auto-any 0.1.0__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.

Potentially problematic release.


This version of auto-any might be problematic. Click here for more details.

Files changed (212) hide show
  1. auto_any-0.1.0/.env.example +52 -0
  2. auto_any-0.1.0/.github/workflows/ci.yml +29 -0
  3. auto_any-0.1.0/.gitignore +14 -0
  4. auto_any-0.1.0/ARCHITECTURE.md +171 -0
  5. auto_any-0.1.0/PKG-INFO +289 -0
  6. auto_any-0.1.0/README.md +248 -0
  7. auto_any-0.1.0/SPEC-2-ENGINE.md +196 -0
  8. auto_any-0.1.0/SPEC-3-FEATURES.md +171 -0
  9. auto_any-0.1.0/SPEC-4-PLATFORM.md +147 -0
  10. auto_any-0.1.0/SPEC-5-VENDORS.md +162 -0
  11. auto_any-0.1.0/SPEC.md +238 -0
  12. auto_any-0.1.0/docs/known-failure-modes.md +98 -0
  13. auto_any-0.1.0/install/README.md +75 -0
  14. auto_any-0.1.0/install/install.ps1 +141 -0
  15. auto_any-0.1.0/install/install.sh +144 -0
  16. auto_any-0.1.0/pyproject.toml +56 -0
  17. auto_any-0.1.0/src/auto_any/__init__.py +11 -0
  18. auto_any-0.1.0/src/auto_any/api/__init__.py +1 -0
  19. auto_any-0.1.0/src/auto_any/api/automations.py +209 -0
  20. auto_any-0.1.0/src/auto_any/api/packs.py +56 -0
  21. auto_any-0.1.0/src/auto_any/api/profiles.py +18 -0
  22. auto_any-0.1.0/src/auto_any/api/receipts.py +38 -0
  23. auto_any-0.1.0/src/auto_any/browsergate/__init__.py +27 -0
  24. auto_any-0.1.0/src/auto_any/browsergate/appium_driver.py +82 -0
  25. auto_any-0.1.0/src/auto_any/browsergate/base.py +45 -0
  26. auto_any-0.1.0/src/auto_any/browsergate/playwright_driver.py +158 -0
  27. auto_any-0.1.0/src/auto_any/browsergate/policy.py +83 -0
  28. auto_any-0.1.0/src/auto_any/browsergate/profile_attach.py +30 -0
  29. auto_any-0.1.0/src/auto_any/browsergate/profile_clone.py +134 -0
  30. auto_any-0.1.0/src/auto_any/browsergate/profile_port.py +28 -0
  31. auto_any-0.1.0/src/auto_any/browsergate/profile_selection.py +273 -0
  32. auto_any-0.1.0/src/auto_any/browsergate/registry.py +11 -0
  33. auto_any-0.1.0/src/auto_any/browsergate/selection.py +49 -0
  34. auto_any-0.1.0/src/auto_any/browsergate/selenium_driver.py +135 -0
  35. auto_any-0.1.0/src/auto_any/browsergate/session.py +38 -0
  36. auto_any-0.1.0/src/auto_any/browsergate/vision_port.py +30 -0
  37. auto_any-0.1.0/src/auto_any/build_info.py +81 -0
  38. auto_any-0.1.0/src/auto_any/capabilities/__init__.py +6 -0
  39. auto_any-0.1.0/src/auto_any/capabilities/base.py +45 -0
  40. auto_any-0.1.0/src/auto_any/capabilities/browser.py +20 -0
  41. auto_any-0.1.0/src/auto_any/capabilities/extract.py +38 -0
  42. auto_any-0.1.0/src/auto_any/capabilities/gate.py +281 -0
  43. auto_any-0.1.0/src/auto_any/capabilities/http.py +14 -0
  44. auto_any-0.1.0/src/auto_any/capabilities/registry.py +10 -0
  45. auto_any-0.1.0/src/auto_any/capabilities/shell.py +14 -0
  46. auto_any-0.1.0/src/auto_any/capabilities/verify.py +33 -0
  47. auto_any-0.1.0/src/auto_any/cli.py +261 -0
  48. auto_any-0.1.0/src/auto_any/conductor/__init__.py +1 -0
  49. auto_any-0.1.0/src/auto_any/conductor/budget.py +29 -0
  50. auto_any-0.1.0/src/auto_any/conductor/conductor.py +10 -0
  51. auto_any-0.1.0/src/auto_any/conductor/prompts/act.md +21 -0
  52. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/act/auth_flow.md +1 -0
  53. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/act/extraction.md +1 -0
  54. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/act/form_fill.md +1 -0
  55. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/act/navigation.md +1 -0
  56. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/act/pagination.md +1 -0
  57. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/act/verification.md +1 -0
  58. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/plan/auth_flow.md +1 -0
  59. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/plan/extraction.md +1 -0
  60. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/plan/form_fill.md +1 -0
  61. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/plan/navigation.md +1 -0
  62. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/plan/pagination.md +1 -0
  63. auto_any-0.1.0/src/auto_any/conductor/prompts/fragments/plan/verification.md +1 -0
  64. auto_any-0.1.0/src/auto_any/conductor/prompts/plan.md +1 -0
  65. auto_any-0.1.0/src/auto_any/conductor/shield.py +31 -0
  66. auto_any-0.1.0/src/auto_any/config.py +109 -0
  67. auto_any-0.1.0/src/auto_any/engine/__init__.py +11 -0
  68. auto_any-0.1.0/src/auto_any/engine/base.py +17 -0
  69. auto_any-0.1.0/src/auto_any/engine/llm.py +142 -0
  70. auto_any-0.1.0/src/auto_any/engine/local.py +262 -0
  71. auto_any-0.1.0/src/auto_any/engine/remote.py +30 -0
  72. auto_any-0.1.0/src/auto_any/main.py +39 -0
  73. auto_any-0.1.0/src/auto_any/mcp_server.py +282 -0
  74. auto_any-0.1.0/src/auto_any/memory/__init__.py +0 -0
  75. auto_any-0.1.0/src/auto_any/memory/site_profiles.py +213 -0
  76. auto_any-0.1.0/src/auto_any/packs/__init__.py +17 -0
  77. auto_any-0.1.0/src/auto_any/packs/base.py +88 -0
  78. auto_any-0.1.0/src/auto_any/packs/builtin/__init__.py +16 -0
  79. auto_any-0.1.0/src/auto_any/packs/builtin/form_fill.py +37 -0
  80. auto_any-0.1.0/src/auto_any/packs/builtin/linkedin_apply.py +47 -0
  81. auto_any-0.1.0/src/auto_any/packs/builtin/login_download.py +41 -0
  82. auto_any-0.1.0/src/auto_any/packs/builtin/scrape_table.py +34 -0
  83. auto_any-0.1.0/src/auto_any/packs/builtin/smoke_test.py +35 -0
  84. auto_any-0.1.0/src/auto_any/packs/compile.py +34 -0
  85. auto_any-0.1.0/src/auto_any/packs/discovery.py +82 -0
  86. auto_any-0.1.0/src/auto_any/packs/registry.py +36 -0
  87. auto_any-0.1.0/src/auto_any/replay/__init__.py +4 -0
  88. auto_any-0.1.0/src/auto_any/replay/distiller.py +98 -0
  89. auto_any-0.1.0/src/auto_any/replay/ledger.py +178 -0
  90. auto_any-0.1.0/src/auto_any/replay/service.py +28 -0
  91. auto_any-0.1.0/src/auto_any/replay/signature.py +15 -0
  92. auto_any-0.1.0/src/auto_any/replay/store.py +20 -0
  93. auto_any-0.1.0/src/auto_any/runtime/__init__.py +1 -0
  94. auto_any-0.1.0/src/auto_any/runtime/container.py +82 -0
  95. auto_any-0.1.0/src/auto_any/schemas.py +31 -0
  96. auto_any-0.1.0/src/auto_any/security.py +58 -0
  97. auto_any-0.1.0/src/auto_any/state/__init__.py +1 -0
  98. auto_any-0.1.0/src/auto_any/state/pending.py +44 -0
  99. auto_any-0.1.0/src/auto_any/taperoom/__init__.py +41 -0
  100. auto_any-0.1.0/src/auto_any/telemetry.py +15 -0
  101. auto_any-0.1.0/src/auto_any/vendors/__init__.py +1 -0
  102. auto_any-0.1.0/src/auto_any/vendors/contracts.py +33 -0
  103. auto_any-0.1.0/src/auto_any/vendors/efa.py +114 -0
  104. auto_any-0.1.0/src/auto_any/vendors/exporter.py +161 -0
  105. auto_any-0.1.0/src/auto_any/vendors/scorer.py +184 -0
  106. auto_any-0.1.0/src/replayable_test_runner/__init__.py +3 -0
  107. auto_any-0.1.0/src/replayable_test_runner/abstractions/__init__.py +53 -0
  108. auto_any-0.1.0/src/replayable_test_runner/abstractions/checkout.py +22 -0
  109. auto_any-0.1.0/src/replayable_test_runner/abstractions/embedding.py +25 -0
  110. auto_any-0.1.0/src/replayable_test_runner/abstractions/executor.py +29 -0
  111. auto_any-0.1.0/src/replayable_test_runner/abstractions/llm.py +55 -0
  112. auto_any-0.1.0/src/replayable_test_runner/abstractions/planner.py +21 -0
  113. auto_any-0.1.0/src/replayable_test_runner/abstractions/pr_source.py +29 -0
  114. auto_any-0.1.0/src/replayable_test_runner/abstractions/reporter.py +19 -0
  115. auto_any-0.1.0/src/replayable_test_runner/abstractions/reviewer.py +22 -0
  116. auto_any-0.1.0/src/replayable_test_runner/abstractions/runner.py +51 -0
  117. auto_any-0.1.0/src/replayable_test_runner/abstractions/signature.py +16 -0
  118. auto_any-0.1.0/src/replayable_test_runner/abstractions/store.py +56 -0
  119. auto_any-0.1.0/src/replayable_test_runner/cli.py +128 -0
  120. auto_any-0.1.0/src/replayable_test_runner/domain/__init__.py +33 -0
  121. auto_any-0.1.0/src/replayable_test_runner/domain/pr.py +32 -0
  122. auto_any-0.1.0/src/replayable_test_runner/domain/receipt.py +95 -0
  123. auto_any-0.1.0/src/replayable_test_runner/domain/results.py +51 -0
  124. auto_any-0.1.0/src/replayable_test_runner/domain/review.py +35 -0
  125. auto_any-0.1.0/src/replayable_test_runner/domain/signature.py +21 -0
  126. auto_any-0.1.0/src/replayable_test_runner/infra/__init__.py +5 -0
  127. auto_any-0.1.0/src/replayable_test_runner/infra/checkout_git.py +48 -0
  128. auto_any-0.1.0/src/replayable_test_runner/infra/fs_safety.py +168 -0
  129. auto_any-0.1.0/src/replayable_test_runner/infra/pr_github.py +91 -0
  130. auto_any-0.1.0/src/replayable_test_runner/infra/signatures.py +73 -0
  131. auto_any-0.1.0/src/replayable_test_runner/infra/state_json.py +130 -0
  132. auto_any-0.1.0/src/replayable_test_runner/infra/store_json.py +112 -0
  133. auto_any-0.1.0/src/replayable_test_runner/infra/vcs.py +34 -0
  134. auto_any-0.1.0/src/replayable_test_runner/pipeline/__init__.py +1 -0
  135. auto_any-0.1.0/src/replayable_test_runner/pipeline/assertions.py +264 -0
  136. auto_any-0.1.0/src/replayable_test_runner/pipeline/container.py +91 -0
  137. auto_any-0.1.0/src/replayable_test_runner/pipeline/executor.py +83 -0
  138. auto_any-0.1.0/src/replayable_test_runner/pipeline/planner.py +281 -0
  139. auto_any-0.1.0/src/replayable_test_runner/pipeline/prompts.py +56 -0
  140. auto_any-0.1.0/src/replayable_test_runner/pipeline/repair.py +243 -0
  141. auto_any-0.1.0/src/replayable_test_runner/pipeline/reporter.py +77 -0
  142. auto_any-0.1.0/src/replayable_test_runner/pipeline/reviewer.py +252 -0
  143. auto_any-0.1.0/src/replayable_test_runner/pipeline/runner.py +296 -0
  144. auto_any-0.1.0/src/replayable_test_runner/providers/__init__.py +8 -0
  145. auto_any-0.1.0/src/replayable_test_runner/providers/embedders.py +75 -0
  146. auto_any-0.1.0/src/replayable_test_runner/providers/llm.py +285 -0
  147. auto_any-0.1.0/src/replayable_test_runner/runners/__init__.py +11 -0
  148. auto_any-0.1.0/src/replayable_test_runner/runners/http.py +522 -0
  149. auto_any-0.1.0/src/replayable_test_runner/runners/http_probe.py +138 -0
  150. auto_any-0.1.0/src/replayable_test_runner/runners/playwright.py +76 -0
  151. auto_any-0.1.0/src/replayable_test_runner/runners/playwright_chromium.py +308 -0
  152. auto_any-0.1.0/src/replayable_test_runner/runners/shell.py +423 -0
  153. auto_any-0.1.0/src/replayable_test_runner/settings.py +58 -0
  154. auto_any-0.1.0/src/replayable_test_runner/support/__init__.py +1 -0
  155. auto_any-0.1.0/src/replayable_test_runner/support/errors.py +23 -0
  156. auto_any-0.1.0/src/replayable_test_runner/support/logging.py +28 -0
  157. auto_any-0.1.0/src/replayable_test_runner/support/net.py +139 -0
  158. auto_any-0.1.0/src/replayable_test_runner/support/registry.py +45 -0
  159. auto_any-0.1.0/tests/test_api.py +196 -0
  160. auto_any-0.1.0/tests/test_artifacts.py +41 -0
  161. auto_any-0.1.0/tests/test_artifacts_per_run.py +229 -0
  162. auto_any-0.1.0/tests/test_budget.py +5 -0
  163. auto_any-0.1.0/tests/test_build_info.py +74 -0
  164. auto_any-0.1.0/tests/test_capability_gate.py +27 -0
  165. auto_any-0.1.0/tests/test_cli.py +25 -0
  166. auto_any-0.1.0/tests/test_distiller.py +7 -0
  167. auto_any-0.1.0/tests/test_e2e_login_real.py +310 -0
  168. auto_any-0.1.0/tests/test_e2e_login_remember.py +234 -0
  169. auto_any-0.1.0/tests/test_e2e_real_browser.py +95 -0
  170. auto_any-0.1.0/tests/test_efa_scorer.py +62 -0
  171. auto_any-0.1.0/tests/test_engine_local.py +61 -0
  172. auto_any-0.1.0/tests/test_engine_multicap.py +39 -0
  173. auto_any-0.1.0/tests/test_exporters.py +57 -0
  174. auto_any-0.1.0/tests/test_extract.py +33 -0
  175. auto_any-0.1.0/tests/test_http_replay.py +198 -0
  176. auto_any-0.1.0/tests/test_human_in_loop.py +98 -0
  177. auto_any-0.1.0/tests/test_krhub_ledger_recovery.py +65 -0
  178. auto_any-0.1.0/tests/test_lane_parity.py +355 -0
  179. auto_any-0.1.0/tests/test_ledger.py +143 -0
  180. auto_any-0.1.0/tests/test_llm_swap.py +188 -0
  181. auto_any-0.1.0/tests/test_login_recovery.py +144 -0
  182. auto_any-0.1.0/tests/test_mcp_replay.py +131 -0
  183. auto_any-0.1.0/tests/test_mcp_server.py +44 -0
  184. auto_any-0.1.0/tests/test_named_profiles.py +65 -0
  185. auto_any-0.1.0/tests/test_packs.py +76 -0
  186. auto_any-0.1.0/tests/test_packs_api.py +41 -0
  187. auto_any-0.1.0/tests/test_packs_new.py +69 -0
  188. auto_any-0.1.0/tests/test_persistence_e2e.py +199 -0
  189. auto_any-0.1.0/tests/test_persistence_paths.py +133 -0
  190. auto_any-0.1.0/tests/test_policy.py +67 -0
  191. auto_any-0.1.0/tests/test_profile_attach.py +90 -0
  192. auto_any-0.1.0/tests/test_profile_clone.py +91 -0
  193. auto_any-0.1.0/tests/test_profile_selection.py +229 -0
  194. auto_any-0.1.0/tests/test_remember_mode.py +171 -0
  195. auto_any-0.1.0/tests/test_replay_executor.py +56 -0
  196. auto_any-0.1.0/tests/test_replay_service.py +58 -0
  197. auto_any-0.1.0/tests/test_run_status_done_requires_progress.py +179 -0
  198. auto_any-0.1.0/tests/test_schema.py +5 -0
  199. auto_any-0.1.0/tests/test_scorers.py +150 -0
  200. auto_any-0.1.0/tests/test_selenium_driver.py +52 -0
  201. auto_any-0.1.0/tests/test_shield.py +5 -0
  202. auto_any-0.1.0/tests/test_signature.py +8 -0
  203. auto_any-0.1.0/tests/test_site_profiles.py +135 -0
  204. auto_any-0.1.0/tests/test_step_count_receipt_alignment.py +207 -0
  205. auto_any-0.1.0/tests/test_taperoom.py +7 -0
  206. auto_any-0.1.0/tests/test_vendor_api.py +59 -0
  207. auto_any-0.1.0/tests/test_vendor_exporters.py +67 -0
  208. auto_any-0.1.0/tests/test_vendor_scorers.py +41 -0
  209. auto_any-0.1.0/tests/test_vendor_selection.py +70 -0
  210. auto_any-0.1.0/tests/test_verify_lane.py +33 -0
  211. auto_any-0.1.0/tests/test_vision_seam.py +24 -0
  212. auto_any-0.1.0/tox.ini +15 -0
@@ -0,0 +1,52 @@
1
+ AANY_SERVICE=auto-any
2
+ AANY_ENVIRONMENT=local
3
+ AANY_LOG_LEVEL=INFO
4
+ AANY_CORS_ORIGINS=["*"]
5
+ AANY_POOL_CONNECTIONS=40
6
+ AANY_POOL_KEEPALIVE=16
7
+ AANY_OUTBOUND_TIMEOUT=60.0
8
+ AANY_ENGINE_MODE=local
9
+ AANY_ENGINE_URL=
10
+ AANY_ENGINE_STREAM_READ_TIMEOUT=300.0
11
+ AANY_LLM_PROVIDER=scripted
12
+ AANY_LLM_MODEL=
13
+ AANY_LLM_API_KEY=
14
+ AANY_LLM_MAX_TOKENS=8192
15
+ AANY_LLM_TEMPERATURE=0.0
16
+ AANY_BROWSER_DRIVER=playwright
17
+ AANY_SELENIUM_BROWSER=chrome
18
+ AANY_SELENIUM_REMOTE_URL=
19
+ AANY_APPIUM_URL=http://127.0.0.1:4723
20
+ AANY_APPIUM_CAPS={}
21
+ AANY_HEADLESS=true
22
+ AANY_MAX_STEPS=40
23
+ AANY_STEP_TIMEOUT=30.0
24
+ AANY_ARTIFACTS_DIR=./artifacts
25
+ AANY_TRACE=true
26
+ AANY_VIDEO=false
27
+ AANY_LEDGER=null
28
+ AANY_SCORER=null
29
+ AANY_VISION_RESOLVER=null
30
+ AANY_CAPABILITIES=["browser","shell","http","extract","verify"]
31
+ AANY_URL_ALLOW=[]
32
+ AANY_URL_DENY=[]
33
+ AANY_RECEIPTS_DIR=./receipts
34
+ AANY_SIMILARITY_THRESHOLD=0.86
35
+ AANY_REPLAY_MODE=auto
36
+ AANY_REPAIR_LADDER=["ollama"]
37
+ AANY_SELF_HEAL=false
38
+ AANY_AUTH_DISABLED=false
39
+ AANY_JWT_AUDIENCE=urn:auto-any
40
+ AANY_JWT_PUBLIC_KEY=
41
+ AANY_TAPE_MODE=off
42
+ AANY_TAPE_DIR=./tapes
43
+ AANY_PROMPT_TOKEN_CEILING=12000
44
+ AANY_ENABLE_METRICS=false
45
+ AANY_OTLP_ENDPOINT=
46
+
47
+
48
+ # SSRF guard: block loopback/RFC1918/link-local/reserved (default off for local automation)
49
+ AANY_BLOCK_PRIVATE_NETWORKS=false
50
+ AANY_PROFILE_MODE=fresh
51
+ AANY_PROFILES_DIR=./profiles
52
+ AANY_PROFILE_MAX_AGE_H=24
@@ -0,0 +1,29 @@
1
+ name: CI
2
+
3
+ on:
4
+ push:
5
+ pull_request:
6
+
7
+ jobs:
8
+ test:
9
+ runs-on: ubuntu-latest
10
+ strategy:
11
+ matrix:
12
+ python-version: ["3.11", "3.12"]
13
+ steps:
14
+ - name: Check out repository
15
+ uses: actions/checkout@v4
16
+ - name: Set up Python
17
+ uses: actions/setup-python@v5
18
+ with:
19
+ python-version: ${{ matrix.python-version }}
20
+ cache: pip
21
+ - name: Install dependencies
22
+ run: |
23
+ python -m pip install --upgrade pip
24
+ python -m pip install -e ".[dev]"
25
+ python -m pip install "replayable-test-runner @ git+https://github.com/karthyick/replayable-test-runner.git"
26
+ - name: Run ruff
27
+ run: ruff check .
28
+ - name: Run tests
29
+ run: pytest -q
@@ -0,0 +1,14 @@
1
+ .venv/
2
+ __pycache__/
3
+ *.pyc
4
+ .env
5
+ .pytest_cache/
6
+ .ruff_cache/
7
+ artifacts/
8
+ receipts/
9
+ tapes/
10
+ uv.lock
11
+ .clinotes/
12
+ dist/
13
+ build/
14
+ *.whl
@@ -0,0 +1,171 @@
1
+ # auto_any — Architecture & Extension Map
2
+
3
+ auto_any is a **general task engine**: an NL goal (or a typed pack call) is
4
+ planned, executed across capabilities, and distilled into a **receipt** that
5
+ replays deterministically (zero-LLM) and self-heals on drift. It is a
6
+ **platform** — every future use-case (LinkedIn apply, app testing, scraping,
7
+ any UI automation) plugs in without editing the core.
8
+
9
+ ## Layers
10
+
11
+ ```
12
+ HTTP /automations (free-text) /packs (typed) /receipts /profiles
13
+ │ │
14
+ ▼ ▼
15
+ CONDUCTOR prompt budget · output shield · tapes (record/replay SSE)
16
+ │
17
+ ▼ AgentEngine PORT ───────────── local think⇄act loop | remote AI service
18
+ │
19
+ ▼ CapabilityGate ── routes Action{capability,kind,params} → RTR StepRunner
20
+ browser · shell · http · extract · verify (+ yours)
21
+ │
22
+ ├─ browsergate: BrowserDriver PORT (Playwright) · ProfilePort (clone/named)
23
+ │ UrlPolicy (SSRF-hardened) · VisionResolver PORT (seam)
24
+ │
25
+ ▼ replay: Signature → Receipt (RTR) → deterministic replay + repair ladder
26
+ → RunLedger (JSONL by default | null | kr_hub | sqlite)
27
+ ```
28
+
29
+ ## Storage layout
30
+
31
+ `AANY_HOME` defaults to `Path.home() / ".auto_any"` and can be overridden with
32
+ the `AANY_HOME` environment variable. Relative receipts and artifacts paths
33
+ resolve under this root; absolute paths are honored verbatim.
34
+
35
+ ```
36
+ ~/.auto_any/
37
+ ├── receipts/ # replayable receipts
38
+ ├── artifacts/<run_id>/ # per-run screenshots + trace.zip
39
+ │ ├── 001_open.png
40
+ │ ├── 002_click.png
41
+ │ └── trace.zip
42
+ ├── ledger.jsonl # append-only run history
43
+ └── site_profiles.json # site → browser-profile mapping
44
+ ```
45
+
46
+ The Playwright driver automatically captures ordered screenshots after browser
47
+ actions (`auto_screenshot=True` by default). Capture failures are logged and do
48
+ not fail the run. An omitted run ID uses a timestamped artifact directory.
49
+
50
+ ## The four PORTS (swap an implementation, don't fork)
51
+
52
+ | Port | File | Ships | Add another by |
53
+ |---|---|---|---|
54
+ | **AgentEngine** | `engine/base.py` | `LocalEngine`, `RemoteEngine` | new class + branch in `engine/__init__.engine_for` |
55
+ | **Capability** | `capabilities/base.py` | browser, shell, http, extract, verify | register a StepRunner + a `Capability` |
56
+ | **BrowserDriver / ProfilePort** | `browsergate/base.py`, `profile_port.py` | Playwright + Clone/Named profile | new driver/port class + registry entry |
57
+ | **VisionResolver** | `browsergate/vision_port.py` | `NullVisionResolver` (DOM-only) | implement `resolve()` for a vision model |
58
+
59
+ ## The three REGISTRIES
60
+
61
+ | Registry | Holds | Populated by |
62
+ |---|---|---|
63
+ | `capabilities.CAPABILITIES` / RTR `STEP_RUNNERS` | execution surfaces | import-time registration |
64
+ | `packs.registry.PLAYBOOKS` | reusable automations | `packs.builtin` + entry points |
65
+
66
+ ## Packs - the extension surface for "support all types"
67
+
68
+ ## Vendor selection
69
+
70
+ ```
71
+ request.driver > PlaybookSpec.driver > AANY_BROWSER_DRIVER > task hint
72
+ override pack hint config nature default
73
+ ```
74
+
75
+ | Selection tier | Source | Example |
76
+ |---|---|---|
77
+ | Request override | `AutomationRequest.driver` | `{"driver":"selenium"}` |
78
+ | Playbook hint | `PlaybookSpec.driver` | `driver="appium"` |
79
+ | Config | `AANY_BROWSER_DRIVER` | `playwright`, `selenium`, `appium`, `auto` |
80
+ | Task nature default | `context["task_hint"]` when config is `auto` | `mobile` → Appium; `cross_browser`/`legacy` → Selenium; otherwise Playwright |
81
+
82
+ `AANY_BROWSER_DRIVER=auto` enables task-nature selection. A non-`auto` config
83
+ value wins over task hints.
84
+
85
+ ### Vendor taxonomy
86
+
87
+ | Role | Built-ins | Seam |
88
+ |---|---|---|
89
+ | Per-action UI control | Playwright, Selenium, Appium | `BrowserDriver` + `DRIVERS` |
90
+ | Whole-spec runners | Cypress, Playwright, Robot | `SpecExporter` + `EXPORTERS` |
91
+ | Finished-run scoring | Null, passmark | `Scorer` + `SCORERS` |
92
+
93
+ Add a vendor by implementing the relevant contract and publishing its class in
94
+ the matching entry-point group. Vendor plugins load during startup; broken
95
+ third-party entry points are logged and skipped.
96
+
97
+ ```toml
98
+ [project.entry-points."auto_any.drivers"]
99
+ my_driver = "my_auto_any_vendor:MyDriver"
100
+
101
+ [project.entry-points."auto_any.exporters"]
102
+ my_exporter = "my_auto_any_vendor:MyExporter"
103
+
104
+ [project.entry-points."auto_any.scorers"]
105
+ my_scorer = "my_auto_any_vendor:MyScorer"
106
+ ```
107
+
108
+ The targets are classes implementing `BrowserDriver`, `SpecExporter`, or
109
+ `Scorer`; the entry-point name becomes the registry name. Install the
110
+ distribution, restart auto-any, and select it through the applicable request,
111
+ pack, or config setting.
112
+
113
+ | Role | Runtime selection |
114
+ |---|---|
115
+ | Driver | precedence diagram above; `AutomationRequest.driver`, `PlaybookSpec.driver`, or `AANY_BROWSER_DRIVER` |
116
+ | Exporter | `POST /receipts/{key}/export` with `{"format":"<name>"}` |
117
+ | Scorer | `AutomationRequest.score` or `AANY_SCORER` |
118
+
119
+ A **Playbook** declares a named, parameterized automation. It only builds a
120
+ `PlaybookRun` (goal + url + capabilities + profile + mode); the standard engine
121
+ executes it, so every pack inherits receipts, replay, self-heal,
122
+ human-in-the-loop, artifacts, ledger, and login-profile reuse for free.
123
+
124
+ ```
125
+ packs/
126
+ ├─ base.py Playbook (ABC) + PlaybookSpec (typed params) + PlaybookRun
127
+ ├─ registry.py PLAYBOOKS: register / get / names
128
+ ├─ discovery.py load_builtin() + load_plugins() (entry-point group auto_any.packs)
129
+ ├─ compile.py to_request(pack, params) -> AutomationRequest
130
+ └─ builtin/ smoke_test · form_fill · linkedin_apply (thin reference packs)
131
+ ```
132
+
133
+ ### How to add a use-case (no core edits)
134
+
135
+ ```python
136
+ from auto_any import Playbook, PlaybookSpec, PlaybookRun, register
137
+
138
+ class MyPack(Playbook):
139
+ spec = PlaybookSpec(
140
+ name="my_task",
141
+ params_schema={"url": {"type": "string", "required": True}},
142
+ capabilities=["browser", "verify"],
143
+ profile_name="my_site", # reuse a logged-in profile
144
+ )
145
+ def build(self, p):
146
+ return PlaybookRun(goal=f"do X on {p['url']}", url=p["url"],
147
+ capabilities=self.spec.capabilities)
148
+
149
+ register(MyPack()) # in-process
150
+ ```
151
+
152
+ Ship it in a separate distribution instead, and it auto-registers:
153
+
154
+ ```toml
155
+ [project.entry-points."auto_any.packs"]
156
+ my_task = "my_pkg:register" # a zero-arg hook, a Playbook subclass, or an instance
157
+ ```
158
+
159
+ Then `GET /packs` lists it with its schema and `POST /packs/my_task/run` executes it.
160
+
161
+ ## "How do I add X?" quick table
162
+
163
+ | Want | Do this | Touches core? |
164
+ |---|---|---|
165
+ | A new use-case (LinkedIn, testing, scrape) | write a **Playbook** pack | no |
166
+ | A new execution surface (e.g. `filesystem`) | register a **StepRunner** + `Capability` | no |
167
+ | A different browser (CDP grid, Edge) | new **BrowserDriver** + registry entry | no |
168
+ | Reuse an existing agent runner | `AANY_ENGINE_MODE=remote` (RemoteEngine) | no |
169
+ | Vision fallback | implement **VisionResolver** | no |
170
+ | Record run history | JSONL at `<AANY_HOME>/ledger.jsonl` by default; `AANY_LEDGER=null` opts out, `AANY_LEDGER=kr_hub` selects kr_hub, and `<path>.db` keeps SQLite | no |
171
+ ```
@@ -0,0 +1,289 @@
1
+ Metadata-Version: 2.4
2
+ Name: auto-any
3
+ Version: 0.1.0
4
+ Summary: Deterministic browser and task automation with replayable receipts
5
+ Requires-Python: >=3.11
6
+ Requires-Dist: anyio>=4.0
7
+ Requires-Dist: fastapi>=0.115
8
+ Requires-Dist: httpx>=0.27
9
+ Requires-Dist: pydantic-settings>=2.4
10
+ Requires-Dist: pydantic>=2.8
11
+ Requires-Dist: pyjwt>=2.8
12
+ Requires-Dist: python-json-logger>=2.0
13
+ Requires-Dist: rich>=13.7
14
+ Requires-Dist: sse-starlette>=2.1
15
+ Requires-Dist: typer>=0.12
16
+ Requires-Dist: uvicorn[standard]>=0.30
17
+ Provides-Extra: anthropic
18
+ Requires-Dist: anthropic>=0.116; extra == 'anthropic'
19
+ Provides-Extra: appium
20
+ Requires-Dist: appium-python-client>=4.0; extra == 'appium'
21
+ Provides-Extra: bedrock
22
+ Requires-Dist: boto3>=1.35; extra == 'bedrock'
23
+ Provides-Extra: browser
24
+ Requires-Dist: playwright>=1.45; extra == 'browser'
25
+ Provides-Extra: dev
26
+ Requires-Dist: asgi-lifespan; extra == 'dev'
27
+ Requires-Dist: httpx; extra == 'dev'
28
+ Requires-Dist: pytest-asyncio>=0.23; extra == 'dev'
29
+ Requires-Dist: pytest>=8; extra == 'dev'
30
+ Requires-Dist: respx>=0.21; extra == 'dev'
31
+ Requires-Dist: ruff>=0.6; extra == 'dev'
32
+ Provides-Extra: efa
33
+ Requires-Dist: rubricon>=0.2.2; extra == 'efa'
34
+ Provides-Extra: mcp
35
+ Requires-Dist: mcp>=1.0; extra == 'mcp'
36
+ Provides-Extra: openai
37
+ Requires-Dist: openai>=1.40; extra == 'openai'
38
+ Provides-Extra: selenium
39
+ Requires-Dist: selenium>=4.20; extra == 'selenium'
40
+ Description-Content-Type: text/markdown
41
+
42
+ # auto-any
43
+
44
+ ```
45
+ goal + URL -> signature -> exact receipt? -> deterministic replay
46
+ \-> local/remote engine -> browser -> distiller -> receipt
47
+ ```
48
+
49
+ Run with `uvicorn auto_any.main:app`. For local development install RTR first with
50
+ `uv pip install -e C:/Users/KR-ultra/Source/code_base/repos/replayable-test-runner`.
51
+ Configuration uses the `AANY_` prefix; see `.env.example`.
52
+
53
+ ## Install / share
54
+
55
+ One command gets a teammate from zero to "auto-any MCP Connected":
56
+
57
+ ```powershell
58
+ pwsh install/install.ps1
59
+ ```
60
+
61
+ macOS, Linux, or Git Bash:
62
+
63
+ ```bash
64
+ bash install/install.sh
65
+ ```
66
+
67
+ See [`install/README.md`](install/README.md) for options (custom venv path, local wheel) and the two `.env` knobs you must set.
68
+
69
+ ## Reuse my Chrome logins
70
+
71
+ To run against a cloned copy of an already-authenticated Chrome profile:
72
+
73
+ ```powershell
74
+ $env:AANY_PROFILE_MODE = "clone"
75
+ $env:AANY_CHROME_SOURCE_PROFILE = "Profile 2" # or Default
76
+ uvicorn auto_any.main:app
77
+ ```
78
+
79
+ The clone is created under `AANY_PROFILES_DIR`, and the source User Data tree is
80
+ never written. Chrome may be left open; a cookie written very recently may not
81
+ have been flushed yet. `AANY_PROFILE_KEEP=false` removes the clone after the run.
82
+ Cloning copies live authentication cookies, so treat the profiles directory as
83
+ sensitive and do not commit or share it. Set `AANY_BLOCK_PRIVATE_NETWORKS=true`
84
+ for hosted deployments; the default is false so operator-owned localhost and
85
+ LAN targets remain usable. For a persistent named profile, use
86
+ `AANY_PROFILE_MODE=named` and select it with `profile: {"name":"work"}` (or use
87
+ a pack that declares `profile_name`, such as `linkedin_apply`). Named profiles
88
+ are managed with `GET /profiles` and `DELETE /profiles/{name}`.
89
+
90
+ ## Capabilities
91
+
92
+ | Capability | Action kinds | Runner |
93
+ |---|---|---|
94
+ | browser | `playwright` | RTR Playwright meta-runner |
95
+ | shell | `shell` | RTR shell runner (allow/deny policy) |
96
+ | http | `http_request` | RTR HTTP runner |
97
+ | extract | `extract` | typed Pydantic model |
98
+ | verify | `verify` | explicit RTR assertion |
99
+
100
+ Add a capability by implementing the `Capability` contract, registering its
101
+ name in `CAPABILITIES`, and registering an RTR `StepRunner` for each action
102
+ kind. `CapabilityGate` then dispatches it without changing the engine loop.
103
+
104
+ ### Rubricon / EFA scoring
105
+
106
+ Install the optional evaluator integration with `pip install 'auto-any[efa]'`
107
+ and select it with `score="efa"` in an automation request. The adapter uses
108
+ Rubricon's dynamic criteria generation and per-criterion evaluator to attach
109
+ RAS, APR, and criterion scores to the run result. It evaluates the completed
110
+ response only; replay remains deterministic and does not invoke an LLM.
111
+
112
+ Paused OTP/CAPTCHA/confirmation turns emit `needs_input`; resume with
113
+ `POST /automations/{run_id}/resume`. Secret values are stored in receipts only
114
+ as `{"$var":"field"}`. Trace artifacts are per-run and video is opt-in.
115
+
116
+ ## Storage layout
117
+
118
+ The storage root is `AANY_HOME`, defaulting to `Path.home() / ".auto_any"`.
119
+ Relative `AANY_RECEIPTS_DIR` and `AANY_ARTIFACTS_DIR` values resolve under this
120
+ root; absolute values are honored as supplied.
121
+
122
+ ```
123
+ ~/.auto_any/
124
+ ├── receipts/ # replayable receipts
125
+ ├── artifacts/<run_id>/ # per-run screenshots + trace.zip
126
+ │ ├── 001_open.png
127
+ │ ├── 002_click.png
128
+ │ └── trace.zip
129
+ ├── ledger.jsonl # append-only run history
130
+ └── site_profiles.json # site → browser-profile mapping
131
+ ```
132
+
133
+ Browser screenshots are captured automatically after browser actions, with
134
+ zero-padded ordered names such as `001_open.png` and `002_click.png`. Capture
135
+ is non-fatal and is enabled by the `auto_screenshot=True` driver constructor
136
+ option. A run without an explicit ID uses a timestamped artifact directory.
137
+ The run ledger is enabled by default as an append-only JSONL file; pass
138
+ `AANY_LEDGER=null` to opt out, `AANY_LEDGER=kr_hub` for the remote ledger, or
139
+ `AANY_LEDGER=<path>.db` to keep using the previous SQLite backend.
140
+
141
+ > **Economics:** replayed runs cost **0 LLM tokens**; planned runs cost **N LLM tokens**.
142
+
143
+ ## MCP server
144
+
145
+ auto_any can run as a STDIO Model Context Protocol (MCP) server so Claude and
146
+ other MCP clients can invoke it in-process.
147
+
148
+ ```powershell
149
+ # install the MCP extra
150
+ uv pip install -e ".[mcp,browser]"
151
+
152
+ # register with Claude Desktop (do not edit ~/.claude manually)
153
+ claude mcp add-json auto-any '{"command": "auto-any-mcp"}'
154
+ ```
155
+
156
+ The server boots one shared headless runtime (`AANY_PROFILE_MODE=fresh`) and
157
+ exposes five tools:
158
+
159
+ | Tool | Purpose |
160
+ |---|---|
161
+ | `auto_any_list_packs` | List installed packs + parameter schemas |
162
+ | `auto_any_run_pack` | Run a pack by name with validated params |
163
+ | `auto_any_run` | Run a free-text goal + URL |
164
+ | `auto_any_list_receipts` | List saved receipts (signature key + title) |
165
+ | `auto_any_replay` | Replay a receipt with zero LLM tokens |
166
+
167
+ Start it manually with `auto-any-mcp`.
168
+
169
+ ## Three Lanes
170
+
171
+ auto_any exposes the same automation engine through three interfaces. All three delegate to the shared `automation_events` generator in `api/automations.py`.
172
+
173
+ | Lane | Command | Description |
174
+ |------|---------|-------------|
175
+ | **CLI** | `auto-any run --goal "..." --url "..."` | Direct command-line execution |
176
+ | **HTTP** | `auto-any serve` (starts uvicorn) | REST API at `POST /automations` |
177
+ | **MCP** | `python -m auto_any.mcp_server` | STDIO tools for Claude Desktop |
178
+
179
+ All lanes accept the same request shape: `goal`, `url`, `mode` (auto/always_plan/replay_only), and return streaming events containing `run_started`, `progress`, `run_finished`, and optionally `receipt_saved`.
180
+
181
+ Additional CLI commands:
182
+ - `auto-any receipts` — list stored receipts (signature key + title)
183
+ - `auto-any replay <signature_key>` — replay a receipt with zero LLM tokens
184
+ - `auto-any version` — print the running build's identity
185
+
186
+ ## Receipt Contract
187
+
188
+ A **receipt** is a deterministic, replayable record of an automation run. It contains:
189
+
190
+ - **signature_key**: SHA256 hash of `goal + url + context` — the identity for exact-match replay
191
+ - **title**: Human-readable description
192
+ - **steps**: Ordered list of actions (goto, click, type, extract, verify...)
193
+ - **url**: Target URL at time of execution
194
+
195
+ **What replay guarantees:**
196
+ - Same goal + same URL → same actions in same order
197
+ - Zero LLM tokens consumed (deterministic playback)
198
+ - Pass/fail per step — if a page element changed, replay reports failure
199
+
200
+ **What replay does NOT guarantee:**
201
+ - External state (logged-in sessions, database data, third-party API responses)
202
+ - Timing/performance — replay is deterministic but not real-time
203
+ - Cross-browser parity — receipts are driver-specific (Playwright vs Selenium)
204
+ - Future site changes — replay fails if UI changed; this is the intended behavior
205
+
206
+ The signature key is the lookup key for both `auto-any receipts` and the MCP `auto_any_replay` tool.
207
+
208
+ ## How to Tell Which Build is Running
209
+
210
+ On 2026-07-26, the MCP server ran for hours importing stale code from a different git worktree while reporting healthy. Nothing revealed which build was actually executing.
211
+
212
+ Two ways to verify the running build:
213
+
214
+ ```bash
215
+ # CLI — prints version, git SHA, dirty flag, and import path
216
+ $ auto-any version
217
+ auto-any 0.1.0 git:bf3545d
218
+
219
+ $ auto-any version --json
220
+ {"version": "0.1.0", "git_sha": "bf3545d", "dirty": false, "import_path": "C:\\Users\\...\\src\\auto_any"}
221
+ ```
222
+
223
+ ```bash
224
+ # MCP — check the startup log line
225
+ $ python -m auto_any.mcp_server
226
+ selected browser driver=playwright tier=config_explicit task_hint=
227
+ MCP server starting: auto-any 0.1.0 git:bf3545d; import_path=C:\Users\...\src\auto_any
228
+ ```
229
+
230
+ The `import_path` field shows the exact directory from which `auto_any` was loaded. If this differs from the expected worktree, the running code is stale.
231
+
232
+ ## Quickstart — adopt it for any UI automation
233
+
234
+ ```bash
235
+ uv venv && uv pip install -e ".[dev,browser]"
236
+ uv run playwright install chromium # real browser lane
237
+
238
+ # list built-in automation packs (smoke_test, form_fill, linkedin_apply)
239
+ uv run auto-any packs
240
+
241
+ # run the API
242
+ uv run auto-any serve --port 8080
243
+ # POST /automations free-text goal (the raw engine)
244
+ # GET /packs installed packs + their param schemas
245
+ # POST /packs/{name}/run typed params -> same pipeline -> receipt
246
+
247
+ # reuse your real Chrome logins (no interactive auth):
248
+ # AANY_PROFILE_MODE=clone AANY_CHROME_SOURCE_PROFILE="Profile 2"
249
+ # real LLM to drive arbitrary UIs:
250
+ # AANY_LLM_PROVIDER=anthropic AANY_LLM_API_KEY=...
251
+ ```
252
+
253
+ ### Packs quickstart
254
+
255
+ The built-in packs are `smoke_test`, `form_fill`, and `linkedin_apply`. For a
256
+ local unauthenticated development server, set `AANY_AUTH_DISABLED=true`, then
257
+ inspect schemas and run `smoke_test` as an SSE stream:
258
+
259
+ ```bash
260
+ export AANY_AUTH_DISABLED=true
261
+ uv run auto-any serve --port 8080
262
+
263
+ curl http://127.0.0.1:8080/packs
264
+ curl -N -X POST http://127.0.0.1:8080/packs/smoke_test/run \
265
+ -H 'Content-Type: application/json' \
266
+ -d '{"base_url":"http://127.0.0.1:8080","checks":[{"path":"/health","expect_text":"ok"}]}'
267
+ ```
268
+
269
+ With authentication enabled, add `-H "Authorization: Bearer <token>"` to both
270
+ requests.
271
+
272
+ Add a new use-case as a **pack** (no core edits) - see `ARCHITECTURE.md`.
273
+
274
+ ## Multi-vendor execution
275
+
276
+ `AANY_BROWSER_DRIVER=playwright|selenium|appium|auto` selects the per-action driver.
277
+ Install optional integrations with `uv pip install -e ".[selenium]"` or `.[appium]`;
278
+ absent SDKs and servers produce an actionable unavailable event. A request can
279
+ override this with `{"driver":"selenium"}`. Export receipts with
280
+ `POST /receipts/{key}/export` and `{"format":"cypress"}` (`playwright` and
281
+ `robot` are also supported). `$var` values become vendor environment lookups and
282
+ are never written as secrets. Set `score` to `passmark` for a tolerant score report.
283
+
284
+ ### Tests
285
+
286
+ ```bash
287
+ uv run pytest -q # hermetic unit suite (fakes only, no browser)
288
+ AANY_E2E=1 uv run pytest tests/test_e2e_real_browser.py # real headless Chromium
289
+ ```