agentinputlab 0.1.0a1__py3-none-any.whl

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 (469) hide show
  1. agentinputlab/__init__.py +28 -0
  2. agentinputlab/algorithms/__init__.py +6 -0
  3. agentinputlab/algorithms/acquire/__init__.py +145 -0
  4. agentinputlab/algorithms/acquire/adaptive_rag/__init__.py +43 -0
  5. agentinputlab/algorithms/acquire/adaptive_rag/replay.py +367 -0
  6. agentinputlab/algorithms/acquire/adaptive_rag/runtime.py +1202 -0
  7. agentinputlab/algorithms/acquire/adaptive_rag_common_support.py +542 -0
  8. agentinputlab/algorithms/acquire/hyde.py +296 -0
  9. agentinputlab/algorithms/acquire/hyde_controlled.py +200 -0
  10. agentinputlab/algorithms/acquire/hyde_external.py +148 -0
  11. agentinputlab/algorithms/acquire/ircot.py +28 -0
  12. agentinputlab/algorithms/acquire/ircot_common_support.py +478 -0
  13. agentinputlab/algorithms/acquire/ircot_native.py +355 -0
  14. agentinputlab/algorithms/acquire/ircot_replay.py +327 -0
  15. agentinputlab/algorithms/acquire/qrag/__init__.py +79 -0
  16. agentinputlab/algorithms/acquire/qrag/bridge.py +828 -0
  17. agentinputlab/algorithms/acquire/qrag/common_support.py +891 -0
  18. agentinputlab/algorithms/acquire/qrag/native_axis.py +71 -0
  19. agentinputlab/algorithms/acquire/qrag/reader_adapter.py +139 -0
  20. agentinputlab/algorithms/acquire/r2ag/__init__.py +57 -0
  21. agentinputlab/algorithms/acquire/r2ag/bridge.py +1312 -0
  22. agentinputlab/algorithms/acquire/r2ag/common_support.py +1029 -0
  23. agentinputlab/algorithms/acquire/r2ag/socket_runtime.py +255 -0
  24. agentinputlab/algorithms/clarify/__init__.py +61 -0
  25. agentinputlab/algorithms/clarify/_structured.py +64 -0
  26. agentinputlab/algorithms/clarify/active_task_disambiguation.py +371 -0
  27. agentinputlab/algorithms/clarify/ask_or_assume/__init__.py +17 -0
  28. agentinputlab/algorithms/clarify/ask_or_assume/monitor.py +410 -0
  29. agentinputlab/algorithms/clarify/future_turns.py +379 -0
  30. agentinputlab/algorithms/clarify/udcs.py +255 -0
  31. agentinputlab/algorithms/clarify/udcs_deterministic.py +196 -0
  32. agentinputlab/algorithms/curate/__init__.py +65 -0
  33. agentinputlab/algorithms/curate/acon/__init__.py +27 -0
  34. agentinputlab/algorithms/curate/acon/common.py +284 -0
  35. agentinputlab/algorithms/curate/acon/common_support.py +136 -0
  36. agentinputlab/algorithms/curate/acon/history_bridge.py +309 -0
  37. agentinputlab/algorithms/curate/acon/observation_bridge.py +290 -0
  38. agentinputlab/algorithms/curate/llmlingua/__init__.py +30 -0
  39. agentinputlab/algorithms/curate/llmlingua/adapters.py +537 -0
  40. agentinputlab/algorithms/curate/llmlingua/budget.py +48 -0
  41. agentinputlab/algorithms/curate/llmlingua/controlled.py +175 -0
  42. agentinputlab/algorithms/curate/llmlingua/multilingual.py +105 -0
  43. agentinputlab/algorithms/curate/llmlingua/span_map.py +167 -0
  44. agentinputlab/algorithms/curate/recomp/__init__.py +47 -0
  45. agentinputlab/algorithms/curate/recomp/external.py +1129 -0
  46. agentinputlab/algorithms/curate/recomp/selective.py +304 -0
  47. agentinputlab/algorithms/decompose/__init__.py +107 -0
  48. agentinputlab/algorithms/decompose/decomp/__init__.py +53 -0
  49. agentinputlab/algorithms/decompose/decomp/common_support.py +478 -0
  50. agentinputlab/algorithms/decompose/decomp/planner_component.py +199 -0
  51. agentinputlab/algorithms/decompose/decomp/worker_bridge.py +416 -0
  52. agentinputlab/algorithms/decompose/llm_compiler/__init__.py +35 -0
  53. agentinputlab/algorithms/decompose/llm_compiler/bridge.py +520 -0
  54. agentinputlab/algorithms/decompose/llm_compiler/common_support.py +504 -0
  55. agentinputlab/algorithms/decompose/plan_and_budget/__init__.py +43 -0
  56. agentinputlab/algorithms/decompose/plan_and_budget/cached_soft_budget.py +362 -0
  57. agentinputlab/algorithms/decompose/plan_and_budget/decomposer_component.py +107 -0
  58. agentinputlab/algorithms/decompose/plan_and_budget/full.py +205 -0
  59. agentinputlab/algorithms/decompose/plan_and_budget/hard_cap.py +246 -0
  60. agentinputlab/algorithms/decompose/plan_and_budget/regenerated_soft_budget.py +448 -0
  61. agentinputlab/algorithms/decompose/plan_and_solve/__init__.py +33 -0
  62. agentinputlab/algorithms/decompose/plan_and_solve/plan_component.py +192 -0
  63. agentinputlab/algorithms/decompose/plan_and_solve/ps_prompt.py +127 -0
  64. agentinputlab/algorithms/decompose/self_ask/__init__.py +27 -0
  65. agentinputlab/algorithms/decompose/self_ask/loop.py +362 -0
  66. agentinputlab/algorithms/decompose/self_ask/planner_component.py +117 -0
  67. agentinputlab/algorithms/reformulate/__init__.py +85 -0
  68. agentinputlab/algorithms/reformulate/infocqr/__init__.py +40 -0
  69. agentinputlab/algorithms/reformulate/infocqr/cache_replay.py +362 -0
  70. agentinputlab/algorithms/reformulate/infocqr/external.py +675 -0
  71. agentinputlab/algorithms/reformulate/infocqr/student.py +189 -0
  72. agentinputlab/algorithms/reformulate/query2doc/__init__.py +42 -0
  73. agentinputlab/algorithms/reformulate/query2doc/controlled_live.py +190 -0
  74. agentinputlab/algorithms/reformulate/query2doc/models.py +95 -0
  75. agentinputlab/algorithms/reformulate/query2doc/official_cached.py +335 -0
  76. agentinputlab/algorithms/reformulate/query2doc/querygym_bridge.py +223 -0
  77. agentinputlab/algorithms/reformulate/query2doc/querygym_runtime.py +326 -0
  78. agentinputlab/algorithms/reformulate/reformer/__init__.py +23 -0
  79. agentinputlab/algorithms/reformulate/reformer/external.py +417 -0
  80. agentinputlab/algorithms/reformulate/rrr.py +696 -0
  81. agentinputlab/annotation/__init__.py +5 -0
  82. agentinputlab/annotation/independent.py +428 -0
  83. agentinputlab/annotation/server.py +233 -0
  84. agentinputlab/annotation/static/app.js +121 -0
  85. agentinputlab/annotation/static/index.html +60 -0
  86. agentinputlab/annotation/static/review.css +2 -0
  87. agentinputlab/annotation/static/review.html +10 -0
  88. agentinputlab/annotation/static/review.js +11 -0
  89. agentinputlab/annotation/static/styles.css +55 -0
  90. agentinputlab/annotation/store.py +444 -0
  91. agentinputlab/api.py +368 -0
  92. agentinputlab/artifacts/__init__.py +15 -0
  93. agentinputlab/artifacts/json_file.py +104 -0
  94. agentinputlab/artifacts/manifest.py +98 -0
  95. agentinputlab/artifacts/store.py +157 -0
  96. agentinputlab/backends/__init__.py +98 -0
  97. agentinputlab/backends/embedding/__init__.py +25 -0
  98. agentinputlab/backends/embedding/base.py +145 -0
  99. agentinputlab/backends/embedding/callable.py +75 -0
  100. agentinputlab/backends/embedding/replay.py +129 -0
  101. agentinputlab/backends/llm/__init__.py +31 -0
  102. agentinputlab/backends/llm/base.py +149 -0
  103. agentinputlab/backends/llm/callable.py +108 -0
  104. agentinputlab/backends/llm/decomp_completion.py +88 -0
  105. agentinputlab/backends/llm/final_content_stop.py +72 -0
  106. agentinputlab/backends/llm/gpt_oss_final_channel.py +84 -0
  107. agentinputlab/backends/llm/openai_compatible.py +332 -0
  108. agentinputlab/backends/llm/transformers.py +171 -0
  109. agentinputlab/backends/retrieval/__init__.py +35 -0
  110. agentinputlab/backends/retrieval/base.py +50 -0
  111. agentinputlab/backends/retrieval/dense.py +777 -0
  112. agentinputlab/backends/retrieval/memory.py +119 -0
  113. agentinputlab/backends/retrieval/sqlite_fts.py +155 -0
  114. agentinputlab/backends/user/__init__.py +15 -0
  115. agentinputlab/backends/user/base.py +66 -0
  116. agentinputlab/backends/user/callable.py +54 -0
  117. agentinputlab/backends/user/replay.py +89 -0
  118. agentinputlab/backends/vector.py +38 -0
  119. agentinputlab/benchmarks/__init__.py +246 -0
  120. agentinputlab/benchmarks/acquisition.py +106 -0
  121. agentinputlab/benchmarks/ale.py +540 -0
  122. agentinputlab/benchmarks/appworld_ul.py +669 -0
  123. agentinputlab/benchmarks/assets.py +226 -0
  124. agentinputlab/benchmarks/base.py +68 -0
  125. agentinputlab/benchmarks/beir_archive.py +235 -0
  126. agentinputlab/benchmarks/beir_sample.py +269 -0
  127. agentinputlab/benchmarks/bfcl_native.py +245 -0
  128. agentinputlab/benchmarks/bfcl_v3.py +607 -0
  129. agentinputlab/benchmarks/cache.py +218 -0
  130. agentinputlab/benchmarks/clariq.py +203 -0
  131. agentinputlab/benchmarks/data/registry.json +321 -0
  132. agentinputlab/benchmarks/data/smoke_assets.json +511 -0
  133. agentinputlab/benchmarks/data/statebench_v0_8_1_task_ids.json +475 -0
  134. agentinputlab/benchmarks/data/toolsandbox_165848b_scenario_names.json +1 -0
  135. agentinputlab/benchmarks/execution.py +1111 -0
  136. agentinputlab/benchmarks/interactive.py +257 -0
  137. agentinputlab/benchmarks/ir_query.py +204 -0
  138. agentinputlab/benchmarks/longbench.py +203 -0
  139. agentinputlab/benchmarks/paired_replay.py +479 -0
  140. agentinputlab/benchmarks/preprocess.py +2313 -0
  141. agentinputlab/benchmarks/production.py +236 -0
  142. agentinputlab/benchmarks/qa.py +963 -0
  143. agentinputlab/benchmarks/registry.py +110 -0
  144. agentinputlab/benchmarks/retrieval.py +459 -0
  145. agentinputlab/benchmarks/runtime.py +509 -0
  146. agentinputlab/benchmarks/statebench.py +938 -0
  147. agentinputlab/benchmarks/tau3.py +725 -0
  148. agentinputlab/benchmarks/toolsandbox.py +626 -0
  149. agentinputlab/benchmarks/trec_dl_sample.py +263 -0
  150. agentinputlab/campaign.py +654 -0
  151. agentinputlab/campaign_runner.py +953 -0
  152. agentinputlab/cli/__init__.py +6 -0
  153. agentinputlab/cli/main.py +933 -0
  154. agentinputlab/connectors/__init__.py +46 -0
  155. agentinputlab/connectors/knowledge/__init__.py +4 -0
  156. agentinputlab/connectors/tools/__init__.py +30 -0
  157. agentinputlab/connectors/tools/base.py +57 -0
  158. agentinputlab/connectors/tools/mcp.py +222 -0
  159. agentinputlab/connectors/tools/memory.py +41 -0
  160. agentinputlab/connectors/tools/openapi.py +695 -0
  161. agentinputlab/connectors/tools/schema.py +102 -0
  162. agentinputlab/contracts/__init__.py +91 -0
  163. agentinputlab/contracts/availability.py +24 -0
  164. agentinputlab/contracts/dependency.py +58 -0
  165. agentinputlab/contracts/enums.py +82 -0
  166. agentinputlab/contracts/errors.py +63 -0
  167. agentinputlab/contracts/evidence.py +43 -0
  168. agentinputlab/contracts/intent.py +110 -0
  169. agentinputlab/contracts/lifecycle.py +164 -0
  170. agentinputlab/contracts/patch.py +167 -0
  171. agentinputlab/contracts/serialization.py +93 -0
  172. agentinputlab/contracts/state.py +127 -0
  173. agentinputlab/contracts/trace.py +59 -0
  174. agentinputlab/contracts/worker.py +42 -0
  175. agentinputlab/core/__init__.py +7 -0
  176. agentinputlab/data/__init__.py +32 -0
  177. agentinputlab/data/checksums.py +9 -0
  178. agentinputlab/data/downloader.py +235 -0
  179. agentinputlab/data/http.py +171 -0
  180. agentinputlab/data/index.py +21 -0
  181. agentinputlab/data/overlap.py +28 -0
  182. agentinputlab/data/preprocess.py +32 -0
  183. agentinputlab/data/storage.py +57 -0
  184. agentinputlab/data/trec_sample.py +427 -0
  185. agentinputlab/evaluation/__init__.py +5 -0
  186. agentinputlab/evaluation/outcomes.py +32 -0
  187. agentinputlab/experiments/__init__.py +278 -0
  188. agentinputlab/experiments/analysis.py +981 -0
  189. agentinputlab/experiments/artifacts.py +687 -0
  190. agentinputlab/experiments/automatic_measurement.py +678 -0
  191. agentinputlab/experiments/auxiliary_matrix.py +203 -0
  192. agentinputlab/experiments/bindings.py +430 -0
  193. agentinputlab/experiments/blueprints.py +147 -0
  194. agentinputlab/experiments/builders/__init__.py +45 -0
  195. agentinputlab/experiments/builders/executable_external_payload.py +423 -0
  196. agentinputlab/experiments/builders/external_process_arena.py +1018 -0
  197. agentinputlab/experiments/builders/registry_intervention.py +463 -0
  198. agentinputlab/experiments/cli.py +158 -0
  199. agentinputlab/experiments/controls.py +340 -0
  200. agentinputlab/experiments/driver_guard.py +205 -0
  201. agentinputlab/experiments/drivers/__init__.py +89 -0
  202. agentinputlab/experiments/drivers/external_arena.py +2908 -0
  203. agentinputlab/experiments/drivers/method_process.py +662 -0
  204. agentinputlab/experiments/drivers/parity_process.py +1484 -0
  205. agentinputlab/experiments/drivers/static_production.py +1574 -0
  206. agentinputlab/experiments/executor.py +448 -0
  207. agentinputlab/experiments/loader.py +605 -0
  208. agentinputlab/experiments/measurement.py +675 -0
  209. agentinputlab/experiments/paper_analysis.py +1818 -0
  210. agentinputlab/experiments/parity_identity.py +605 -0
  211. agentinputlab/experiments/plan.py +277 -0
  212. agentinputlab/experiments/production_builders.py +3739 -0
  213. agentinputlab/experiments/readiness.py +432 -0
  214. agentinputlab/experiments/schema.py +514 -0
  215. agentinputlab/experiments/secrets.py +177 -0
  216. agentinputlab/experiments/stateful_analysis.py +1738 -0
  217. agentinputlab/experiments/transfer_analysis.py +1316 -0
  218. agentinputlab/experiments/unseen_analysis.py +1025 -0
  219. agentinputlab/integrations/__init__.py +1 -0
  220. agentinputlab/integrations/action_space/__init__.py +1 -0
  221. agentinputlab/integrations/external_workers/__init__.py +76 -0
  222. agentinputlab/integrations/external_workers/ace.py +431 -0
  223. agentinputlab/integrations/external_workers/base.py +181 -0
  224. agentinputlab/integrations/external_workers/decomp_live.py +249 -0
  225. agentinputlab/integrations/external_workers/decomp_planner.py +30 -0
  226. agentinputlab/integrations/external_workers/infocqr_student.py +280 -0
  227. agentinputlab/integrations/external_workers/ircot.py +576 -0
  228. agentinputlab/integrations/external_workers/llmcompiler_live.py +360 -0
  229. agentinputlab/integrations/external_workers/longllmlingua.py +162 -0
  230. agentinputlab/integrations/external_workers/method.py +314 -0
  231. agentinputlab/integrations/external_workers/process.py +678 -0
  232. agentinputlab/integrations/external_workers/supervisor.py +1071 -0
  233. agentinputlab/integrations/full_architecture/__init__.py +1 -0
  234. agentinputlab/integrations/full_architecture/hyde.py +212 -0
  235. agentinputlab/integrations/full_architecture/plan_and_budget.py +188 -0
  236. agentinputlab/integrations/live_history.py +222 -0
  237. agentinputlab/integrations/prompt_optimization/__init__.py +1 -0
  238. agentinputlab/integrations/training/__init__.py +1 -0
  239. agentinputlab/metrics/__init__.py +50 -0
  240. agentinputlab/metrics/catalog.py +313 -0
  241. agentinputlab/metrics/harm.py +95 -0
  242. agentinputlab/metrics/paired.py +40 -0
  243. agentinputlab/metrics/qa.py +76 -0
  244. agentinputlab/metrics/text.py +33 -0
  245. agentinputlab/metrics/third_party.py +161 -0
  246. agentinputlab/models/__init__.py +4 -0
  247. agentinputlab/observability/__init__.py +56 -0
  248. agentinputlab/observability/budget.py +316 -0
  249. agentinputlab/observability/instrumentation.py +754 -0
  250. agentinputlab/observability/models.py +403 -0
  251. agentinputlab/observability/monitor.py +380 -0
  252. agentinputlab/optimizers/__init__.py +1 -0
  253. agentinputlab/optimizers/clarify/__init__.py +12 -0
  254. agentinputlab/optimizers/context/__init__.py +22 -0
  255. agentinputlab/optimizers/decompose/__init__.py +64 -0
  256. agentinputlab/optimizers/evidence/__init__.py +22 -0
  257. agentinputlab/optimizers/rewrite/__init__.py +30 -0
  258. agentinputlab/parity/__init__.py +5 -0
  259. agentinputlab/parity/base.py +206 -0
  260. agentinputlab/production_campaign.py +763 -0
  261. agentinputlab/py.typed +1 -0
  262. agentinputlab/recipes/__init__.py +333 -0
  263. agentinputlab/recipes/ace.py +418 -0
  264. agentinputlab/recipes/ace_bfcl.py +273 -0
  265. agentinputlab/recipes/ace_common_support.py +169 -0
  266. agentinputlab/recipes/ace_native_plan.py +322 -0
  267. agentinputlab/recipes/ace_online.py +711 -0
  268. agentinputlab/recipes/ace_retrieval.py +330 -0
  269. agentinputlab/recipes/ace_statebench.py +382 -0
  270. agentinputlab/recipes/ace_tau.py +463 -0
  271. agentinputlab/recipes/ace_toolsandbox.py +432 -0
  272. agentinputlab/recipes/acon_guideline.py +22 -0
  273. agentinputlab/recipes/adaptive_rag.py +822 -0
  274. agentinputlab/recipes/llmlingua2_inference.py +146 -0
  275. agentinputlab/recipes/llmlingua2_training.py +796 -0
  276. agentinputlab/recipes/mce.py +1486 -0
  277. agentinputlab/recipes/mce_final.py +335 -0
  278. agentinputlab/recipes/mce_text_fit.py +499 -0
  279. agentinputlab/recipes/mce_text_only.py +298 -0
  280. agentinputlab/recipes/qrag.py +729 -0
  281. agentinputlab/recipes/qrag_inference.py +364 -0
  282. agentinputlab/recipes/r2ag_features.py +236 -0
  283. agentinputlab/recipes/r2ag_inference.py +503 -0
  284. agentinputlab/recipes/r2ag_training.py +948 -0
  285. agentinputlab/recipes/recomp_abstractive.py +592 -0
  286. agentinputlab/recipes/recomp_extractive.py +525 -0
  287. agentinputlab/recipes/reformer.py +17 -0
  288. agentinputlab/recipes/rqrag.py +1516 -0
  289. agentinputlab/recipes/rqrag_common_support.py +193 -0
  290. agentinputlab/recipes/rqrag_completion.py +147 -0
  291. agentinputlab/recipes/skillopt.py +462 -0
  292. agentinputlab/recipes/skillopt_common_support.py +154 -0
  293. agentinputlab/recipes/skillopt_environment.py +165 -0
  294. agentinputlab/recipes/skillopt_environment_cli.py +5 -0
  295. agentinputlab/recipes/skillopt_retrieval.py +188 -0
  296. agentinputlab/recipes/smartsearch.py +897 -0
  297. agentinputlab/recipes/smartsearch_curriculum.py +283 -0
  298. agentinputlab/recipes/smartsearch_grpo_contracts.py +58 -0
  299. agentinputlab/recipes/smartsearch_grpo_data.py +314 -0
  300. agentinputlab/recipes/smartsearch_grpo_meter.py +324 -0
  301. agentinputlab/recipes/smartsearch_grpo_prefix.py +144 -0
  302. agentinputlab/recipes/smartsearch_grpo_rollout.py +119 -0
  303. agentinputlab/recipes/smartsearch_grpo_tools.py +364 -0
  304. agentinputlab/recipes/smartsearch_grpo_training.py +305 -0
  305. agentinputlab/recipes/smartsearch_grpo_transport.py +248 -0
  306. agentinputlab/recipes/smartsearch_grpo_worker.py +364 -0
  307. agentinputlab/recipes/smartsearch_inference.py +524 -0
  308. agentinputlab/recipes/smartsearch_native_data.py +172 -0
  309. agentinputlab/recipes/smartsearch_policy.py +616 -0
  310. agentinputlab/recipes/smartsearch_policy_worker.py +9 -0
  311. agentinputlab/recipes/smartsearch_sampling.py +232 -0
  312. agentinputlab/recipes/smartsearch_teacher.py +203 -0
  313. agentinputlab/recipes/smartsearch_training.py +265 -0
  314. agentinputlab/registry/__init__.py +46 -0
  315. agentinputlab/registry/beta.py +727 -0
  316. agentinputlab/registry/data/algorithms.json +1959 -0
  317. agentinputlab/registry/data/auxiliary_method_profiles.json +34 -0
  318. agentinputlab/registry/loader.py +100 -0
  319. agentinputlab/registry/result_manifest.py +184 -0
  320. agentinputlab/registry/runtime.py +195 -0
  321. agentinputlab/registry/schema.py +449 -0
  322. agentinputlab/runner/__init__.py +30 -0
  323. agentinputlab/runner/balance.py +248 -0
  324. agentinputlab/runner/cache.py +135 -0
  325. agentinputlab/runner/composite.py +130 -0
  326. agentinputlab/runner/engine.py +285 -0
  327. agentinputlab/runner/runtime_guard.py +268 -0
  328. agentinputlab/runner/worker.py +80 -0
  329. agentinputlab/runtime_bootstrap.py +288 -0
  330. agentinputlab/selectors/__init__.py +196 -0
  331. agentinputlab/selectors/advanced.py +1514 -0
  332. agentinputlab/selectors/artifacts.py +1087 -0
  333. agentinputlab/selectors/base.py +112 -0
  334. agentinputlab/selectors/campaigns.py +1893 -0
  335. agentinputlab/selectors/catalog.py +86 -0
  336. agentinputlab/selectors/discovery.py +1860 -0
  337. agentinputlab/selectors/evaluation.py +1348 -0
  338. agentinputlab/selectors/features.py +383 -0
  339. agentinputlab/selectors/learned.py +502 -0
  340. agentinputlab/selectors/no_op.py +28 -0
  341. agentinputlab/selectors/outcome_table.py +475 -0
  342. agentinputlab/selectors/producer.py +853 -0
  343. agentinputlab/selectors/rules.py +109 -0
  344. agentinputlab/selectors/runtime.py +420 -0
  345. agentinputlab/studies/__init__.py +16 -0
  346. agentinputlab/studies/design.py +165 -0
  347. agentinputlab/studies/paired.py +322 -0
  348. agentinputlab/studio/__init__.py +6 -0
  349. agentinputlab/studio/campaigns.py +819 -0
  350. agentinputlab/studio/engine.py +1408 -0
  351. agentinputlab/studio/model_backends.py +425 -0
  352. agentinputlab/studio/prepared.py +410 -0
  353. agentinputlab/studio/server.py +613 -0
  354. agentinputlab/studio/static/app.js +397 -0
  355. agentinputlab/studio/static/campaigns.html +175 -0
  356. agentinputlab/studio/static/campaigns.js +426 -0
  357. agentinputlab/studio/static/index.html +192 -0
  358. agentinputlab/studio/static/result.html +103 -0
  359. agentinputlab/studio/static/result.js +273 -0
  360. agentinputlab/studio/static/styles.css +467 -0
  361. agentinputlab-0.1.0a1.data/data/share/agentinputlab/CITATION.cff +22 -0
  362. agentinputlab-0.1.0a1.dist-info/METADATA +392 -0
  363. agentinputlab-0.1.0a1.dist-info/RECORD +469 -0
  364. agentinputlab-0.1.0a1.dist-info/WHEEL +5 -0
  365. agentinputlab-0.1.0a1.dist-info/entry_points.txt +6 -0
  366. agentinputlab-0.1.0a1.dist-info/licenses/LICENSE +21 -0
  367. agentinputlab-0.1.0a1.dist-info/licenses/NOTICE +31 -0
  368. agentinputlab-0.1.0a1.dist-info/licenses/THIRD_PARTY_NOTICES.md +94 -0
  369. agentinputlab-0.1.0a1.dist-info/top_level.txt +2 -0
  370. workers/README.md +94 -0
  371. workers/ace/README.md +261 -0
  372. workers/ace/__init__.py +1 -0
  373. workers/ace/compatibility_manifest.json +39 -0
  374. workers/ace/driver.py +1682 -0
  375. workers/acon/README.md +215 -0
  376. workers/acon/__init__.py +5 -0
  377. workers/acon/compatibility_manifest.json +93 -0
  378. workers/acon/driver.py +1859 -0
  379. workers/acon/fitted.py +395 -0
  380. workers/acon/guideline.py +470 -0
  381. workers/acon/guideline_source_manifest.json +20 -0
  382. workers/adaptive_rag/README.md +96 -0
  383. workers/adaptive_rag/__init__.py +15 -0
  384. workers/adaptive_rag/compatibility_manifest.json +61 -0
  385. workers/adaptive_rag/driver.py +1470 -0
  386. workers/adaptive_rag/fixture_runtime.py +163 -0
  387. workers/adaptive_rag/runtime.py +252 -0
  388. workers/ask_or_assume/README.md +124 -0
  389. workers/ask_or_assume/__init__.py +1 -0
  390. workers/ask_or_assume/compatibility_manifest.json +36 -0
  391. workers/ask_or_assume/driver.py +1227 -0
  392. workers/decomp/README.md +133 -0
  393. workers/decomp/UPSTREAM_LICENSE.Apache-2.0.txt +201 -0
  394. workers/decomp/__init__.py +5 -0
  395. workers/decomp/compatibility_manifest.json +205 -0
  396. workers/decomp/driver.py +1015 -0
  397. workers/decomp/live.py +137 -0
  398. workers/decomp/runtime.py +231 -0
  399. workers/infocqr/README.md +106 -0
  400. workers/infocqr/__init__.py +1 -0
  401. workers/infocqr/compatibility_manifest.json +84 -0
  402. workers/infocqr/driver.py +1766 -0
  403. workers/ircot/README.md +167 -0
  404. workers/ircot/__init__.py +7 -0
  405. workers/ircot/compatibility_manifest.json +121 -0
  406. workers/ircot/controlled.py +457 -0
  407. workers/ircot/driver.py +1373 -0
  408. workers/ircot/fixture_bindings.json +39 -0
  409. workers/ircot/fixture_config.json +48 -0
  410. workers/ircot/fixture_prompt.txt +2 -0
  411. workers/ircot/native_qa.py +398 -0
  412. workers/ircot/qa_compatibility_manifest.json +57 -0
  413. workers/ircot/runtime.py +97 -0
  414. workers/llmcompiler/README.md +160 -0
  415. workers/llmcompiler/__init__.py +1 -0
  416. workers/llmcompiler/compatibility_manifest.json +45 -0
  417. workers/llmcompiler/driver.py +1514 -0
  418. workers/llmcompiler/export_prompts.py +74 -0
  419. workers/llmcompiler/live.py +174 -0
  420. workers/llmcompiler/patch_manifest.json +147 -0
  421. workers/mce/README.md +198 -0
  422. workers/mce/__init__.py +29 -0
  423. workers/mce/compatibility_manifest.json +115 -0
  424. workers/mce/driver.py +2585 -0
  425. workers/mce/final_eval.py +173 -0
  426. workers/official_arenas/README.md +25 -0
  427. workers/official_arenas/__init__.py +1 -0
  428. workers/official_arenas/capture_records.py +110 -0
  429. workers/official_arenas/driver.py +619 -0
  430. workers/official_arenas/protocol.py +334 -0
  431. workers/official_arenas/providers/__init__.py +1 -0
  432. workers/official_arenas/providers/_live_history.py +228 -0
  433. workers/official_arenas/providers/_message_compat.py +77 -0
  434. workers/official_arenas/providers/appworld_ul.py +196 -0
  435. workers/official_arenas/providers/fixture.py +112 -0
  436. workers/official_arenas/providers/statebench.py +960 -0
  437. workers/official_arenas/providers/tau3.py +1672 -0
  438. workers/official_arenas/providers/tau_retail.py +586 -0
  439. workers/official_arenas/providers/toolsandbox.py +1321 -0
  440. workers/official_arenas/providers/tooltalk.py +699 -0
  441. workers/official_arenas/retail_capture.py +188 -0
  442. workers/official_arenas/retail_driver.py +275 -0
  443. workers/qrag/README.md +172 -0
  444. workers/qrag/__init__.py +1 -0
  445. workers/qrag/compatibility_manifest.json +64 -0
  446. workers/qrag/driver.py +673 -0
  447. workers/r2ag/README.md +154 -0
  448. workers/r2ag/__init__.py +5 -0
  449. workers/r2ag/compatibility_manifest.json +55 -0
  450. workers/r2ag/driver.py +1287 -0
  451. workers/r2ag/feature_schema.json +8 -0
  452. workers/r2ag/runtime.py +589 -0
  453. workers/recomp/README.md +120 -0
  454. workers/recomp/__init__.py +1 -0
  455. workers/recomp/compatibility_manifest.json +43 -0
  456. workers/recomp/driver.py +1653 -0
  457. workers/reformer/README.md +125 -0
  458. workers/reformer/__init__.py +1 -0
  459. workers/reformer/compatibility_manifest.json +68 -0
  460. workers/reformer/driver.py +1459 -0
  461. workers/reformer/induction.py +457 -0
  462. workers/rrr/README.md +126 -0
  463. workers/rrr/__init__.py +1 -0
  464. workers/rrr/compatibility_manifest.json +57 -0
  465. workers/rrr/driver.py +1133 -0
  466. workers/smartsearch/README.md +331 -0
  467. workers/smartsearch/__init__.py +1 -0
  468. workers/smartsearch/compatibility_manifest.json +41 -0
  469. workers/smartsearch/driver.py +1643 -0
@@ -0,0 +1,193 @@
1
+ """RQ-RAG tree output mapped to the shared reformulation seam.
2
+
3
+ The controlled RQ-RAG implementation produces a terminal answer and selected
4
+ retrieval evidence. The paper taxonomy classifies RQ-RAG by its agent-visible
5
+ query refinement. This adapter therefore runs the complete controlled tree on
6
+ a read-only view of the visible request, keeps only query actions on the
7
+ model-selected path, and appends those refinements to the original query. It
8
+ never replaces the downstream agent, forwards retrieved evidence or terminal
9
+ answers, exposes tool schemas, or exposes scorer metadata to the optimizer.
10
+ """
11
+
12
+ from __future__ import annotations
13
+
14
+ from dataclasses import dataclass, field
15
+
16
+ from agentinputlab.contracts import (
17
+ CompositeResult,
18
+ ExecutionTrace,
19
+ FrozenAgentState,
20
+ InputPatch,
21
+ InterventionContext,
22
+ PatchOp,
23
+ PatchOperation,
24
+ SemanticFamily,
25
+ TraceEvent,
26
+ TracePhase,
27
+ )
28
+ from agentinputlab.contracts.serialization import content_hash
29
+
30
+ from .rqrag import RQ_RAG_UPSTREAM_COMMIT, RQRAGMethodControls, RQRAGMethodTreeSearch
31
+
32
+ RQ_RAG_COMMON_SUPPORT_RELEASE_ID = "recipe.rqrag.common_support_queries"
33
+ RQ_RAG_COMMON_SUPPORT_MAPPING = "append_selected_query_refinements.v1"
34
+
35
+
36
+ @dataclass(slots=True)
37
+ class RQRAGCommonSupportQueries:
38
+ """Run the controlled tree and append only selected-path query refinements."""
39
+
40
+ controls: RQRAGMethodControls = field(default_factory=RQRAGMethodControls)
41
+ method_id: str = "recipe.rqrag"
42
+ release_id: str = RQ_RAG_COMMON_SUPPORT_RELEASE_ID
43
+ family: SemanticFamily = SemanticFamily.REFORMULATE
44
+
45
+ def propose(self, state: FrozenAgentState, context: InterventionContext) -> InputPatch:
46
+ optimization_state = FrozenAgentState(
47
+ query=state.query,
48
+ task_id=state.task_id,
49
+ instruction=state.instruction,
50
+ context=state.context,
51
+ messages=state.messages,
52
+ tool_names=(),
53
+ metadata={},
54
+ budget=state.budget,
55
+ )
56
+ result = RQRAGMethodTreeSearch(controls=self.controls).run(
57
+ optimization_state,
58
+ context,
59
+ )
60
+ selected_path = tuple(result.metadata["selected_path"])
61
+ refinements = _selected_query_refinements(result, selected_path=selected_path)
62
+ appended_query = _serialize_refinements(refinements)
63
+ selected_document_ids = tuple(result.metadata["selected_document_ids"])
64
+ operation = (
65
+ PatchOperation(
66
+ PatchOp.APPEND_QUERY,
67
+ appended_query,
68
+ target="query",
69
+ rationale=(
70
+ "Expose only RQ-RAG query actions on the model-selected path while "
71
+ "preserving the original request and downstream agent."
72
+ ),
73
+ )
74
+ if appended_query
75
+ else PatchOperation(
76
+ PatchOp.NO_OP,
77
+ rationale="RQ-RAG selected the root direct-response branch with no query action.",
78
+ )
79
+ )
80
+ trace = ExecutionTrace(
81
+ method_id=self.method_id,
82
+ release_id=self.release_id,
83
+ events=(
84
+ TraceEvent(
85
+ TracePhase.PREPARE,
86
+ "rqrag_common_support_view",
87
+ {
88
+ "source_state_hash": state.state_hash,
89
+ "optimization_state_hash": optimization_state.state_hash,
90
+ "tool_schema_count": len(state.tool_names),
91
+ "tool_schemas_forwarded": False,
92
+ "metadata_forwarded": False,
93
+ "upstream_commit": RQ_RAG_UPSTREAM_COMMIT,
94
+ },
95
+ ),
96
+ *result.trace.events,
97
+ TraceEvent(
98
+ TracePhase.FINALIZE,
99
+ "rqrag_selected_queries_appended",
100
+ {
101
+ "mapping": RQ_RAG_COMMON_SUPPORT_MAPPING,
102
+ "operation": operation.op.value,
103
+ "selected_path": selected_path,
104
+ "selected_query_refinements": refinements,
105
+ "selected_refinements_hash": content_hash(refinements),
106
+ "terminal_answer_forwarded": False,
107
+ "retrieved_evidence_forwarded": False,
108
+ },
109
+ ),
110
+ ),
111
+ counters={
112
+ **dict(result.trace.counters),
113
+ "selected_query_refinements": len(refinements),
114
+ "optimizer_tool_schema_count": 0,
115
+ },
116
+ warnings=(
117
+ *result.trace.warnings,
118
+ "controlled_query_only_mapping_not_released_7b_or_paper_result_parity",
119
+ ),
120
+ )
121
+ return InputPatch(
122
+ method_id=self.method_id,
123
+ release_id=self.release_id,
124
+ source_state_hash=state.state_hash,
125
+ operations=(operation,),
126
+ trace=trace,
127
+ implementation_identity=(
128
+ f"controlled_method_tree:chanchimin/RQ-RAG@{RQ_RAG_UPSTREAM_COMMIT}+"
129
+ f"{RQ_RAG_COMMON_SUPPORT_MAPPING}"
130
+ ),
131
+ profile_kind="controlled_adapter",
132
+ metadata={
133
+ "common_support_mapping": RQ_RAG_COMMON_SUPPORT_MAPPING,
134
+ "method_tree_release_id": result.release_id,
135
+ "method_tree_implementation_identity": result.implementation_identity,
136
+ "selected_node_id": result.metadata["selected_node_id"],
137
+ "selected_path": selected_path,
138
+ "selected_query_refinements": refinements,
139
+ "selected_refinements_hash": content_hash(refinements),
140
+ "selected_document_ids_internal_only": selected_document_ids,
141
+ "selection_mode": result.metadata["selection_mode"],
142
+ "score_source": result.metadata["score_source"],
143
+ "termination": result.metadata["termination"],
144
+ "comparison_eligible": result.metadata["comparison_eligible"],
145
+ "tree_decode": result.metadata["tree_decode"],
146
+ "oracle": result.metadata["oracle"],
147
+ "tool_schemas_forwarded": False,
148
+ "state_metadata_forwarded": False,
149
+ "terminal_answer_forwarded": False,
150
+ "retrieved_evidence_forwarded": False,
151
+ "paper_checkpoint_parity": False,
152
+ "paper_result_parity_claimed": False,
153
+ },
154
+ )
155
+
156
+
157
+ def _selected_query_refinements(
158
+ result: CompositeResult,
159
+ *,
160
+ selected_path: tuple[str, ...],
161
+ ) -> tuple[tuple[str, str], ...]:
162
+ selected = set(selected_path)
163
+ refinements: list[tuple[str, str]] = []
164
+ for event in result.trace.events:
165
+ if event.name != "tree_node":
166
+ continue
167
+ payload = event.payload
168
+ node_id = payload.get("node_id")
169
+ query_type = payload.get("query_type")
170
+ query = payload.get("query")
171
+ if (
172
+ node_id in selected
173
+ and query_type in {"rewrite", "decompose", "disambiguate"}
174
+ and isinstance(query, str)
175
+ and query.strip()
176
+ ):
177
+ refinements.append((str(query_type), query.strip()))
178
+ return tuple(refinements)
179
+
180
+
181
+ def _serialize_refinements(refinements: tuple[tuple[str, str], ...]) -> str:
182
+ if not refinements:
183
+ return ""
184
+ lines = ["RQ-RAG selected query refinement path:"]
185
+ lines.extend(f"- [{kind}] {query}" for kind, query in refinements)
186
+ return "\n".join(lines)
187
+
188
+
189
+ __all__ = [
190
+ "RQ_RAG_COMMON_SUPPORT_MAPPING",
191
+ "RQ_RAG_COMMON_SUPPORT_RELEASE_ID",
192
+ "RQRAGCommonSupportQueries",
193
+ ]
@@ -0,0 +1,147 @@
1
+ """Explicit raw-completion scoring for the controlled RQ-RAG tree.
2
+
3
+ One echoed completion supplies prompt and generated-token likelihoods. No extra
4
+ teacher-forcing request, hidden retry, chat-template substitution or local model
5
+ copy is made. This is a backend-neutral controlled profile, not released-7B
6
+ tokenizer or checkpoint parity.
7
+ """
8
+
9
+ from __future__ import annotations
10
+
11
+ import math
12
+ import os
13
+ from collections.abc import Mapping, Sequence
14
+ from dataclasses import dataclass
15
+ from typing import Any
16
+
17
+ from agentinputlab.backends.llm.base import GenerationRequest, GenerationResponse, ModelCapability
18
+ from agentinputlab.backends.llm.openai_compatible import (
19
+ OpenAICompatibleBackend,
20
+ _contains_exact_text,
21
+ _parse_generation_response,
22
+ )
23
+ from agentinputlab.contracts import CapabilityError, ValidationError
24
+ from agentinputlab.contracts.serialization import to_jsonable
25
+
26
+ _SAMPLING_EXTRAS = frozenset({
27
+ "top_p", "top_k", "min_p", "repetition_penalty", "presence_penalty", "frequency_penalty",
28
+ })
29
+
30
+
31
+ def _log_probability(value: object) -> float:
32
+ if isinstance(value, bool) or not isinstance(value, (int, float)):
33
+ raise ValueError("missing token log probability")
34
+ result = float(value)
35
+ if not math.isfinite(result) or result > 0:
36
+ raise ValueError("token log probability must be finite and non-positive")
37
+ return result
38
+
39
+
40
+ def echoed_path_score(data: Mapping[str, Any], prompt: str) -> tuple[str, dict[str, Any]]:
41
+ """Reject incomplete prompt likelihoods rather than call completion-only PPL full-path."""
42
+ choice = data["choices"][0]
43
+ text = choice["text"]
44
+ if not isinstance(text, str) or not text.startswith(prompt):
45
+ raise ValueError("echoed completion does not preserve the exact input prompt")
46
+ usage, logs = data["usage"], choice["logprobs"]
47
+ prompt_count, completion_count = usage["prompt_tokens"], usage["completion_tokens"]
48
+ if any(isinstance(n, bool) or not isinstance(n, int) or n < 0
49
+ for n in (prompt_count, completion_count)) or prompt_count < 1:
50
+ raise ValueError("echoed completion has invalid native usage")
51
+ values, top = logs["token_logprobs"], logs["top_logprobs"]
52
+ if (not isinstance(values, Sequence) or isinstance(values, (str, bytes))
53
+ or not isinstance(top, Sequence) or isinstance(top, (str, bytes))
54
+ or len(values) != len(top)
55
+ or not prompt_count < len(values) <= prompt_count + completion_count):
56
+ raise ValueError("echoed completion does not contain full prompt and answer likelihoods")
57
+ path_values = []
58
+ for index, value in enumerate(values):
59
+ if index == 0 and value is None:
60
+ continue # The unconditional first prompt token has no predecessor.
61
+ path_values.append(_log_probability(value))
62
+ if not path_values:
63
+ raise ValueError("echoed path has no scored tokens")
64
+ confidence = []
65
+ for candidates in top[prompt_count:]:
66
+ if not isinstance(candidates, Mapping) or not candidates:
67
+ raise ValueError("answer likelihoods require non-empty top-token probabilities")
68
+ confidence.append(math.exp(max(_log_probability(value) for value in candidates.values())))
69
+ perplexity = math.exp(-math.fsum(path_values) / len(path_values))
70
+ return text[len(prompt):], {
71
+ "confidence": math.fsum(confidence) / len(confidence),
72
+ "perplexity": perplexity,
73
+ "perplexity_scope": "full_path",
74
+ "scoring_profile": "raw_completion_echo_all_returned_path_tokens.v1",
75
+ "source": "native_echoed_prompt_and_completion_logprobs",
76
+ "prompt_tokens": prompt_count,
77
+ "scored_path_tokens": len(path_values),
78
+ "scored_answer_tokens": len(confidence),
79
+ "unreturned_stop_tokens_excluded": prompt_count + completion_count - len(values),
80
+ "first_unconditional_token_excluded": values[0] is None,
81
+ "chat_template_applied": False,
82
+ "released_7b_checkpoint_parity": False,
83
+ }
84
+
85
+
86
+ @dataclass(slots=True)
87
+ class RQRAGCompletionBackend(OpenAICompatibleBackend):
88
+ """Opt-in /completions client with exact full-path likelihood accounting.
89
+
90
+ Callers supply the RQ-RAG serialized prompt. Chat messages, tools, JSON mode,
91
+ unknown body overrides and absent native prompt logprobs fail closed. Every
92
+ successful ``generate`` is exactly one physical API/model generation.
93
+ """
94
+
95
+ def __post_init__(self) -> None:
96
+ OpenAICompatibleBackend.__post_init__(self)
97
+ if set(self.extra_body) - _SAMPLING_EXTRAS:
98
+ raise ValidationError("RQ-RAG raw completion accepts sampling extras only")
99
+ self.capabilities = frozenset({
100
+ ModelCapability.COMPLETION, ModelCapability.LOGPROBS, ModelCapability.SEED,
101
+ })
102
+
103
+ def generate(self, request: GenerationRequest) -> GenerationResponse:
104
+ if (not request.prompt or request.messages or request.system or request.tools
105
+ or request.response_format is not None):
106
+ raise CapabilityError("RQ-RAG raw completion requires a prompt without chat/tools/JSON mode")
107
+ payload = dict(to_jsonable({
108
+ **dict(self.extra_body), "model": self.model_id, "prompt": request.prompt,
109
+ "max_tokens": request.max_tokens, "temperature": request.temperature,
110
+ "echo": True, "logprobs": max(5, request.top_logprobs or 0),
111
+ "add_special_tokens": False,
112
+ }))
113
+ if request.seed is not None:
114
+ payload["seed"] = request.seed
115
+ if request.stop:
116
+ payload["stop"] = list(request.stop)
117
+ key = self.api_key or os.environ.get(self.api_key_env, "")
118
+ headers = {"Content-Type": "application/json", **dict(self.extra_headers)}
119
+ if key:
120
+ headers["Authorization"] = f"Bearer {key}"
121
+ data, failed = None, False
122
+ try:
123
+ data = self.transport(f"{self.base_url}/completions", headers, payload, self.timeout_seconds)
124
+ except Exception:
125
+ failed = True
126
+ if failed:
127
+ raise RuntimeError("RQ-RAG completion transport failed; no automatic retry")
128
+ if key and _contains_exact_text(data, key):
129
+ raise RuntimeError("RQ-RAG completion response contains the configured credential")
130
+ response = None
131
+ try:
132
+ text, score = echoed_path_score(data, request.prompt)
133
+ parsed = _parse_generation_response(data, fallback_model=self.model_id)
134
+ response = GenerationResponse(text=text, model=parsed.model,
135
+ finish_reason=parsed.finish_reason, usage=parsed.usage,
136
+ raw={**dict(data), "rqrag_score": score})
137
+ except (AttributeError, IndexError, KeyError, OverflowError, TypeError, ValueError):
138
+ failed = True
139
+ if failed or response is None:
140
+ raise RuntimeError("RQ-RAG completion lacks valid full-path probabilities")
141
+ return response
142
+
143
+ def fingerprint(self) -> Mapping[str, Any]:
144
+ return {**dict(OpenAICompatibleBackend.fingerprint(self)),
145
+ "backend": "rqrag_raw_completion_echo",
146
+ "scoring_profile": "raw_completion_echo_all_returned_path_tokens.v1",
147
+ "chat_template_applied": False, "automatic_retry": False}