argus-testing 0.8.0__tar.gz → 0.9.0__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.
Files changed (180) hide show
  1. {argus_testing-0.8.0 → argus_testing-0.9.0}/AGENTS.md +10 -1
  2. {argus_testing-0.8.0 → argus_testing-0.9.0}/CHANGELOG.md +17 -0
  3. {argus_testing-0.8.0 → argus_testing-0.9.0}/PKG-INFO +46 -16
  4. {argus_testing-0.8.0 → argus_testing-0.9.0}/README.md +45 -15
  5. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/monitoring.py +8 -0
  6. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/manager.py +8 -0
  7. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/browser.py +10 -0
  8. argus_testing-0.9.0/argus/entrypoints/mcp/overhead.py +40 -0
  9. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/entrypoints/mcp/results.py +21 -3
  10. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/entrypoints/mcp/server.py +33 -5
  11. argus_testing-0.9.0/argus/entrypoints/mcp/snapshots.py +77 -0
  12. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/gateway.py +8 -1
  13. argus_testing-0.9.0/argus/evaluation/overhead.py +121 -0
  14. argus_testing-0.9.0/argus/install_check.py +68 -0
  15. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/mcp_server.py +1 -1
  16. argus_testing-0.9.0/argus/measurement.py +61 -0
  17. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/reporter.py +2 -0
  18. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/settings.py +2 -2
  19. {argus_testing-0.8.0 → argus_testing-0.9.0}/docs/agent-protocol.md +48 -0
  20. {argus_testing-0.8.0 → argus_testing-0.9.0}/docs/architecture.md +13 -0
  21. {argus_testing-0.8.0 → argus_testing-0.9.0}/docs/evaluation.md +41 -0
  22. argus_testing-0.9.0/examples/response-overhead.json +170 -0
  23. {argus_testing-0.8.0 → argus_testing-0.9.0}/llms-install.md +10 -0
  24. {argus_testing-0.8.0 → argus_testing-0.9.0}/pyproject.toml +1 -1
  25. {argus_testing-0.8.0 → argus_testing-0.9.0}/server.json +2 -2
  26. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_monitoring.py +24 -0
  27. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_release_metadata.py +9 -0
  28. argus_testing-0.9.0/tests/test_response_efficiency.py +235 -0
  29. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_stdio_protocol.py +41 -0
  30. {argus_testing-0.8.0 → argus_testing-0.9.0}/.gitignore +0 -0
  31. {argus_testing-0.8.0 → argus_testing-0.9.0}/CONTRIBUTING.md +0 -0
  32. {argus_testing-0.8.0 → argus_testing-0.9.0}/LICENSE +0 -0
  33. {argus_testing-0.8.0 → argus_testing-0.9.0}/PRIVACY.md +0 -0
  34. {argus_testing-0.8.0 → argus_testing-0.9.0}/SECURITY.md +0 -0
  35. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/__init__.py +0 -0
  36. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/_compat.py +0 -0
  37. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/access.py +0 -0
  38. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/__init__.py +0 -0
  39. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/assets/__init__.py +0 -0
  40. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/assets/dashboard.css +0 -0
  41. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/assets/dashboard.html +0 -0
  42. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/assets/dashboard.js +0 -0
  43. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/journal.py +0 -0
  44. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/logging.py +0 -0
  45. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/reporting/__init__.py +0 -0
  46. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/reporting/common.py +0 -0
  47. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/reporting/html.py +0 -0
  48. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/reporting/json.py +0 -0
  49. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/reporting/junit.py +0 -0
  50. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/reporting/privacy.py +0 -0
  51. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/reporting/sarif.py +0 -0
  52. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/scripts/__init__.py +0 -0
  53. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/scripts/check_layout_js.js +0 -0
  54. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/scripts/extract_elements_js.js +0 -0
  55. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/scripts/extract_page_content_js.js +0 -0
  56. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/scripts/extract_performance_js.js +0 -0
  57. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/scripts/find_visible_element_js.js +0 -0
  58. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/adapters/scripts/inspect_element_js.js +0 -0
  59. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/__init__.py +0 -0
  60. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/actions.py +0 -0
  61. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/capsules.py +0 -0
  62. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/capture.py +0 -0
  63. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/context.py +0 -0
  64. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/coverage.py +0 -0
  65. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/events.py +0 -0
  66. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/evidence.py +0 -0
  67. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/expectations.py +0 -0
  68. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/findings.py +0 -0
  69. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/finish.py +0 -0
  70. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/guidance.py +0 -0
  71. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/inspection.py +0 -0
  72. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/journeys.py +0 -0
  73. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/lifecycle.py +0 -0
  74. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/navigation.py +0 -0
  75. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/network_storage.py +0 -0
  76. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/observation.py +0 -0
  77. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/operations.py +0 -0
  78. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/payloads.py +0 -0
  79. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/probes.py +0 -0
  80. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/progress.py +0 -0
  81. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/regression.py +0 -0
  82. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/screen.py +0 -0
  83. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/screenshots.py +0 -0
  84. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/session.py +0 -0
  85. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/telemetry.py +0 -0
  86. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/text.py +0 -0
  87. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/verification.py +0 -0
  88. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/application/verifiers.py +0 -0
  89. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/bench/__init__.py +0 -0
  90. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/bench/__main__.py +0 -0
  91. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/bench/agent_runner.py +0 -0
  92. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/bench/runner.py +0 -0
  93. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/bench/scenarios_buggytasks.py +0 -0
  94. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/bench/scenarios_darkshop.py +0 -0
  95. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/bootstrap.py +0 -0
  96. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/cleanup.py +0 -0
  97. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/cli.py +0 -0
  98. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/config.py +0 -0
  99. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/detector.py +0 -0
  100. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/differ.py +0 -0
  101. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/domain/__init__.py +0 -0
  102. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/domain/contracts.py +0 -0
  103. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/domain/detector.py +0 -0
  104. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/domain/models.py +0 -0
  105. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/domain/replay.py +0 -0
  106. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/domain/verification.py +0 -0
  107. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/entrypoints/__init__.py +0 -0
  108. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/entrypoints/mcp/__init__.py +0 -0
  109. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/entrypoints/mcp/catalog.py +0 -0
  110. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/__init__.py +0 -0
  111. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/__main__.py +0 -0
  112. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/adjudication.py +0 -0
  113. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/agent_fixture.py +0 -0
  114. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/browser_baseline.py +0 -0
  115. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/codex_runner.py +0 -0
  116. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/fixture.py +0 -0
  117. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/models.py +0 -0
  118. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/scoring.py +0 -0
  119. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/evaluation/smoke.py +0 -0
  120. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/explorer.py +0 -0
  121. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/models.py +0 -0
  122. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/planner.py +0 -0
  123. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/ports.py +0 -0
  124. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/resolver.py +0 -0
  125. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/runtime.py +0 -0
  126. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/screen/__init__.py +0 -0
  127. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/screen/backend.py +0 -0
  128. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/screen/permissions.py +0 -0
  129. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/screen/safety.py +0 -0
  130. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/screen/validate.py +0 -0
  131. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/security.py +0 -0
  132. {argus_testing-0.8.0 → argus_testing-0.9.0}/argus/storage.py +0 -0
  133. {argus_testing-0.8.0 → argus_testing-0.9.0}/docs/decisions.md +0 -0
  134. {argus_testing-0.8.0 → argus_testing-0.9.0}/docs/evaluation-0.8.md +0 -0
  135. {argus_testing-0.8.0 → argus_testing-0.9.0}/docs/implementation-0.7.md +0 -0
  136. {argus_testing-0.8.0 → argus_testing-0.9.0}/examples/agent-evaluation-optimized-evidence.json +0 -0
  137. {argus_testing-0.8.0 → argus_testing-0.9.0}/examples/agent-evaluation-optimized.json +0 -0
  138. {argus_testing-0.8.0 → argus_testing-0.9.0}/examples/agent-evaluation-pilot-evidence.json +0 -0
  139. {argus_testing-0.8.0 → argus_testing-0.9.0}/examples/agent-evaluation-pilot.json +0 -0
  140. {argus_testing-0.8.0 → argus_testing-0.9.0}/examples/focus.yaml +0 -0
  141. {argus_testing-0.8.0 → argus_testing-0.9.0}/examples/mcp-config.json +0 -0
  142. {argus_testing-0.8.0 → argus_testing-0.9.0}/glama.json +0 -0
  143. {argus_testing-0.8.0 → argus_testing-0.9.0}/human-eye-fixture/app.py +0 -0
  144. {argus_testing-0.8.0 → argus_testing-0.9.0}/scripts/check_distributions.py +0 -0
  145. {argus_testing-0.8.0 → argus_testing-0.9.0}/test-site/app.py +0 -0
  146. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/__init__.py +0 -0
  147. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/conftest.py +0 -0
  148. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_action_targeting.py +0 -0
  149. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_adjudication.py +0 -0
  150. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_agent_results.py +0 -0
  151. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_agent_workflow.py +0 -0
  152. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_architecture_runtime.py +0 -0
  153. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_bench_report.py +0 -0
  154. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_browser_shadow.py +0 -0
  155. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_codex_evaluation.py +0 -0
  156. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_cross_stack.py +0 -0
  157. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_detector_console.py +0 -0
  158. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_detector_network.py +0 -0
  159. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_distribution_contents.py +0 -0
  160. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_evaluation.py +0 -0
  161. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_expectation_oracle.py +0 -0
  162. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_final_hardening.py +0 -0
  163. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_journal_regression.py +0 -0
  164. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_mcp_contract.py +0 -0
  165. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_regression_cli.py +0 -0
  166. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_replay_minimize.py +0 -0
  167. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_replay_receipt.py +0 -0
  168. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_replay_trace.py +0 -0
  169. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_reporter_badge.py +0 -0
  170. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_reproduction_receipt.py +0 -0
  171. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_resolver.py +0 -0
  172. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_runtime_limits.py +0 -0
  173. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_screen_permissions.py +0 -0
  174. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_screen_safety.py +0 -0
  175. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_security_boundaries.py +0 -0
  176. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_session_lifecycle.py +0 -0
  177. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_state_capsules.py +0 -0
  178. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_storage_cleanup.py +0 -0
  179. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_substrate.py +0 -0
  180. {argus_testing-0.8.0 → argus_testing-0.9.0}/tests/test_visual_inspection.py +0 -0
@@ -27,6 +27,15 @@ argus-mcp --doctor # macOS Screen Recording + Accessibility check
27
27
 
28
28
  The default profile is `core`. For structured-result-aware hosts, use
29
29
  `--tool-profile agent --response-mode compact` for 19 focused tools.
30
+
31
+ `--response-mode delta` is opt-in for snapshot-aware clients. Only observation
32
+ fields listed in the protocol may be inherited from an exact `base_id`; action
33
+ and verification results are current-call-only. If context or a base is missing,
34
+ call `observe` for full state. Never infer a passing check from unchanged fields.
35
+ Use `python -m argus.install_check` to verify installed versions and stdio without
36
+ a browser or model, and `python -m argus.evaluation.overhead` for deterministic
37
+ byte measurements. Bytes are not model tokens. Monitoring retains numeric
38
+ overhead only; do not add page content or arguments to metrics.
30
39
  Use `--tool-profile screen` for native
31
40
  macOS review and `--tool-profile full` only when the broader network,
32
41
  storage, tabs, crawl, and coordinate surface is needed. The
@@ -73,7 +82,7 @@ result includes captured stages. See [the protocol](docs/agent-protocol.md).
73
82
  | `type_into(description, text)` / `select_into(description, value)` | Inputs and dropdowns by description. |
74
83
  | `test_action(target, expect=...)` | Click + before/after diff in one call. Pass `expect` to PREDICT the outcome ({"count":{"label":"tasks","delta":1}}, {"gains":"Buy milk"}, {"removes":...}, {"text_present":...}, {"toast":...}, {"url_changed":true}) and Argus reports MATCH / SURPRISE — a surprise is a bug lead. Also shows CROSS-STACK: which requests the click fired (methods/statuses) and a CHECK nudge when a message appeared without a matching write. |
75
84
  | `verify_persistence(expect, target_text, after_url)` | Forces a fresh GET; returns target presence and the page observation. Check persistence once, repeat only to resolve uncertainty. A matching symptom is not proof of a semantic defect. |
76
- | `capsule_save(name, liveness_marker)` / `capsule_restore(name)` | Snapshot the logged-in/seeded state (cookies+storage) after minting it through the UI, then restore it later (with a mandatory live/stale re-check). Restore is a CLEAN replace, so save→branch A→restore→branch B runs two journeys from a byte-identical state for differential testing. |
85
+ | `capsule_save(name, liveness_marker)` / `capsule_restore(name)` | Snapshot cookies and page storage after establishing state through the UI. Restore replaces the captured client state and checks a saved liveness marker when provided; without one, validity remains unverified. It does not reset backend data or guarantee identical application state across journeys. |
77
86
  | `regression_check()` | Re-test the findings journaled in prior runs against the CURRENT build: STILL-PRESENT / NO-LONGER-REPRODUCES / INCONCLUSIVE. "Did my fix land, did anything come back?" |
78
87
  | `inspect_element(description)` | Computed styles + ARIA + outerHTML + truncation flag for interactive or visible non-interactive content. |
79
88
  | `check_layout()` | Bounded viewport signals for horizontal overflow, clipped text, small targets, and fixed/sticky overlays. |
@@ -1,5 +1,22 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.9.0 - 2026-10-25
4
+
5
+ - Report the Argus product version in the MCP handshake instead of the SDK
6
+ version. Add a no-browser installed-package stdio check and run it against
7
+ fresh release wheels outside the source checkout.
8
+ - Remove redundant output-schema display titles without changing validation
9
+ constraints or tool display titles.
10
+ - Add opt-in delta observations with explicit base IDs, field removals, full
11
+ resynchronization, lifecycle invalidation and bounded redacted retention.
12
+ Preserve errors, current-call verdicts, image blocks and compatibility defaults.
13
+ - Measure response bytes and instrumented browser waits, observation, screenshots,
14
+ report export and encoding. Expose numeric totals through loopback status and
15
+ Prometheus; keep page content and input values out of monitoring.
16
+ - Add reproducible, lossless encoding comparisons with navigation controls and
17
+ gateway discovery/response measurements. Do not equate byte savings with model
18
+ token savings or claim a new model effectiveness evaluation.
19
+
3
20
  ## 0.8.0 - 2026-10-25
4
21
 
5
22
  - Add an optional 19-tool agent profile with concise discovery descriptions,
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.5
2
2
  Name: argus-testing
3
- Version: 0.8.0
3
+ Version: 0.9.0
4
4
  Summary: Agent-first testing MCP with evidence-backed coverage, persistence checks and reproduction receipts
5
5
  Project-URL: Homepage, https://yichenwu.dev/projects/argus
6
6
  Project-URL: Documentation, https://github.com/chriswu727/argus#readme
@@ -67,6 +67,8 @@ Argus is an [MCP](https://modelcontextprotocol.io/) server that gives your codin
67
67
  [![Capability ceiling](https://img.shields.io/badge/bench-34%2F34-brightgreen)](#benchmarks)
68
68
  [![License: MIT](https://img.shields.io/badge/license-MIT-green)](LICENSE)
69
69
 
70
+ [![Argus Testing — Autonomous QA MCP server](https://mcprush.com/badge/argus-testing-autonomous-qa-mcp/card.svg)](https://mcprush.com/to/argus-testing-autonomous-qa-mcp?ref=badge)
71
+
70
72
  [Product page](https://yichenwu.dev/projects/argus) · [Quick start](#quick-start) · [Why Argus](#why-argus-is-different) · [Compared](#how-it-compares) · [Tools](#tool-surface) · [Benchmarks](#benchmarks)
71
73
 
72
74
  </div>
@@ -94,10 +96,14 @@ the phase summary in the final result. See the [agent protocol](docs/agent-proto
94
96
  - **Conservative checks:** missing or malformed evidence stays unknown, not a
95
97
  guessed pass or a fabricated bug. Findings link to stable session references.
96
98
  - **Focused agent mode:** `argus-mcp --tool-profile agent --response-mode compact`
97
- exposes 19 common testing tools with concise discovery descriptions and keeps structured evidence
99
+ exposes 19 common testing tools with concise discovery descriptions, preserves structured evidence,
98
100
  and avoids repeating migrated payloads as prose. Compatible mode remains the default.
99
101
  - **Visible progress:** real phases, goal counts and check counts over MCP;
100
102
  `--monitor-port 8765` adds the local live dashboard without exposing goal text.
103
+ - **Measured response overhead (0.9):** smaller output schemas, numeric payload
104
+ and component-timing counters, and opt-in `--response-mode delta` for clients
105
+ that retain observation snapshots. Missing bases require a full `observe`;
106
+ verification outcomes and image blocks are never replaced by cached verdicts.
101
107
 
102
108
  Does installing Argus improve an agent? Measure it with the
103
109
  [paired evaluation harness](docs/evaluation.md): the same model, task, fixture,
@@ -111,6 +117,11 @@ declared-goal coverage. Argus used more time and tokens. On four repeated pilot
111
117
  cases, the targeted update reduced Argus's median calls from 17 to 9; this small
112
118
  synthetic sample does not establish general superiority.
113
119
 
120
+ The no-model [response encoding benchmark](docs/evaluation.md#response-overhead)
121
+ compares identical synthetic evidence, including navigation invalidation. Delta
122
+ mode reduces aggregate result JSON by 47.09% versus compact mode across those
123
+ 24 responses; this measures bytes, **not agent token usage or testing quality**.
124
+
114
125
  <details>
115
126
  <summary><b>Live monitoring from a real local fixture run</b></summary>
116
127
 
@@ -148,10 +159,10 @@ flowchart LR
148
159
  A(["observe"]) --> B{"looks wrong?"}
149
160
  B -->|not sure| C["act: click · type · resize · verify"]
150
161
  C --> A
151
- B -->|bug| D["verify_persistence — reload from a clean state"]
152
- D -->|symptom repeats| E(["VERIFIED"])
153
- D -->|symptom gone or uncertain| F(["not certified"])
154
- E --> G[["report: HTML · JSON · JUnit · SARIF"]]
162
+ B -->|suspected defect| D["verify_persistence — inspect a fresh load"]
163
+ D --> E["record_bug with an independent recheck"]
164
+ E --> F(["reproduced / not reproduced / inconclusive"])
165
+ F --> G[["report: HTML · JSON · JUnit · SARIF"]]
155
166
  ```
156
167
 
157
168
  The agent supplies the judgment. Argus supplies concise QA guidance, a description-keyed tool surface (`click_what("Login button")`, not `click(7)`), a goal coverage ledger, and a **reproduction-receipt engine** that distinguishes observed, reproduced and inconclusive symptoms.
@@ -212,7 +223,7 @@ The button adds Argus to Cursor; run the Chromium installation command above onc
212
223
 
213
224
  The default `core` profile exposes the primary web-testing workflow without flooding the host with every specialist tool. Use `uvx --from argus-testing argus-mcp --list-tools` to inspect the selected profile, `--tool-profile screen` for native macOS testing, or `--tool-profile full` for the entire advanced surface. `ARGUS_TOOL_PROFILE` provides the same setting through the environment.
214
225
 
215
- For agents that read structured MCP results, use the focused 0.8 configuration:
226
+ For agents that read structured MCP results, use the focused configuration:
216
227
 
217
228
  ```bash
218
229
  uvx --from argus-testing argus-mcp --tool-profile agent --response-mode compact --monitor-port 8765
@@ -222,6 +233,21 @@ Read returned observations before issuing another `observe`. Batch related goals
222
233
  with `coverage_update(goal=["edit profile", "reload profile"], status="in_progress")`;
223
234
  finish them with observed evidence, not an assumed pass. The monitor stays local.
224
235
 
236
+ `--response-mode delta` is an additional opt-in for clients that can retain and
237
+ merge snapshot IDs. Keep `compact` if the host cannot reliably resolve bases.
238
+ See the [snapshot contract](docs/agent-protocol.md#incremental-observations).
239
+
240
+ To check installation, product version, stdio handshake, discovery and structured
241
+ errors without opening a browser or calling a model:
242
+
243
+ ```bash
244
+ uvx --from argus-testing python -m argus.install_check
245
+ ```
246
+
247
+ Editable installs must be reinstalled after a version change so their package
248
+ metadata matches the source. MCP `serverInfo.version` now reports Argus's version,
249
+ not the MCP SDK version.
250
+
225
251
  Then just ask, in your agent session:
226
252
 
227
253
  > **"Test my app at http://localhost:3000 — find real bugs."**
@@ -313,9 +339,9 @@ selected URL is printed to **stderr**, so stdio MCP messages stay intact.
313
339
  | `/` | Local live dashboard |
314
340
  | `/healthz` | HTTP process liveness |
315
341
  | `/readyz` | Runtime initialized and not shutting down; 503 otherwise |
316
- | `/api/status` | Versioned state snapshot, counts and capture drops |
342
+ | `/api/status` | Versioned state snapshot, counts, capture drops and numeric overhead |
317
343
  | `/api/events` | Server-Sent Events, `Last-Event-ID` reconnect and gap notices |
318
- | `/metrics` | Prometheus text counters and operation duration sums/counts |
344
+ | `/metrics` | Prometheus operation, response-byte and component-timing counters |
319
345
 
320
346
  The monitor binds only to `127.0.0.1`, rejects unrelated Host/Origin headers,
321
347
  and has no action endpoints. It emits no page contents, URLs, typed values,
@@ -361,16 +387,20 @@ application data; see [Privacy](PRIVACY.md).
361
387
 
362
388
  ## Why Argus is different
363
389
 
364
- **Existing testing tools only test what you script.** Playwright and Cypress run the assertions you wrote. Argus *discovers* bugs you didn't think to test for — and then does the thing an LLM alone can't be trusted to do: **proves them.**
390
+ **Give exploratory testing a repeatable evidence workflow.** An agent can use
391
+ Argus to investigate journeys beyond prewritten checks, inspect what changed,
392
+ and independently recheck suspected symptoms. The agent still decides whether
393
+ the observed behavior violates the application's requirements; a reproduction
394
+ receipt is supporting evidence, not a guarantee of correctness or zero false positives.
365
395
 
366
396
  | | |
367
397
  |---|---|
368
398
  | **Agent-directed & black-box** | Give your agent a URL and scope. It explores through Argus without requiring application source access or prewritten test scripts. |
369
399
  | **Coverage contract** | Optional natural-language goals, user constraints, discovered pages, and time budget stay visible throughout the session and in the final report. |
370
- | **Reproduction receipts** | Before certifying a bug, it re-loads the page from a clean state and re-confirms the symptom. Engineered for **zero false-certifications.** |
371
- | **Finds human-eye bugs** | Fake "Only 3 left!" scarcity, a "Saved" toast that doesn't save, a sale badge where the price didn't drop, a stale navbar after a rename. Static analysis catches none of these. |
372
- | **Discover → guard** | Findings are journaled; `argus-regression` re-checks them on every build with **zero LLM cost** and a non-zero exit — a real CI gate against known bugs coming back. |
373
- | **Machine-readable** | Every report also emits JSON, JUnit, and SARIF — so findings gate a pipeline and surface as inline **GitHub PR annotations.** |
400
+ | **Reproduction receipts** | Independent checks distinguish reproduced, not-reproduced and inconclusive symptoms. Reloads preserve authentication and storage by default; replay can repeat writes. |
401
+ | **Runtime evidence** | Inspect misleading success messages, cross-page inconsistencies and persistence failures in the running app. The agent interprets this evidence against the intended behavior. |
402
+ | **Discover → guard** | Add `argus-regression` to CI to recheck eligible journaled findings without model calls. Exit codes distinguish reproduced symptoms, operational failures and inconclusive checks. |
403
+ | **Machine-readable** | JSON, JUnit and SARIF exports support downstream processing. CI gates and code-scanning uploads require integration; URL-based findings do not guarantee source-line PR annotations. |
374
404
 
375
405
  ---
376
406
 
@@ -424,7 +454,7 @@ not the basis for current product claims. See [methodology](docs/evaluation.md).
424
454
 
425
455
  **BuggyTasks** (`:5555`) — 22 mechanical bugs in a task app: console errors, dead links, fake delete (UI says "deleted!" but data persists on refresh), auth bypass, NaN dates, off-by-one counts, race conditions. The "scripted E2E could find these" tier.
426
456
 
427
- **DarkShop** (`:5556`) — 12 human-eye bugs in a polished-looking store: hardcoded "Only 3 left!" scarcity, `-50%` badges where sale price equals original, a "free shipping over $50" banner contradicted by a flat `$5` at checkout, inverted visual hierarchy ("Add to Cart" demoted under a prominent "Subscribe"), cross-page state drift (rename sticks on /account, navbar greeting doesn't). **Static analysis catches roughly none of these** — they require an agent that reads the page and *reasons.*
457
+ **DarkShop** (`:5556`) — 12 seeded runtime and visual scenarios in a polished-looking store: hardcoded "Only 3 left!" scarcity, `-50%` badges where sale price equals original, a "free shipping over $50" banner contradicted by a flat `$5` at checkout, inverted visual hierarchy ("Add to Cart" demoted under a prominent "Subscribe"), cross-page state drift (rename sticks on /account, navbar greeting doesn't). These fixtures exercise comparisons between visible claims and observed behavior; they are not a benchmark against static analysis tools.
428
458
 
429
459
  ```bash
430
460
  python test-site/app.py # BuggyTasks :5555
@@ -475,7 +505,7 @@ uvx --from argus-testing argus-mcp --tool-profile full --list-tools
475
505
  | `test_action` · `test_form` | Perform a description-keyed action or form submission and return the resulting state diff in one round trip. |
476
506
  | `check_links` · `check_performance` | Probe current-page internal links and expose raw browser performance metrics without auto-certifying generic audit findings. |
477
507
  | `regression_check` | Re-test journaled findings for the current origin without requiring another discovery pass. |
478
- | `record_bug` · `record_observation` | Record a reproducible defect with evidence and receipt, or keep a qualitative review note separate from certified bugs. |
508
+ | `record_bug` · `record_observation` | Record a finding with evidence and its recheck outcome, or keep a qualitative review note separate. Inspect the receipt rather than treating recording as confirmation. |
479
509
  | `end_session` | Close the active session and emit HTML, JSON, JUnit, and SARIF reports. |
480
510
 
481
511
  </details>
@@ -17,6 +17,8 @@ Argus is an [MCP](https://modelcontextprotocol.io/) server that gives your codin
17
17
  [![Capability ceiling](https://img.shields.io/badge/bench-34%2F34-brightgreen)](#benchmarks)
18
18
  [![License: MIT](https://img.shields.io/badge/license-MIT-green)](LICENSE)
19
19
 
20
+ [![Argus Testing — Autonomous QA MCP server](https://mcprush.com/badge/argus-testing-autonomous-qa-mcp/card.svg)](https://mcprush.com/to/argus-testing-autonomous-qa-mcp?ref=badge)
21
+
20
22
  [Product page](https://yichenwu.dev/projects/argus) · [Quick start](#quick-start) · [Why Argus](#why-argus-is-different) · [Compared](#how-it-compares) · [Tools](#tool-surface) · [Benchmarks](#benchmarks)
21
23
 
22
24
  </div>
@@ -44,10 +46,14 @@ the phase summary in the final result. See the [agent protocol](docs/agent-proto
44
46
  - **Conservative checks:** missing or malformed evidence stays unknown, not a
45
47
  guessed pass or a fabricated bug. Findings link to stable session references.
46
48
  - **Focused agent mode:** `argus-mcp --tool-profile agent --response-mode compact`
47
- exposes 19 common testing tools with concise discovery descriptions and keeps structured evidence
49
+ exposes 19 common testing tools with concise discovery descriptions, preserves structured evidence,
48
50
  and avoids repeating migrated payloads as prose. Compatible mode remains the default.
49
51
  - **Visible progress:** real phases, goal counts and check counts over MCP;
50
52
  `--monitor-port 8765` adds the local live dashboard without exposing goal text.
53
+ - **Measured response overhead (0.9):** smaller output schemas, numeric payload
54
+ and component-timing counters, and opt-in `--response-mode delta` for clients
55
+ that retain observation snapshots. Missing bases require a full `observe`;
56
+ verification outcomes and image blocks are never replaced by cached verdicts.
51
57
 
52
58
  Does installing Argus improve an agent? Measure it with the
53
59
  [paired evaluation harness](docs/evaluation.md): the same model, task, fixture,
@@ -61,6 +67,11 @@ declared-goal coverage. Argus used more time and tokens. On four repeated pilot
61
67
  cases, the targeted update reduced Argus's median calls from 17 to 9; this small
62
68
  synthetic sample does not establish general superiority.
63
69
 
70
+ The no-model [response encoding benchmark](docs/evaluation.md#response-overhead)
71
+ compares identical synthetic evidence, including navigation invalidation. Delta
72
+ mode reduces aggregate result JSON by 47.09% versus compact mode across those
73
+ 24 responses; this measures bytes, **not agent token usage or testing quality**.
74
+
64
75
  <details>
65
76
  <summary><b>Live monitoring from a real local fixture run</b></summary>
66
77
 
@@ -98,10 +109,10 @@ flowchart LR
98
109
  A(["observe"]) --> B{"looks wrong?"}
99
110
  B -->|not sure| C["act: click · type · resize · verify"]
100
111
  C --> A
101
- B -->|bug| D["verify_persistence — reload from a clean state"]
102
- D -->|symptom repeats| E(["VERIFIED"])
103
- D -->|symptom gone or uncertain| F(["not certified"])
104
- E --> G[["report: HTML · JSON · JUnit · SARIF"]]
112
+ B -->|suspected defect| D["verify_persistence — inspect a fresh load"]
113
+ D --> E["record_bug with an independent recheck"]
114
+ E --> F(["reproduced / not reproduced / inconclusive"])
115
+ F --> G[["report: HTML · JSON · JUnit · SARIF"]]
105
116
  ```
106
117
 
107
118
  The agent supplies the judgment. Argus supplies concise QA guidance, a description-keyed tool surface (`click_what("Login button")`, not `click(7)`), a goal coverage ledger, and a **reproduction-receipt engine** that distinguishes observed, reproduced and inconclusive symptoms.
@@ -162,7 +173,7 @@ The button adds Argus to Cursor; run the Chromium installation command above onc
162
173
 
163
174
  The default `core` profile exposes the primary web-testing workflow without flooding the host with every specialist tool. Use `uvx --from argus-testing argus-mcp --list-tools` to inspect the selected profile, `--tool-profile screen` for native macOS testing, or `--tool-profile full` for the entire advanced surface. `ARGUS_TOOL_PROFILE` provides the same setting through the environment.
164
175
 
165
- For agents that read structured MCP results, use the focused 0.8 configuration:
176
+ For agents that read structured MCP results, use the focused configuration:
166
177
 
167
178
  ```bash
168
179
  uvx --from argus-testing argus-mcp --tool-profile agent --response-mode compact --monitor-port 8765
@@ -172,6 +183,21 @@ Read returned observations before issuing another `observe`. Batch related goals
172
183
  with `coverage_update(goal=["edit profile", "reload profile"], status="in_progress")`;
173
184
  finish them with observed evidence, not an assumed pass. The monitor stays local.
174
185
 
186
+ `--response-mode delta` is an additional opt-in for clients that can retain and
187
+ merge snapshot IDs. Keep `compact` if the host cannot reliably resolve bases.
188
+ See the [snapshot contract](docs/agent-protocol.md#incremental-observations).
189
+
190
+ To check installation, product version, stdio handshake, discovery and structured
191
+ errors without opening a browser or calling a model:
192
+
193
+ ```bash
194
+ uvx --from argus-testing python -m argus.install_check
195
+ ```
196
+
197
+ Editable installs must be reinstalled after a version change so their package
198
+ metadata matches the source. MCP `serverInfo.version` now reports Argus's version,
199
+ not the MCP SDK version.
200
+
175
201
  Then just ask, in your agent session:
176
202
 
177
203
  > **"Test my app at http://localhost:3000 — find real bugs."**
@@ -263,9 +289,9 @@ selected URL is printed to **stderr**, so stdio MCP messages stay intact.
263
289
  | `/` | Local live dashboard |
264
290
  | `/healthz` | HTTP process liveness |
265
291
  | `/readyz` | Runtime initialized and not shutting down; 503 otherwise |
266
- | `/api/status` | Versioned state snapshot, counts and capture drops |
292
+ | `/api/status` | Versioned state snapshot, counts, capture drops and numeric overhead |
267
293
  | `/api/events` | Server-Sent Events, `Last-Event-ID` reconnect and gap notices |
268
- | `/metrics` | Prometheus text counters and operation duration sums/counts |
294
+ | `/metrics` | Prometheus operation, response-byte and component-timing counters |
269
295
 
270
296
  The monitor binds only to `127.0.0.1`, rejects unrelated Host/Origin headers,
271
297
  and has no action endpoints. It emits no page contents, URLs, typed values,
@@ -311,16 +337,20 @@ application data; see [Privacy](PRIVACY.md).
311
337
 
312
338
  ## Why Argus is different
313
339
 
314
- **Existing testing tools only test what you script.** Playwright and Cypress run the assertions you wrote. Argus *discovers* bugs you didn't think to test for — and then does the thing an LLM alone can't be trusted to do: **proves them.**
340
+ **Give exploratory testing a repeatable evidence workflow.** An agent can use
341
+ Argus to investigate journeys beyond prewritten checks, inspect what changed,
342
+ and independently recheck suspected symptoms. The agent still decides whether
343
+ the observed behavior violates the application's requirements; a reproduction
344
+ receipt is supporting evidence, not a guarantee of correctness or zero false positives.
315
345
 
316
346
  | | |
317
347
  |---|---|
318
348
  | **Agent-directed & black-box** | Give your agent a URL and scope. It explores through Argus without requiring application source access or prewritten test scripts. |
319
349
  | **Coverage contract** | Optional natural-language goals, user constraints, discovered pages, and time budget stay visible throughout the session and in the final report. |
320
- | **Reproduction receipts** | Before certifying a bug, it re-loads the page from a clean state and re-confirms the symptom. Engineered for **zero false-certifications.** |
321
- | **Finds human-eye bugs** | Fake "Only 3 left!" scarcity, a "Saved" toast that doesn't save, a sale badge where the price didn't drop, a stale navbar after a rename. Static analysis catches none of these. |
322
- | **Discover → guard** | Findings are journaled; `argus-regression` re-checks them on every build with **zero LLM cost** and a non-zero exit — a real CI gate against known bugs coming back. |
323
- | **Machine-readable** | Every report also emits JSON, JUnit, and SARIF — so findings gate a pipeline and surface as inline **GitHub PR annotations.** |
350
+ | **Reproduction receipts** | Independent checks distinguish reproduced, not-reproduced and inconclusive symptoms. Reloads preserve authentication and storage by default; replay can repeat writes. |
351
+ | **Runtime evidence** | Inspect misleading success messages, cross-page inconsistencies and persistence failures in the running app. The agent interprets this evidence against the intended behavior. |
352
+ | **Discover → guard** | Add `argus-regression` to CI to recheck eligible journaled findings without model calls. Exit codes distinguish reproduced symptoms, operational failures and inconclusive checks. |
353
+ | **Machine-readable** | JSON, JUnit and SARIF exports support downstream processing. CI gates and code-scanning uploads require integration; URL-based findings do not guarantee source-line PR annotations. |
324
354
 
325
355
  ---
326
356
 
@@ -374,7 +404,7 @@ not the basis for current product claims. See [methodology](docs/evaluation.md).
374
404
 
375
405
  **BuggyTasks** (`:5555`) — 22 mechanical bugs in a task app: console errors, dead links, fake delete (UI says "deleted!" but data persists on refresh), auth bypass, NaN dates, off-by-one counts, race conditions. The "scripted E2E could find these" tier.
376
406
 
377
- **DarkShop** (`:5556`) — 12 human-eye bugs in a polished-looking store: hardcoded "Only 3 left!" scarcity, `-50%` badges where sale price equals original, a "free shipping over $50" banner contradicted by a flat `$5` at checkout, inverted visual hierarchy ("Add to Cart" demoted under a prominent "Subscribe"), cross-page state drift (rename sticks on /account, navbar greeting doesn't). **Static analysis catches roughly none of these** — they require an agent that reads the page and *reasons.*
407
+ **DarkShop** (`:5556`) — 12 seeded runtime and visual scenarios in a polished-looking store: hardcoded "Only 3 left!" scarcity, `-50%` badges where sale price equals original, a "free shipping over $50" banner contradicted by a flat `$5` at checkout, inverted visual hierarchy ("Add to Cart" demoted under a prominent "Subscribe"), cross-page state drift (rename sticks on /account, navbar greeting doesn't). These fixtures exercise comparisons between visible claims and observed behavior; they are not a benchmark against static analysis tools.
378
408
 
379
409
  ```bash
380
410
  python test-site/app.py # BuggyTasks :5555
@@ -425,7 +455,7 @@ uvx --from argus-testing argus-mcp --tool-profile full --list-tools
425
455
  | `test_action` · `test_form` | Perform a description-keyed action or form submission and return the resulting state diff in one round trip. |
426
456
  | `check_links` · `check_performance` | Probe current-page internal links and expose raw browser performance metrics without auto-certifying generic audit findings. |
427
457
  | `regression_check` | Re-test journaled findings for the current origin without requiring another discovery pass. |
428
- | `record_bug` · `record_observation` | Record a reproducible defect with evidence and receipt, or keep a qualitative review note separate from certified bugs. |
458
+ | `record_bug` · `record_observation` | Record a finding with evidence and its recheck outcome, or keep a qualitative review note separate. Inspect the receipt rather than treating recording as confirmation. |
429
459
  | `end_session` | Close the active session and emit HTML, JSON, JUnit, and SARIF reports. |
430
460
 
431
461
  </details>
@@ -132,6 +132,14 @@ class Monitor:
132
132
  f"argus_operation_duration_seconds_count{{operation={label}}} {count}",
133
133
  ]
134
134
  )
135
+ for name, value in sorted(runtime.response_totals.items()):
136
+ lines.append(f"argus_response_total{{measurement={json.dumps(name)}}} {value}")
137
+ for name, value in sorted(runtime.component_seconds.items()):
138
+ label = json.dumps(name)
139
+ lines.append(f"argus_component_seconds_sum{{component={label}}} {value}")
140
+ lines.append(
141
+ f"argus_component_seconds_count{{component={label}}} {runtime.component_calls[name]}"
142
+ )
135
143
  return PlainTextResponse("\n".join(lines) + "\n", media_type="text/plain; version=0.0.4")
136
144
 
137
145
  async def asset(self, request: Request) -> Response:
@@ -48,6 +48,9 @@ class SessionManager:
48
48
  self.calls: Counter[str] = Counter()
49
49
  self.failures: Counter[str] = Counter()
50
50
  self.durations: dict[str, float] = defaultdict(float)
51
+ self.response_totals: Counter[str] = Counter()
52
+ self.component_seconds: Counter[str] = Counter()
53
+ self.component_calls: Counter[str] = Counter()
51
54
  self.current_operation: str | None = None
52
55
  self.operation_started: float | None = None
53
56
  self.started = time.monotonic()
@@ -262,6 +265,11 @@ class SessionManager:
262
265
  "screenshots": len(session.screenshots),
263
266
  "capture_drops": drops,
264
267
  "events_dropped": self.events.dropped,
268
+ "overhead": {
269
+ "response_totals": dict(self.response_totals),
270
+ "component_seconds": dict(self.component_seconds),
271
+ "component_calls": dict(self.component_calls),
272
+ },
265
273
  }
266
274
 
267
275
  async def shutdown(self) -> None:
@@ -8,6 +8,7 @@ from playwright.async_api import Browser, BrowserContext, Page, async_playwright
8
8
 
9
9
  from argus.access import AccessPolicy
10
10
  from argus.domain.contracts import OperationError
11
+ from argus.measurement import timed
11
12
  from argus.security import _redact as _redact
12
13
  from argus.security import _redact_headers as _redact_headers
13
14
  from argus.security import redact_body
@@ -205,6 +206,7 @@ class BrowserDriver:
205
206
  async def click_resolved(self, index: int, elements: list[InteractiveElement]) -> None:
206
207
  await self._locator(index, elements).click(timeout=5000)
207
208
 
209
+ @timed("browser_wait")
208
210
  async def settle(self) -> None:
209
211
  await self._page.wait_for_load_state("networkidle", timeout=10_000)
210
212
 
@@ -807,6 +809,7 @@ class BrowserDriver:
807
809
 
808
810
  # -- waits --
809
811
 
812
+ @timed("browser_wait")
810
813
  async def wait_for_text(self, text: str, timeout_s: float = 10.0) -> bool:
811
814
  """Poll until the given text appears anywhere in the page body."""
812
815
  try:
@@ -819,6 +822,7 @@ class BrowserDriver:
819
822
  except Exception:
820
823
  return False
821
824
 
825
+ @timed("browser_wait")
822
826
  async def wait_for_request(
823
827
  self,
824
828
  url_substring: str,
@@ -854,6 +858,7 @@ class BrowserDriver:
854
858
 
855
859
  # -- navigation --
856
860
 
861
+ @timed("browser_wait")
857
862
  async def _settle_network(self, page=None, timeout: int = 6_000):
858
863
  """Best-effort wait for network to go idle. A never-resolving fetch (a
859
864
  stuck 'Loading…' spinner) means networkidle NEVER fires, so this must
@@ -885,6 +890,7 @@ class BrowserDriver:
885
890
 
886
891
  # -- state extraction --
887
892
 
893
+ @timed("observation")
888
894
  async def get_state(self, page=None) -> PageState:
889
895
  page = page or self._page
890
896
  if page is None:
@@ -1222,6 +1228,7 @@ class BrowserDriver:
1222
1228
  except Exception:
1223
1229
  pass
1224
1230
 
1231
+ @timed("browser_wait")
1225
1232
  async def _settle_dom(self, cap_ms: int = 1500, quiet_ms: int = 300) -> None:
1226
1233
  """Wait until the DOM stops mutating for `quiet_ms` (capped at `cap_ms`).
1227
1234
  networkidle doesn't cover pure-JS DOM updates that fire after a click —
@@ -1245,6 +1252,7 @@ class BrowserDriver:
1245
1252
  except Exception:
1246
1253
  pass # navigation destroyed the context, eval blocked, etc. — best-effort
1247
1254
 
1255
+ @timed("browser_wait")
1248
1256
  async def _settle_visual(self, cap_ms: int = 1200) -> None:
1249
1257
  try:
1250
1258
  await self._page.evaluate(
@@ -1652,12 +1660,14 @@ class BrowserDriver:
1652
1660
  await self._page.evaluate("(d) => window.scrollBy(0, d)", int(px))
1653
1661
  await asyncio.sleep(0.4)
1654
1662
 
1663
+ @timed("screenshot")
1655
1664
  async def screenshot(self, path: str, full_page: bool = False) -> str:
1656
1665
  Path(path).parent.mkdir(parents=True, exist_ok=True)
1657
1666
  await self._settle_visual()
1658
1667
  await self._page.screenshot(path=path, full_page=full_page)
1659
1668
  return path
1660
1669
 
1670
+ @timed("screenshot")
1661
1671
  async def element_screenshot(self, selector: str, path: str) -> Optional[str]:
1662
1672
  """Capture just the bounds of one element (Playwright crops natively)."""
1663
1673
  Path(path).parent.mkdir(parents=True, exist_ok=True)
@@ -0,0 +1,40 @@
1
+ from __future__ import annotations
2
+
3
+ import json
4
+
5
+ from mcp.types import CallToolResult, Tool
6
+
7
+
8
+ def json_bytes(value) -> int:
9
+ return len(json.dumps(value, ensure_ascii=False, separators=(",", ":")).encode("utf-8"))
10
+
11
+
12
+ def response_bytes(result: CallToolResult) -> dict[str, int]:
13
+ return {
14
+ "result_json_bytes": len(result.model_dump_json(by_alias=True, exclude_none=True).encode()),
15
+ "text_utf8_bytes": sum(
16
+ len(block.text.encode()) for block in result.content if block.type == "text"
17
+ ),
18
+ "structured_json_bytes": json_bytes(result.structuredContent)
19
+ if result.structuredContent is not None
20
+ else 0,
21
+ "image_base64_bytes": sum(
22
+ len(block.data.encode()) for block in result.content if block.type == "image"
23
+ ),
24
+ "images": sum(block.type == "image" for block in result.content),
25
+ }
26
+
27
+
28
+ def discovery_bytes(tools: list[Tool], instructions: str = "") -> dict[str, int]:
29
+ return {
30
+ "tools": len(tools),
31
+ "tools_json_bytes": json_bytes(
32
+ [tool.model_dump(mode="json", by_alias=True, exclude_none=True) for tool in tools]
33
+ ),
34
+ "instructions_utf8_bytes": len(instructions.encode()),
35
+ "description_utf8_bytes": sum(len((tool.description or "").encode()) for tool in tools),
36
+ "input_schema_json_bytes": sum(json_bytes(tool.inputSchema) for tool in tools),
37
+ "output_schema_json_bytes": sum(
38
+ json_bytes(tool.outputSchema) for tool in tools if tool.outputSchema
39
+ ),
40
+ }
@@ -4,14 +4,25 @@ from typing import Any, Literal
4
4
 
5
5
  from mcp.server.fastmcp.utilities.types import Image
6
6
  from mcp.types import CallToolResult, TextContent
7
- from pydantic import BaseModel, Field
7
+ from pydantic import BaseModel, ConfigDict, Field
8
8
 
9
9
  from argus.application.progress import Progress
10
10
  from argus.domain.contracts import ImageArtifact, OperationResult, ToolText
11
+ from argus.measurement import timed
11
12
  from argus.security import redacted_copy
12
13
 
14
+ from .snapshots import ObservationSnapshots
15
+
16
+
17
+ def _omit_schema_titles(schema: dict) -> None:
18
+ schema.pop("title", None)
19
+ for field_schema in schema.get("properties", {}).values():
20
+ field_schema.pop("title", None)
21
+
13
22
 
14
23
  class AgentResult(BaseModel):
24
+ model_config = ConfigDict(json_schema_extra=_omit_schema_titles)
25
+
15
26
  schema_version: Literal[1] = 1
16
27
  operation: str
17
28
  status: Literal["completed", "error"]
@@ -36,8 +47,13 @@ _RECOVERY = {
36
47
  }
37
48
 
38
49
 
50
+ @timed("response_encode")
39
51
  def to_mcp_result(
40
- name: str, result: OperationResult, progress: Progress, mode: str = "compatible"
52
+ name: str,
53
+ result: OperationResult,
54
+ progress: Progress,
55
+ mode: str = "compatible",
56
+ snapshots: ObservationSnapshots | None = None,
41
57
  ) -> CallToolResult:
42
58
  value = result.value
43
59
  code = result.error.value if result.error else None
@@ -56,13 +72,15 @@ def to_mcp_result(
56
72
  ]
57
73
  if artifacts:
58
74
  data["artifacts"] = artifacts
59
- if mode == "compact" and not code and isinstance(value, ToolText) and data:
75
+ if mode in {"compact", "delta"} and not code and isinstance(value, ToolText) and data:
60
76
  content = [
61
77
  TextContent(
62
78
  type="text",
63
79
  text=f"{name}: completed. Read structuredContent.data for evidence; completion is not a test verdict.",
64
80
  )
65
81
  ]
82
+ if mode == "delta" and snapshots is not None:
83
+ data = snapshots.encode(data, name, result.run_id, error=bool(code))
66
84
  envelope = AgentResult(
67
85
  operation=name,
68
86
  status="error" if code else "completed",
@@ -8,10 +8,12 @@ from mcp.server.fastmcp.exceptions import ToolError
8
8
  from mcp.server.fastmcp.utilities.types import Image
9
9
  from mcp.types import ToolAnnotations
10
10
 
11
+ from argus import __version__
11
12
  from argus.application.manager import SessionManager
12
13
  from argus.application.operations import OPERATIONS
13
14
  from argus.application.progress import progress_scope
14
15
  from argus.domain.contracts import ImageArtifact
16
+ from argus.measurement import measurement_scope
15
17
 
16
18
  from .catalog import (
17
19
  _AGENT_TOOL_DESCRIPTIONS,
@@ -22,7 +24,9 @@ from .catalog import (
22
24
  _SCREEN_TOOL_NAMES,
23
25
  _TOOL_TITLES,
24
26
  )
27
+ from .overhead import response_bytes
25
28
  from .results import AgentResult, to_mcp_result
29
+ from .snapshots import FIELDS, ObservationSnapshots
26
30
 
27
31
  INSTRUCTIONS = (
28
32
  "Start with start_session; follow its protocol. Exercise real journeys and "
@@ -60,16 +64,29 @@ def tool_names(profile: str) -> set[str]:
60
64
  raise ValueError("tool profile must be agent, core, screen, or full")
61
65
 
62
66
 
63
- def bind_tool(runtime: SessionManager, operation):
67
+ def bind_tool(runtime: SessionManager, operation, snapshots: ObservationSnapshots | None = None):
64
68
  @wraps(operation)
65
69
  async def call(*args, ctx: Context | None = None, **kwargs):
66
70
  async def notify(sequence: int, message: str) -> None:
67
71
  if ctx is not None:
68
72
  await ctx.report_progress(sequence, message=message)
69
73
 
70
- with progress_scope(notify if ctx is not None else None) as progress:
71
- result = await runtime.execute(operation.__name__, operation, *args, **kwargs)
72
- return to_mcp_result(operation.__name__, result, progress, runtime.settings.response_mode)
74
+ with measurement_scope() as measurements:
75
+ with progress_scope(notify if ctx is not None else None) as progress:
76
+ try:
77
+ result = await runtime.execute(operation.__name__, operation, *args, **kwargs)
78
+ except BaseException:
79
+ if snapshots is not None:
80
+ snapshots.clear()
81
+ raise
82
+ encoded = to_mcp_result(
83
+ operation.__name__, result, progress, runtime.settings.response_mode, snapshots
84
+ )
85
+ runtime.response_totals.update(response_bytes(encoded))
86
+ runtime.response_totals["responses"] += 1
87
+ runtime.component_seconds.update(measurements.seconds)
88
+ runtime.component_calls.update(measurements.calls)
89
+ return encoded
73
90
 
74
91
  signature = inspect.signature(operation, eval_str=True)
75
92
  parameters = [
@@ -88,10 +105,21 @@ def bind_tool(runtime: SessionManager, operation):
88
105
 
89
106
  def create_server(runtime: SessionManager, profile: str = "core") -> FastMCP:
90
107
  server = FastMCP("argus", instructions=INSTRUCTIONS)
108
+ # FastMCP 1.x does not expose the low-level server's product version argument.
109
+ server._mcp_server.version = __version__
110
+ snapshots = ObservationSnapshots()
111
+ if runtime.settings.response_mode == "delta":
112
+ server._mcp_server.instructions += (
113
+ " Delta mode: retain observation snapshots by ID. In delta snapshots, inherit only "
114
+ + ", ".join(sorted(FIELDS))
115
+ + "; replace returned fields and delete snapshot.removed fields. base_id must match. "
116
+ "Other fields are current-call-only. On a missing base call observe for full state. "
117
+ "Never infer unchanged from an unavailable base."
118
+ )
91
119
  for name in sorted(tool_names(profile)):
92
120
  read_only = name in _READ_ONLY_TOOL_NAMES
93
121
  server.add_tool(
94
- bind_tool(runtime, OPERATIONS[name]),
122
+ bind_tool(runtime, OPERATIONS[name], snapshots),
95
123
  name=name,
96
124
  description=_AGENT_TOOL_DESCRIPTIONS[name] if profile == "agent" else None,
97
125
  title=_TOOL_TITLES[name],