.0/tests/improve/test_ranked_evidence.py +124 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_report_card.py +21 -4
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_runtime.py +26 -0
- agent_usage_tracker-0.4.0/tests/improve/test_session_observability.py +43 -0
- agent_usage_tracker-0.4.0/tests/improve/test_skill_proposals.py +70 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_store.py +9 -3
- agent_usage_tracker-0.4.0/tests/improve/test_studies.py +946 -0
- agent_usage_tracker-0.4.0/tests/improve/test_study_candidates.py +132 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_t3.py +223 -3
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_trial.py +46 -0
- agent_usage_tracker-0.4.0/tests/improve/test_turn_provenance.py +105 -0
- agent_usage_tracker-0.4.0/tests/improve/test_turn_revisions.py +222 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_validation.py +41 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label.py +10 -1
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_models.py +2 -1
- agent_usage_tracker-0.4.0/tests/outcomes/conftest.py +25 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/outcomes/test_outcome_git.py +24 -21
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/outcomes/test_outcomes.py +39 -39
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_backup.py +1 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_git_days.py +14 -5
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_local_page.py +15 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_publish_improve.py +18 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_redact.py +33 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/test_observation_server.py +14 -3
- agent_usage_tracker-0.2.0/PKG-INFO +0 -410
- agent_usage_tracker-0.2.0/README.md +0 -378
- agent_usage_tracker-0.2.0/docs/first-run.md +0 -64
- agent_usage_tracker-0.2.0/docs/improvement-studies.md +0 -169
- agent_usage_tracker-0.2.0/src/agent_usage_tracker/improve/evidence.py +0 -322
- agent_usage_tracker-0.2.0/src/agent_usage_tracker/improve/study_cli.py +0 -46
- agent_usage_tracker-0.2.0/tests/improve/test_studies.py +0 -405
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/.gitignore +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/CONTRACT.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/DEPLOY.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/LICENSE +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/config.example.toml +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/collectors-to-add.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/harness-coverage.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/human-html/2026-11-08-understanding-implicit-state-machines.html +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/human-html/2026-11-09-status-state-reliability-fixes.html +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/human-html/docs-index.json +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/human-html/index.html +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/local-labels.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/new-laptop.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/observations.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/quota.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/research/codexbar-learnings.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/research/otel-lane.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/research/sdk-usage-capture.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/superpowers/plans/2026-07-22-aut-showcase-metrics.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/docs/superpowers/specs/2026-07-22-aut-showcase-metrics-design.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/README.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/claude/commands/outcome.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/launchd/aut-scan.sh +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/launchd/dev.aut.backup.plist +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/launchd/dev.aut.improve-audit.plist +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/launchd/dev.aut.improve-draft.plist +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/launchd/dev.aut.retry.plist +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/launchd/dev.aut.scan.plist +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/observations/Dockerfile +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/observations/Dockerfile.dockerignore +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/observations/README.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/observations/aut-observations.service +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/sql/turn-snapshot-grant.sql +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/systemd/agent-usage-tracker.service +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/systemd/agent-usage-tracker.timer +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/packaging/windows/register-task.ps1 +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/__init__.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/_util.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/activity_common.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/activitywatch.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/claude_code.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/claude_code_activity.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/commits.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/cursor.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/cursor_api.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/cursor_sdk.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/generic_sdk.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/pi.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/session_digest.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/collectors/t3.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/config.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/coverage.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/coverage_catalog.json +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/cursor_run.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/demo.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/demo_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/diagnostics.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/episode_store.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/event.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/git_day_cache.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/git_days.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/git_safe.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/harness_inventory.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/history_steers.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/__init__.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/desktop.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/groups.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/inbox.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/qualify.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/rubric.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/rubrics/avoidable_recovery.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/rubrics/code_quality.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/rubrics/completion_supported.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/rubrics/redundant_tests.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/rubrics/verbosity.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/schedule.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/scores.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/improve/search_check.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/inspector.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/__init__.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/backends.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/bakeoff.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/baseline.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/clef.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/context.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/core.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/covered_run.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/evalset.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/gold.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/local.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/local_parallel.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/merge.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/outcome_coverage.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/replay.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/run.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/tasks.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/label/teach.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/manual_time.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/observation_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/observation_report.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/observation_server.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/observations.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/operational_status.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/outcome_git.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/outcomes.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/overview.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/overview_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/pricing.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/pricing_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/push.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/quota.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/quota_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/repo_keys.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/review.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/scan.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/scheduled.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sdk.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/setup_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/setup_schedule.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/README.md +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/apply.sh +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/compose.yaml +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/env.example +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/grafana/dashboards/agents-effort.json +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/grafana/dashboards/agents-usage.json +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/grafana/provisioning/dashboards/aut.yaml +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/grafana/provisioning/datasources/agents.yaml +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/sql/01_schema.sql +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/sql/02_price.sql +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/sql/03_effort.sql +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/sql/04_grants.sql +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sink_stack/sql/refresh.sql +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/sources.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/state.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/summary.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/summary_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/telemetry.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/src/agent_usage_tracker/usage_price.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/codex_pi_helpers.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_activity.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_activitywatch_macos.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_claude_code.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_claude_profiles.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_codex.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_codex_pi.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_collectors.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_commits.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_cursor.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_cursor_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_cursor_cli_cache.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_cursor_sdk.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_generic_sdk.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_links.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_links_claude.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_pi.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_read_horizon.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_session_digest.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/collectors/test_t3.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/conftest.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/fixtures/claude-statusline-quota.json +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/__init__.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/conftest.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_case_context.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_case_view.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_desktop.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_draft.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_evidence.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_evidence_citations.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_groups.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_inbox.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_judge.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_judge_choices.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_judge_deadline.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_judge_native_deadline.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_judge_queue.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_judge_sampling.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_judge_stall.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_outputs.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_proposals.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_qualify.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_rubric.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_schedule.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_scores.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_search_check.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/improve/test_studies_decision.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/label_helpers.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_context_paths.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_covered_run.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_cursor_origin.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_history_steers.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_bakeoff.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_baseline.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_context.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_core.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_evalset.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_gap.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_gold.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_local.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_parallel.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_reliability.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_replay.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_tasks.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_teach.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_label_writing.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/label/test_outcome_coverage.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/outcomes/test_outcome_reliability.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/fake_psycopg.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/price_fixture.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_coverage.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_coverage_diagnostics.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_coverage_scan.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_cursor_run.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_delivery_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_delivery_notices.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_demo.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_first_run.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_git_day_cache.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_harness_inventory.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_heartbeat.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_inspect.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_ledger_zone.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_net.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_overview.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_overview_quota.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_pricing.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_push.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_push_outbox.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_push_snapshots.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_quota.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_repo_keys.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_review.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_scan_wrapper.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_sessions.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_sink.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_sink_sql.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_state.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_summary.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/pipeline/test_summary_native_counts.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/test_observation_cli.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/test_observation_report.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/test_observations.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/test_observe_entrypoint.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/test_public_projection.py +0 -0
- {agent_usage_tracker-0.2.0 → agent_usage_tracker-0.4.0}/tests/test_telemetry.py +0 -0
|
@@ -0,0 +1,121 @@
|
|
|
1
|
+
Metadata-Version: 2.5
|
|
2
|
+
Name: agent-usage-tracker
|
|
3
|
+
Version: 0.4.0
|
|
4
|
+
Summary: Track coding-agent usage and outcomes from local logs, find recurring problems, and test proposed fixes before adopting them.
|
|
5
|
+
Project-URL: Homepage, https://github.com/rhnfzl/agent-usage-tracker
|
|
6
|
+
Project-URL: Issues, https://github.com/rhnfzl/agent-usage-tracker/issues
|
|
7
|
+
License: MIT
|
|
8
|
+
License-File: LICENSE
|
|
9
|
+
Keywords: claude-code,codex,coding-agent,cursor,observability,pi,tokens,usage
|
|
10
|
+
Classifier: Development Status :: 4 - Beta
|
|
11
|
+
Classifier: Environment :: Console
|
|
12
|
+
Classifier: Operating System :: OS Independent
|
|
13
|
+
Classifier: Programming Language :: Python :: 3
|
|
14
|
+
Classifier: Topic :: Software Development
|
|
15
|
+
Requires-Python: >=3.11
|
|
16
|
+
Requires-Dist: tzdata; sys_platform == 'win32'
|
|
17
|
+
Provides-Extra: code
|
|
18
|
+
Requires-Dist: pydantic-monty<1.7,>=1.6; extra == 'code'
|
|
19
|
+
Provides-Extra: label
|
|
20
|
+
Requires-Dist: huggingface-hub>=1.3; (sys_platform == 'darwin' and platform_machine == 'arm64') and extra == 'label'
|
|
21
|
+
Requires-Dist: mlx-lm<0.33,>=0.32.2; (sys_platform == 'darwin' and platform_machine == 'arm64') and extra == 'label'
|
|
22
|
+
Requires-Dist: mlx>=0.31.2; (sys_platform == 'darwin' and platform_machine == 'arm64') and extra == 'label'
|
|
23
|
+
Requires-Dist: psycopg[binary]>=3.1; extra == 'label'
|
|
24
|
+
Provides-Extra: mcp
|
|
25
|
+
Requires-Dist: mcp<2,>=1.30; extra == 'mcp'
|
|
26
|
+
Provides-Extra: merge
|
|
27
|
+
Requires-Dist: huggingface-hub>=1.3; extra == 'merge'
|
|
28
|
+
Requires-Dist: peft>=0.21; extra == 'merge'
|
|
29
|
+
Requires-Dist: torch>=2.14; extra == 'merge'
|
|
30
|
+
Requires-Dist: transformers>=5.17; extra == 'merge'
|
|
31
|
+
Provides-Extra: postgres
|
|
32
|
+
Requires-Dist: psycopg[binary]>=3.1; extra == 'postgres'
|
|
33
|
+
Provides-Extra: test
|
|
34
|
+
Requires-Dist: pytest; extra == 'test'
|
|
35
|
+
Requires-Dist: pytest-xdist==3.8.0; extra == 'test'
|
|
36
|
+
Description-Content-Type: text/markdown
|
|
37
|
+
|
|
38
|
+
# Agent Usage Tracker (`aut`)
|
|
39
|
+
|
|
40
|
+
[](https://github.com/rhnfzl/agent-usage-tracker/actions/workflows/ci.yml)
|
|
41
|
+
[](https://pypi.org/project/agent-usage-tracker/)
|
|
42
|
+
[](https://github.com/rhnfzl/agent-usage-tracker/blob/main/LICENSE)
|
|
43
|
+
|
|
44
|
+
**See where your coding agents spend tokens, how the work ends, and what keeps going wrong.**
|
|
45
|
+
|
|
46
|
+
AUT reads the logs Claude Code, Codex, Cursor, and Pi already save. Start with a local usage report or CSV. Add your own database for session history across machines. The optional improvement workflow links recurring problems to the evidence and drafts changes you can review.
|
|
47
|
+
|
|
48
|
+
## Quick start
|
|
49
|
+
|
|
50
|
+
With [uv](https://docs.astral.sh/uv/getting-started/installation/) installed, try the demo with synthetic logs:
|
|
51
|
+
|
|
52
|
+
```sh
|
|
53
|
+
uvx agent-usage-tracker demo
|
|
54
|
+
```
|
|
55
|
+
|
|
56
|
+
The demo finds two agent records totaling **1,140 tokens**. It shows input, output and cache separately, and reports missing model prices as unknown. Try the same report with your own logs, or export them as a CSV:
|
|
57
|
+
|
|
58
|
+
```sh
|
|
59
|
+
uvx agent-usage-tracker summary --period week --group-by agent
|
|
60
|
+
uvx agent-usage-tracker --agents claude_code,codex,cursor_cli,pi export --out usage.csv
|
|
61
|
+
```
|
|
62
|
+
|
|
63
|
+
Local reporting runs without a database, a provider account, or a model. The CLI runs on macOS, Linux, and Windows with Python 3.11 to 3.15. See [first run](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/first-run.md) for installation, source checks, and the private browser overview.
|
|
64
|
+
|
|
65
|
+
## What you can track
|
|
66
|
+
|
|
67
|
+
| Question | Where to start |
|
|
68
|
+
|---|---|
|
|
69
|
+
| Which agents and models am I using? How much is cached? | Local summary and CSV export. |
|
|
70
|
+
| Which sessions belong to the same work? What happened to their commits? | Your own TimescaleDB and Grafana stack, with Git outcome tracking. |
|
|
71
|
+
| How much of my work is building, debugging, planning, or review? | Optional local models estimate work categories and leave uncertain cases unknown. |
|
|
72
|
+
| What if I lose this laptop? | Optional encrypted backups of agent logs and AUT settings with restic. |
|
|
73
|
+
|
|
74
|
+
## From a repeated problem to a fix you can review
|
|
75
|
+
|
|
76
|
+

|
|
77
|
+
|
|
78
|
+
For example, an agent says the tests passed. AUT can surface the tool result behind that claim. If the result is missing, the judgment can stay unknown. When judgments identify the same problem across distinct tasks, AUT can group those cases and draft a small instruction or config change in an isolated copy. Run a check you trust against the original and the candidate, then decide whether to adopt it.
|
|
79
|
+
|
|
80
|
+
Start by freezing a small sample of recent sessions. This step runs locally without a model:
|
|
81
|
+
|
|
82
|
+
```sh
|
|
83
|
+
uvx agent-usage-tracker improve audit --days 7 --limit 30
|
|
84
|
+
```
|
|
85
|
+
|
|
86
|
+
The [improvement guide](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement.md) covers private evidence, advisory judgments, drafts, and the review inbox. The [study guide](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement-studies.md) explains how to test a candidate. Source edits require an explicit decision. A passing check supports the tested change; measuring an improvement in real work requires comparable follow-up.
|
|
87
|
+
|
|
88
|
+
You can also ask a coding agent to explore that evidence directly. Run `aut evidence capture` for a private full-text audit, then search and compare runs through the CLI or MCP. Choose what the client can read and link useful findings to a repair you can test. See the [agent evidence guide](docs/evidence.md).
|
|
89
|
+
|
|
90
|
+
## How I run AUT
|
|
91
|
+
|
|
92
|
+
I run AUT on an Apple silicon Mac to track agent usage, classify work, and find recurring problems. AUT sends usage metadata to a self-hosted database. The local labeler combines that metadata with request and reply text on the Mac.
|
|
93
|
+
|
|
94
|
+

|
|
95
|
+
|
|
96
|
+
| Job | Models I use | Method |
|
|
97
|
+
| --- | --- | --- |
|
|
98
|
+
| What my message is doing | Elephant 1, PPLX Decider Flash, Kev 4B | Weighted scores |
|
|
99
|
+
| What kind of work I’m doing | PPLX Decider Flash, Elephant 1, Clef Flash | Majority vote |
|
|
100
|
+
| Where work with no code changes stands | Elephant 1, Kev 4B, Clef Flash | Weighted scores |
|
|
101
|
+
| Draft a fix for a recurring problem | Inkling 2 Mini via llama.cpp | Low reasoning effort, then human review |
|
|
102
|
+
|
|
103
|
+
The label recipes leave uncertain cases unknown. I run the larger Inkling model only when a recurring problem is eligible for a draft. Each proposal goes to a private inbox for review.
|
|
104
|
+
|
|
105
|
+
You can adapt this setup by starting with one [local decision model](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/local-labels.md#ongoing-local-estimates) per job and checking it against examples you’ve reviewed. Add [offline drafting](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement.md#drafted-proposals) when you have a recurring problem worth testing a fix for.
|
|
106
|
+
|
|
107
|
+
## Collectors
|
|
108
|
+
|
|
109
|
+
Claude Code, Codex, and Pi supply native token records. Cursor CLI supplies prompts and activity. For Cursor token totals, choose one of the IDE, API, or SDK sources to avoid double-counting. T3 Code activity and ActivityWatch are optional. See the [collector matrix](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/reference.md#collectors) and [coverage guide](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/harness-coverage.md).
|
|
110
|
+
|
|
111
|
+
## Privacy model
|
|
112
|
+
|
|
113
|
+
Normal collection works from existing logs. If you configure a sink, it receives usage and activity metadata such as counts, project basenames, and commit IDs. Prompt text and tool output stay off that path. Optional remote drafting sends selected file text and review summaries after redaction. Remote drafting requires a separate opt-in for conversation excerpts. Evidence access uses the content mode you choose for that connection or export. See the [privacy and network-call reference](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/reference.md#privacy-model) and [drafting guide](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement.md#drafted-proposals).
|
|
114
|
+
|
|
115
|
+
Usage is measured in tokens. Dollar estimates apply API list rates and are separate from subscription billing. AUT is a CLI with an optional self-hosted dashboard.
|
|
116
|
+
|
|
117
|
+
## Documentation
|
|
118
|
+
|
|
119
|
+
- **Set up:** [First run](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/first-run.md) · [Self-hosted stack](https://github.com/rhnfzl/agent-usage-tracker/blob/main/src/agent_usage_tracker/sink_stack/README.md) · [Deploy runbook](https://github.com/rhnfzl/agent-usage-tracker/blob/main/DEPLOY.md) · [Timers](https://github.com/rhnfzl/agent-usage-tracker/blob/main/packaging/README.md)
|
|
120
|
+
- **Use it:** [Command and configuration reference](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/reference.md) · [Local labels](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/local-labels.md) · [Quota and statusline](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/quota.md) · [Backup and recovery](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/new-laptop.md)
|
|
121
|
+
- **Improve it:** [Private improvement](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement.md) · [Controlled studies](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement-studies.md) · [Observation imports](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/observations.md) · [Development](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/development.md) · [Data contract](https://github.com/rhnfzl/agent-usage-tracker/blob/main/CONTRACT.md)
|
|
@@ -0,0 +1,84 @@
|
|
|
1
|
+
# Agent Usage Tracker (`aut`)
|
|
2
|
+
|
|
3
|
+
[](https://github.com/rhnfzl/agent-usage-tracker/actions/workflows/ci.yml)
|
|
4
|
+
[](https://pypi.org/project/agent-usage-tracker/)
|
|
5
|
+
[](https://github.com/rhnfzl/agent-usage-tracker/blob/main/LICENSE)
|
|
6
|
+
|
|
7
|
+
**See where your coding agents spend tokens, how the work ends, and what keeps going wrong.**
|
|
8
|
+
|
|
9
|
+
AUT reads the logs Claude Code, Codex, Cursor, and Pi already save. Start with a local usage report or CSV. Add your own database for session history across machines. The optional improvement workflow links recurring problems to the evidence and drafts changes you can review.
|
|
10
|
+
|
|
11
|
+
## Quick start
|
|
12
|
+
|
|
13
|
+
With [uv](https://docs.astral.sh/uv/getting-started/installation/) installed, try the demo with synthetic logs:
|
|
14
|
+
|
|
15
|
+
```sh
|
|
16
|
+
uvx agent-usage-tracker demo
|
|
17
|
+
```
|
|
18
|
+
|
|
19
|
+
The demo finds two agent records totaling **1,140 tokens**. It shows input, output and cache separately, and reports missing model prices as unknown. Try the same report with your own logs, or export them as a CSV:
|
|
20
|
+
|
|
21
|
+
```sh
|
|
22
|
+
uvx agent-usage-tracker summary --period week --group-by agent
|
|
23
|
+
uvx agent-usage-tracker --agents claude_code,codex,cursor_cli,pi export --out usage.csv
|
|
24
|
+
```
|
|
25
|
+
|
|
26
|
+
Local reporting runs without a database, a provider account, or a model. The CLI runs on macOS, Linux, and Windows with Python 3.11 to 3.15. See [first run](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/first-run.md) for installation, source checks, and the private browser overview.
|
|
27
|
+
|
|
28
|
+
## What you can track
|
|
29
|
+
|
|
30
|
+
| Question | Where to start |
|
|
31
|
+
|---|---|
|
|
32
|
+
| Which agents and models am I using? How much is cached? | Local summary and CSV export. |
|
|
33
|
+
| Which sessions belong to the same work? What happened to their commits? | Your own TimescaleDB and Grafana stack, with Git outcome tracking. |
|
|
34
|
+
| How much of my work is building, debugging, planning, or review? | Optional local models estimate work categories and leave uncertain cases unknown. |
|
|
35
|
+
| What if I lose this laptop? | Optional encrypted backups of agent logs and AUT settings with restic. |
|
|
36
|
+
|
|
37
|
+
## From a repeated problem to a fix you can review
|
|
38
|
+
|
|
39
|
+

|
|
40
|
+
|
|
41
|
+
For example, an agent says the tests passed. AUT can surface the tool result behind that claim. If the result is missing, the judgment can stay unknown. When judgments identify the same problem across distinct tasks, AUT can group those cases and draft a small instruction or config change in an isolated copy. Run a check you trust against the original and the candidate, then decide whether to adopt it.
|
|
42
|
+
|
|
43
|
+
Start by freezing a small sample of recent sessions. This step runs locally without a model:
|
|
44
|
+
|
|
45
|
+
```sh
|
|
46
|
+
uvx agent-usage-tracker improve audit --days 7 --limit 30
|
|
47
|
+
```
|
|
48
|
+
|
|
49
|
+
The [improvement guide](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement.md) covers private evidence, advisory judgments, drafts, and the review inbox. The [study guide](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement-studies.md) explains how to test a candidate. Source edits require an explicit decision. A passing check supports the tested change; measuring an improvement in real work requires comparable follow-up.
|
|
50
|
+
|
|
51
|
+
You can also ask a coding agent to explore that evidence directly. Run `aut evidence capture` for a private full-text audit, then search and compare runs through the CLI or MCP. Choose what the client can read and link useful findings to a repair you can test. See the [agent evidence guide](docs/evidence.md).
|
|
52
|
+
|
|
53
|
+
## How I run AUT
|
|
54
|
+
|
|
55
|
+
I run AUT on an Apple silicon Mac to track agent usage, classify work, and find recurring problems. AUT sends usage metadata to a self-hosted database. The local labeler combines that metadata with request and reply text on the Mac.
|
|
56
|
+
|
|
57
|
+

|
|
58
|
+
|
|
59
|
+
| Job | Models I use | Method |
|
|
60
|
+
| --- | --- | --- |
|
|
61
|
+
| What my message is doing | Elephant 1, PPLX Decider Flash, Kev 4B | Weighted scores |
|
|
62
|
+
| What kind of work I’m doing | PPLX Decider Flash, Elephant 1, Clef Flash | Majority vote |
|
|
63
|
+
| Where work with no code changes stands | Elephant 1, Kev 4B, Clef Flash | Weighted scores |
|
|
64
|
+
| Draft a fix for a recurring problem | Inkling 2 Mini via llama.cpp | Low reasoning effort, then human review |
|
|
65
|
+
|
|
66
|
+
The label recipes leave uncertain cases unknown. I run the larger Inkling model only when a recurring problem is eligible for a draft. Each proposal goes to a private inbox for review.
|
|
67
|
+
|
|
68
|
+
You can adapt this setup by starting with one [local decision model](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/local-labels.md#ongoing-local-estimates) per job and checking it against examples you’ve reviewed. Add [offline drafting](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement.md#drafted-proposals) when you have a recurring problem worth testing a fix for.
|
|
69
|
+
|
|
70
|
+
## Collectors
|
|
71
|
+
|
|
72
|
+
Claude Code, Codex, and Pi supply native token records. Cursor CLI supplies prompts and activity. For Cursor token totals, choose one of the IDE, API, or SDK sources to avoid double-counting. T3 Code activity and ActivityWatch are optional. See the [collector matrix](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/reference.md#collectors) and [coverage guide](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/harness-coverage.md).
|
|
73
|
+
|
|
74
|
+
## Privacy model
|
|
75
|
+
|
|
76
|
+
Normal collection works from existing logs. If you configure a sink, it receives usage and activity metadata such as counts, project basenames, and commit IDs. Prompt text and tool output stay off that path. Optional remote drafting sends selected file text and review summaries after redaction. Remote drafting requires a separate opt-in for conversation excerpts. Evidence access uses the content mode you choose for that connection or export. See the [privacy and network-call reference](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/reference.md#privacy-model) and [drafting guide](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement.md#drafted-proposals).
|
|
77
|
+
|
|
78
|
+
Usage is measured in tokens. Dollar estimates apply API list rates and are separate from subscription billing. AUT is a CLI with an optional self-hosted dashboard.
|
|
79
|
+
|
|
80
|
+
## Documentation
|
|
81
|
+
|
|
82
|
+
- **Set up:** [First run](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/first-run.md) · [Self-hosted stack](https://github.com/rhnfzl/agent-usage-tracker/blob/main/src/agent_usage_tracker/sink_stack/README.md) · [Deploy runbook](https://github.com/rhnfzl/agent-usage-tracker/blob/main/DEPLOY.md) · [Timers](https://github.com/rhnfzl/agent-usage-tracker/blob/main/packaging/README.md)
|
|
83
|
+
- **Use it:** [Command and configuration reference](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/reference.md) · [Local labels](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/local-labels.md) · [Quota and statusline](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/quota.md) · [Backup and recovery](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/new-laptop.md)
|
|
84
|
+
- **Improve it:** [Private improvement](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement.md) · [Controlled studies](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/improvement-studies.md) · [Observation imports](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/observations.md) · [Development](https://github.com/rhnfzl/agent-usage-tracker/blob/main/docs/development.md) · [Data contract](https://github.com/rhnfzl/agent-usage-tracker/blob/main/CONTRACT.md)
|
|
Binary file
|
|
Binary file
|
|
@@ -0,0 +1,69 @@
|
|
|
1
|
+
# Development
|
|
2
|
+
|
|
3
|
+
From a checkout:
|
|
4
|
+
|
|
5
|
+
In restricted agent sessions, keep uv's cache inside the checkout before running these commands:
|
|
6
|
+
|
|
7
|
+
```bash
|
|
8
|
+
export UV_CACHE_DIR="$PWD/.agent-harness/uv-cache"
|
|
9
|
+
```
|
|
10
|
+
|
|
11
|
+
```bash
|
|
12
|
+
uv venv
|
|
13
|
+
uv pip install -e ".[test,postgres]" ruff==0.15.11 mypy==2.4.1
|
|
14
|
+
uv run --no-sync pytest -q
|
|
15
|
+
uv run --no-sync ruff check .
|
|
16
|
+
uv run --no-sync mypy
|
|
17
|
+
```
|
|
18
|
+
|
|
19
|
+
For feedback before committing, run the same Ruff and mypy commands as CI:
|
|
20
|
+
|
|
21
|
+
```bash
|
|
22
|
+
uv run --no-sync ruff check . && uv run --no-sync mypy
|
|
23
|
+
```
|
|
24
|
+
|
|
25
|
+
You can add that line to your existing pre-commit hook. It checks the working tree.
|
|
26
|
+
For a full test run with the same three workers as CI:
|
|
27
|
+
|
|
28
|
+
```bash
|
|
29
|
+
uv run --no-sync pytest -q -n 3 --dist worksteal --max-worker-restart=0 --durations=15
|
|
30
|
+
```
|
|
31
|
+
|
|
32
|
+
The [CI workflow](../.github/workflows/ci.yml) tests Linux, macOS, and Windows across supported Python versions.
|
|
33
|
+
It also checks Ruff, mypy, the sink SQL, and the Compose configuration.
|
|
34
|
+
The `CI required` check passes once all validation jobs succeed.
|
|
35
|
+
In GitHub's branch protection for `main`, require a pull request and the `CI required` check from GitHub Actions.
|
|
36
|
+
Require branches to be up to date before merging, and apply the rule to administrators.
|
|
37
|
+
CI also runs on `main` to verify the landed commit.
|
|
38
|
+
SQL integration tests require a scratch database on a TimescaleDB server and `psql`, through `AUT_TEST_PG_DSN` with a literal IP address and port; otherwise those tests skip.
|
|
39
|
+
|
|
40
|
+
## Local preflight and packages
|
|
41
|
+
|
|
42
|
+
Run these commands on macOS or Linux with `uv`, Docker Compose, `psql`, and Perl with `shasum` available.
|
|
43
|
+
|
|
44
|
+
```bash
|
|
45
|
+
./scripts/local-ci.sh quick
|
|
46
|
+
./scripts/local-ci.sh check
|
|
47
|
+
./scripts/local-ci.sh package HEAD
|
|
48
|
+
```
|
|
49
|
+
|
|
50
|
+
`quick` runs Ruff. `check` runs Ruff, mypy, the two suite time zones, the MCP suite, and the SQL and Compose checks.
|
|
51
|
+
These checks run concurrently. The preflight reserves cores for other checks and uses up to six pytest workers per time zone.
|
|
52
|
+
Set `AUT_CI_WORKERS` to lower the worker count per time zone if the machine is busy.
|
|
53
|
+
It uses a disposable TimescaleDB container on a random loopback port.
|
|
54
|
+
The preflight defaults to Python 3.11. Set `AUT_CI_PYTHON` to test another supported version.
|
|
55
|
+
For the Mac batch publisher, pass `./scripts/local-ci.sh quick` as `--light` and `./scripts/local-ci.sh check` as `--full`.
|
|
56
|
+
The full local preflight fails above 60 seconds, including dependency setup and cleanup.
|
|
57
|
+
It lets every check finish and reports the results before rejecting an over-budget run.
|
|
58
|
+
GitHub still runs the required platform matrix for the pull request. That run has its own runner queue and timing.
|
|
59
|
+
|
|
60
|
+
`package` builds a wheel from the source distribution of the named commit.
|
|
61
|
+
It checks both distributions, installs the wheel in isolation, and saves the artifacts under `dist/local-<short-commit>/`.
|
|
62
|
+
The directory contains `SOURCE_COMMIT` and `SHA256SUMS` for deployment verification.
|
|
63
|
+
Build from the merged commit when preparing a deployment candidate.
|
|
64
|
+
Check the bundle with `shasum -a 256 -c SHA256SUMS` from its output directory.
|
|
65
|
+
Use the verified wheel with the [standalone runtime installer](improvement.md#install-a-stable-runtime) to stage an upgrade before switching a timer.
|
|
66
|
+
|
|
67
|
+
To release, raise `__version__` in `src/agent_usage_tracker/__init__.py`, merge, and push the tag `v<version>`.
|
|
68
|
+
The [release workflow](../.github/workflows/release.yml) builds it and publishes it to PyPI through trusted publishing.
|
|
69
|
+
Use synthetic fixtures in public contributions. Keep session logs, personal labels, and private benchmarks out of commits.
|
|
@@ -0,0 +1,281 @@
|
|
|
1
|
+
# Investigate coding-agent sessions with AUT
|
|
2
|
+
|
|
3
|
+
Search across agent sessions, cite captured events, and compare a failed attempt with a clean one.
|
|
4
|
+
Use the CLI from a shell or give an agent the same operations through MCP. Both read a selected
|
|
5
|
+
audit or a portable evidence bundle.
|
|
6
|
+
|
|
7
|
+
## Capture a project
|
|
8
|
+
|
|
9
|
+
```sh
|
|
10
|
+
aut evidence capture --project /path/to/repo --days 7 --limit 20
|
|
11
|
+
aut evidence --audit AUDIT_ID describe
|
|
12
|
+
aut evidence --audit AUDIT_ID search --status error
|
|
13
|
+
aut evidence --audit AUDIT_ID search --query "Operation not permitted"
|
|
14
|
+
aut evidence --audit AUDIT_ID search --mode ranked --query "permissions cache"
|
|
15
|
+
aut evidence --audit AUDIT_ID session SESSION_ID
|
|
16
|
+
aut evidence --audit AUDIT_ID event SESSION_ID EVENT_ID
|
|
17
|
+
aut evidence --audit AUDIT_ID commits COMMIT_SHA
|
|
18
|
+
aut evidence --audit AUDIT_ID compare FAILED_SESSION_ID CLEAN_SESSION_ID
|
|
19
|
+
```
|
|
20
|
+
|
|
21
|
+
Exact mode is the default. It matches literal substrings. Ranked mode uses SQLite FTS5 to match words
|
|
22
|
+
in any order and puts exact phrases first. A running MCP server reuses an in-memory index for
|
|
23
|
+
queries and pages that use the same filters. It rebuilds the index when the filters or source
|
|
24
|
+
files change. Each CLI invocation builds a fresh index. Narrow the filters if the index would
|
|
25
|
+
exceed 50,000 events or 64 MiB. Metadata-only scope supports filters without text search.
|
|
26
|
+
|
|
27
|
+
Copy the full audit, session, and event IDs from the JSON results. Follow search's `next_cursor`
|
|
28
|
+
with the same filters. An empty search page can still have a cursor when the scan budget runs out.
|
|
29
|
+
Session results list event references. Supported native Codex turns link the completed turn to its
|
|
30
|
+
recorded model and workspace. Metadata-only clients can see the status and source location.
|
|
31
|
+
Read a referenced event and follow `next_offset` to read its
|
|
32
|
+
remaining text. Commit lookup cites a matched Git call and result. A short recorded prefix can
|
|
33
|
+
match more than one commit. AUT reports the prefix as captured. Redacted and metadata scopes omit
|
|
34
|
+
command-derived IDs that match a redaction rule. Comparison shows event counts and the first difference in the observed tool
|
|
35
|
+
sequence. A missing result stays unknown. The comparison does not infer why two runs differed.
|
|
36
|
+
|
|
37
|
+
For a completed T3 run with a ready workspace checkpoint, AUT resolves its local Git ref
|
|
38
|
+
when it captures the audit. Open the boundary event to see the checkpoint time and, with
|
|
39
|
+
full or redacted access, the ref and commit. This is the workspace state saved by T3 after a run.
|
|
40
|
+
Ordinary-work studies still verify turn-start setup from the evidence recorded for that turn.
|
|
41
|
+
If you already have a checkpoint SHA, use
|
|
42
|
+
`aut evidence --audit AUDIT_ID commits CHECKPOINT_SHA --source checkpoint` to find its run.
|
|
43
|
+
MCP accepts the same `source` option.
|
|
44
|
+
|
|
45
|
+
The project filter includes its Git worktrees and excludes sessions with recorded directories in
|
|
46
|
+
other repositories. T3 applies the thread limit within that project.
|
|
47
|
+
The filter cannot infer where unrecorded tool activity took place. Select explicit
|
|
48
|
+
sessions before sharing externally. Use `--workspace` or `--tool` to filter on recorded labels. Capture uses AUT's existing Claude Code, Codex, Pi,
|
|
49
|
+
Cursor CLI, and T3 Code sources and preserves their coverage limits. It skips recently active
|
|
50
|
+
sources by default. Session and event IDs refer to frozen evidence. A fresh capture can include
|
|
51
|
+
new events and produce a new ID.
|
|
52
|
+
|
|
53
|
+
If a session reports a byte cap, capture the project again with a larger per-source budget.
|
|
54
|
+
For example, use `aut evidence --content redacted capture --project /path/to/repo --max-mib 64`.
|
|
55
|
+
The default is 16 MiB. The audit records the budget and reports any remaining capture gaps.
|
|
56
|
+
|
|
57
|
+
## Choose what the client can read
|
|
58
|
+
|
|
59
|
+
Put `--content` before the operation, for example
|
|
60
|
+
`aut evidence --audit AUDIT_ID --content redacted event SESSION_ID EVENT_ID`.
|
|
61
|
+
|
|
62
|
+
| Mode | Result |
|
|
63
|
+
| --- | --- |
|
|
64
|
+
| `full` (default) | Unredacted text available from the selected capture. |
|
|
65
|
+
| `redacted` | Text with recognized secrets and identifiers masked. Review it before sharing. |
|
|
66
|
+
| `metadata` | Event references, kinds, counts, and coverage. Text search is unavailable. |
|
|
67
|
+
|
|
68
|
+
Running `aut evidence capture` authorizes a new private copy of available text in the AUT improvement
|
|
69
|
+
store. It retains normalized text, subject to each source's capture budget and exclusions.
|
|
70
|
+
Existing audits may contain only clipped, redacted text. Such events report `snapshot_redacted`
|
|
71
|
+
when the client requests their text. This includes a restored snapshot without its full-text record.
|
|
72
|
+
Capture the source again if it is still available.
|
|
73
|
+
|
|
74
|
+
Full text lives in separate private records. Existing judges, drafts, and the usage sink still
|
|
75
|
+
receive their existing inputs. Default restic backups exclude these records in the standard store
|
|
76
|
+
and the store named by `AUT_IMPROVE_ROOT`. Add the full-content directory to `backup_excludes`
|
|
77
|
+
when a different `--root` sits inside a backed-up path. Use `aut evidence --content redacted capture` to save only the
|
|
78
|
+
standard redacted snapshots. Bundles and MCP clients can narrow their content mode; they cannot
|
|
79
|
+
recover text omitted at capture or export. A local process can still pass a response to a hosted
|
|
80
|
+
model through its client. Choose the content mode and client together.
|
|
81
|
+
|
|
82
|
+
## Connect a local MCP client
|
|
83
|
+
|
|
84
|
+
Install the optional MCP extra in the environment that supplies `aut`.
|
|
85
|
+
|
|
86
|
+
```sh
|
|
87
|
+
pip install 'agent-usage-tracker[mcp]'
|
|
88
|
+
```
|
|
89
|
+
|
|
90
|
+
Set the client's MCP command to the absolute path to `aut`. Pass these arguments:
|
|
91
|
+
|
|
92
|
+
```json
|
|
93
|
+
["evidence", "--audit", "AUDIT_ID", "mcp"]
|
|
94
|
+
```
|
|
95
|
+
|
|
96
|
+
Use the same `--root` as capture if you use a custom store. Add `--content redacted` before
|
|
97
|
+
`mcp` to mask text. The server exposes `describe`, `search`, `session`, `event`, `compare`,
|
|
98
|
+
`commits`, `case`, and `ordinary_work` with the same scope and results as the CLI.
|
|
99
|
+
It provides read operations only.
|
|
100
|
+
|
|
101
|
+
An example prompt for the connected agent:
|
|
102
|
+
|
|
103
|
+
> Find recent setup failures and a comparable successful attempt. Cite the failed call and its
|
|
104
|
+
> result, list the differences you can observe, and propose one change we can test. State any
|
|
105
|
+
> missing context that affects the diagnosis.
|
|
106
|
+
|
|
107
|
+
## Read the evidence from another machine
|
|
108
|
+
|
|
109
|
+
Export the exact sessions you want to share:
|
|
110
|
+
|
|
111
|
+
```sh
|
|
112
|
+
aut evidence --audit AUDIT_ID --content redacted export investigation.json SESSION_ID OTHER_SESSION_ID
|
|
113
|
+
aut evidence --bundle investigation.json search --status error
|
|
114
|
+
```
|
|
115
|
+
|
|
116
|
+
Transfer the file through a channel you choose. A bundle is private plain JSON with an integrity
|
|
117
|
+
checksum. It can be opened without the original logs or store. AUT refuses to overwrite an existing
|
|
118
|
+
bundle. Treat a bundle as saved data. Revoke client access by stopping the server or replacing its
|
|
119
|
+
token. Copies recipients already have remain readable.
|
|
120
|
+
|
|
121
|
+
On a private host, load a bundle or use an audit already on that host. Supply a unique random
|
|
122
|
+
URL-safe token of 32 to 128 characters through `AUT_EVIDENCE_TOKEN`. For example, generate it with
|
|
123
|
+
`python -c 'import secrets; print(secrets.token_urlsafe(32))'` and store it with the service's secrets.
|
|
124
|
+
|
|
125
|
+
```sh
|
|
126
|
+
aut evidence --bundle investigation.json mcp --transport http --port 8765
|
|
127
|
+
ssh -N -L 8765:127.0.0.1:8765 user@host
|
|
128
|
+
```
|
|
129
|
+
|
|
130
|
+
Run the first command on the evidence host and the second on the client machine. Configure a client
|
|
131
|
+
that supports custom bearer headers to use `http://127.0.0.1:8765/mcp` with that token. The service
|
|
132
|
+
binds to loopback and scopes one token to the selected evidence. It supports explicit bearer
|
|
133
|
+
credentials over the tunnel. It does not run an OAuth authorization server or a multi-tenant
|
|
134
|
+
service. The host's normal AUT usage sink does not provide transcript access.
|
|
135
|
+
|
|
136
|
+
## Turn an observation into a repair
|
|
137
|
+
|
|
138
|
+
Freeze selected attempts in the private store. This creates the same type of case AUT uses for
|
|
139
|
+
proposals and studies. It does not run a judge.
|
|
140
|
+
|
|
141
|
+
```sh
|
|
142
|
+
aut evidence --audit AUDIT_ID freeze-case FAILED_SESSION_ID FAILED_EVENT_ID
|
|
143
|
+
aut evidence --audit AUDIT_ID freeze-case CLEAN_SESSION_ID CLEAN_EVENT_ID
|
|
144
|
+
aut evidence --audit AUDIT_ID compare FAILED_CASE_ID CLEAN_CASE_ID --cases
|
|
145
|
+
```
|
|
146
|
+
|
|
147
|
+
A case from a successful run can preserve a workflow worth reusing. Case results list commit IDs
|
|
148
|
+
observed in matched Git tool results when the capture contains them. Use the failing case when
|
|
149
|
+
drafting a repair. Set an explicit scope and target for the proposed change.
|
|
150
|
+
|
|
151
|
+
```sh
|
|
152
|
+
aut improve propose --case FAILED_CASE_ID --scope /path/to/repo \
|
|
153
|
+
--target /path/to/repo/AGENTS.md --before-file before.txt --after-file after.txt
|
|
154
|
+
```
|
|
155
|
+
|
|
156
|
+
Use the same improvement root for both commands. The
|
|
157
|
+
[study workflow](improvement-studies.md) links the candidate to the original evidence. Test an
|
|
158
|
+
isolated baseline and candidate with the same acceptance check. Record the result and compare
|
|
159
|
+
fresh eligible work before claiming a lasting benefit. The source configuration changes only
|
|
160
|
+
through the existing review and adoption process.
|
|
161
|
+
|
|
162
|
+
## Propose a project skill from successful work
|
|
163
|
+
|
|
164
|
+
Open the successful attempts and freeze their case IDs. Ask the reviewing agent to write a draft
|
|
165
|
+
`SKILL.md` with a matching `name`, a `description`, and the instructions. Cite at least two independent
|
|
166
|
+
sessions from the same project. Each citation identifies a successful result matched to a captured
|
|
167
|
+
call. AUT records the supporting cases and any observed commit IDs.
|
|
168
|
+
|
|
169
|
+
```sh
|
|
170
|
+
aut evidence --audit AUDIT_ID propose-skill --scope /path/to/repo \
|
|
171
|
+
--target .agents/skills/check-setup/SKILL.md --draft-file draft.md \
|
|
172
|
+
--proof CASE_ID:EVENT_ID --proof OTHER_CASE_ID:OTHER_EVENT_ID \
|
|
173
|
+
--candidate /path/outside/the/repo
|
|
174
|
+
```
|
|
175
|
+
|
|
176
|
+
Review the proposal with `aut improve inbox list` and the existing review workflow. AUT can prepare
|
|
177
|
+
an addition or update, and stage it in the independent candidate. A new file has a null baseline
|
|
178
|
+
hash. To trial the addition, explicitly list the file with a null value in `files.baseline`. List its
|
|
179
|
+
SHA-256 hash in `files.candidate`. This binds the trial to the absence of the skill before the change.
|
|
180
|
+
A successful tool result is evidence for the cited step. Review whether the proposed instructions
|
|
181
|
+
apply to the whole workflow before you adopt the skill.
|
|
182
|
+
|
|
183
|
+
## Run a programmable investigation
|
|
184
|
+
|
|
185
|
+
A recipe can use evidence from one session to decide what to search for in another.
|
|
186
|
+
Give it explicit audits or bundles, then return citations, a timeline, a hypothesis,
|
|
187
|
+
or a statistic with its counting rule.
|
|
188
|
+
|
|
189
|
+
```sh
|
|
190
|
+
pip install 'agent-usage-tracker[code]'
|
|
191
|
+
aut evidence code-catalog
|
|
192
|
+
aut evidence --audit FIRST --content redacted run-code follow.py --include-audit SECOND --save
|
|
193
|
+
```
|
|
194
|
+
|
|
195
|
+
For example, put this program in `follow.py`. It takes a lead from a match in the
|
|
196
|
+
first audit and searches the second. The last expression becomes the JSON result.
|
|
197
|
+
|
|
198
|
+
```python
|
|
199
|
+
seed = read("search", query="setup failed", limit=1)
|
|
200
|
+
found = []
|
|
201
|
+
if seed["data"]["items"]:
|
|
202
|
+
query = seed["data"]["items"][0]["preview"].split()[-1]
|
|
203
|
+
cursor = ""
|
|
204
|
+
while cursor is not None:
|
|
205
|
+
page = read("search", scope="2", lead=seed["receipt"], query=query, cursor=cursor)
|
|
206
|
+
for hit in page["data"]["items"]:
|
|
207
|
+
found.append({"session_id": hit["session_id"], "event_id": hit["event_id"]})
|
|
208
|
+
cursor = page["data"]["next_cursor"]
|
|
209
|
+
found
|
|
210
|
+
```
|
|
211
|
+
|
|
212
|
+
Use `-` in place of the path to read source from stdin. Pass a JSON object with
|
|
213
|
+
`--params` to populate the program's `params` variable. The `scopes` variable lists
|
|
214
|
+
the actual coverage and content grant per source. Scope `1` is the main audit or
|
|
215
|
+
bundle. Additional audits come next, followed by additional bundles. Each type
|
|
216
|
+
retains command-line order. Bundles can narrow the content grant. Code mode
|
|
217
|
+
defaults to redacted access. Existing direct commands keep their own defaults.
|
|
218
|
+
|
|
219
|
+
The catalog lists the read operations, their parameters, and defaults. Set `lead`
|
|
220
|
+
to an earlier read's receipt to record which observation motivated a later read.
|
|
221
|
+
Each successful read records available event references, a result digest, and
|
|
222
|
+
whether a further page was offered. Continue an empty page if it has a cursor.
|
|
223
|
+
Check scope and session coverage before drawing conclusions. A returned citation
|
|
224
|
+
identifies an observed record. A person or a reviewed rule must check what it
|
|
225
|
+
supports. Snapshot IDs may repeat across audits or change after a recapture.
|
|
226
|
+
When comparing distinct source sessions, use an available source identity and
|
|
227
|
+
report cases where it is unknown. The readers expose recorded Git outcomes and
|
|
228
|
+
study evidence. They require independently verified links for joins with usage
|
|
229
|
+
or records outside the selected evidence.
|
|
230
|
+
|
|
231
|
+
Recipes run in fresh local Monty sessions. The runner limits source size, host
|
|
232
|
+
calls, evidence bytes, output, and wall time. Failed execution reports an error.
|
|
233
|
+
Recipes use the Python subset supported by Monty. They get no general host
|
|
234
|
+
filesystem, network, or shell operations. Plan a stopping rule for recursive
|
|
235
|
+
searches, and report which leads you left open if you stop early.
|
|
236
|
+
|
|
237
|
+
Save a run with `--save` to keep its program, parameters, scopes, and read receipts
|
|
238
|
+
in the private AUT store. Saving can retain sensitive text in a program or its
|
|
239
|
+
parameters. The default AUT backup exclusions cover this directory under the
|
|
240
|
+
standard store and `AUT_IMPROVE_ROOT`. Add a backup exclusion when using a different
|
|
241
|
+
`--root` inside a backed-up directory.
|
|
242
|
+
|
|
243
|
+
```sh
|
|
244
|
+
aut evidence --audit FIRST replay-code RUN_ID --include-audit SECOND
|
|
245
|
+
aut evidence --audit NEW_FIRST run-code --recipe RUN_ID --include-audit NEW_SECOND
|
|
246
|
+
```
|
|
247
|
+
|
|
248
|
+
Replay checks the original scope and compares the result and read record with the
|
|
249
|
+
saved digest. Deleted inputs can prevent replay. Changes in AUT or Monty versions
|
|
250
|
+
can change the digest. Use `--recipe` to run the same program on new inputs and
|
|
251
|
+
optionally pass new parameters.
|
|
252
|
+
|
|
253
|
+
Install `agent-usage-tracker[mcp,code]` to use a recipe through MCP. If you use
|
|
254
|
+
the pinned-wheel runtime installer, pass `--code` to include both optional dependencies. Add `--code`
|
|
255
|
+
to the server arguments after `mcp` and include any additional scopes there.
|
|
256
|
+
The server adds `code_catalog` and `run_code(source, params)`. The client receives
|
|
257
|
+
the result under the connection's grant. Restart the MCP server to select a
|
|
258
|
+
changed bundle. A client connected to a hosted model can
|
|
259
|
+
forward that result. Session observability settings govern a separate delivery
|
|
260
|
+
path.
|
|
261
|
+
|
|
262
|
+
## Send session evidence to a configured receiver
|
|
263
|
+
|
|
264
|
+
Set `AUT_SESSION_OBSERVABILITY_URL` to an HTTPS endpoint that accepts AUT evidence
|
|
265
|
+
bundle JSON by POST. Loopback HTTP is available for local development. Set
|
|
266
|
+
`AUT_SESSION_OBSERVABILITY_TOKEN` if your receiver requires bearer authentication.
|
|
267
|
+
AUT uses a dedicated endpoint for session evidence. The existing usage sink
|
|
268
|
+
continues to receive usage data.
|
|
269
|
+
|
|
270
|
+
With a destination configured, `aut evidence capture` sends the new audit with
|
|
271
|
+
redacted content by default. Set `AUT_SESSION_OBSERVABILITY=metadata` for event
|
|
272
|
+
references and coverage, `full` for available raw session text, or `off` to stop
|
|
273
|
+
delivery. Use `capture --no-observe` to keep a single capture local. A capture or
|
|
274
|
+
bundle can further restrict what is sent. Redaction masks recognized patterns.
|
|
275
|
+
Review the content policy and the receiver's retention before sharing private work.
|
|
276
|
+
|
|
277
|
+
A delivery failure appears in the capture's JSON as `observability.status=failed`.
|
|
278
|
+
The private audit remains available. Retry with
|
|
279
|
+
`aut evidence --audit AUDIT_ID observe`, or select a bundle to deliver a smaller
|
|
280
|
+
set of sessions. The request carries an `Idempotency-Key` that matches the bundle
|
|
281
|
+
digest. A receiver can use it to deduplicate identical deliveries.
|