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.
- agentinputlab/__init__.py +28 -0
- agentinputlab/algorithms/__init__.py +6 -0
- agentinputlab/algorithms/acquire/__init__.py +145 -0
- agentinputlab/algorithms/acquire/adaptive_rag/__init__.py +43 -0
- agentinputlab/algorithms/acquire/adaptive_rag/replay.py +367 -0
- agentinputlab/algorithms/acquire/adaptive_rag/runtime.py +1202 -0
- agentinputlab/algorithms/acquire/adaptive_rag_common_support.py +542 -0
- agentinputlab/algorithms/acquire/hyde.py +296 -0
- agentinputlab/algorithms/acquire/hyde_controlled.py +200 -0
- agentinputlab/algorithms/acquire/hyde_external.py +148 -0
- agentinputlab/algorithms/acquire/ircot.py +28 -0
- agentinputlab/algorithms/acquire/ircot_common_support.py +478 -0
- agentinputlab/algorithms/acquire/ircot_native.py +355 -0
- agentinputlab/algorithms/acquire/ircot_replay.py +327 -0
- agentinputlab/algorithms/acquire/qrag/__init__.py +79 -0
- agentinputlab/algorithms/acquire/qrag/bridge.py +828 -0
- agentinputlab/algorithms/acquire/qrag/common_support.py +891 -0
- agentinputlab/algorithms/acquire/qrag/native_axis.py +71 -0
- agentinputlab/algorithms/acquire/qrag/reader_adapter.py +139 -0
- agentinputlab/algorithms/acquire/r2ag/__init__.py +57 -0
- agentinputlab/algorithms/acquire/r2ag/bridge.py +1312 -0
- agentinputlab/algorithms/acquire/r2ag/common_support.py +1029 -0
- agentinputlab/algorithms/acquire/r2ag/socket_runtime.py +255 -0
- agentinputlab/algorithms/clarify/__init__.py +61 -0
- agentinputlab/algorithms/clarify/_structured.py +64 -0
- agentinputlab/algorithms/clarify/active_task_disambiguation.py +371 -0
- agentinputlab/algorithms/clarify/ask_or_assume/__init__.py +17 -0
- agentinputlab/algorithms/clarify/ask_or_assume/monitor.py +410 -0
- agentinputlab/algorithms/clarify/future_turns.py +379 -0
- agentinputlab/algorithms/clarify/udcs.py +255 -0
- agentinputlab/algorithms/clarify/udcs_deterministic.py +196 -0
- agentinputlab/algorithms/curate/__init__.py +65 -0
- agentinputlab/algorithms/curate/acon/__init__.py +27 -0
- agentinputlab/algorithms/curate/acon/common.py +284 -0
- agentinputlab/algorithms/curate/acon/common_support.py +136 -0
- agentinputlab/algorithms/curate/acon/history_bridge.py +309 -0
- agentinputlab/algorithms/curate/acon/observation_bridge.py +290 -0
- agentinputlab/algorithms/curate/llmlingua/__init__.py +30 -0
- agentinputlab/algorithms/curate/llmlingua/adapters.py +537 -0
- agentinputlab/algorithms/curate/llmlingua/budget.py +48 -0
- agentinputlab/algorithms/curate/llmlingua/controlled.py +175 -0
- agentinputlab/algorithms/curate/llmlingua/multilingual.py +105 -0
- agentinputlab/algorithms/curate/llmlingua/span_map.py +167 -0
- agentinputlab/algorithms/curate/recomp/__init__.py +47 -0
- agentinputlab/algorithms/curate/recomp/external.py +1129 -0
- agentinputlab/algorithms/curate/recomp/selective.py +304 -0
- agentinputlab/algorithms/decompose/__init__.py +107 -0
- agentinputlab/algorithms/decompose/decomp/__init__.py +53 -0
- agentinputlab/algorithms/decompose/decomp/common_support.py +478 -0
- agentinputlab/algorithms/decompose/decomp/planner_component.py +199 -0
- agentinputlab/algorithms/decompose/decomp/worker_bridge.py +416 -0
- agentinputlab/algorithms/decompose/llm_compiler/__init__.py +35 -0
- agentinputlab/algorithms/decompose/llm_compiler/bridge.py +520 -0
- agentinputlab/algorithms/decompose/llm_compiler/common_support.py +504 -0
- agentinputlab/algorithms/decompose/plan_and_budget/__init__.py +43 -0
- agentinputlab/algorithms/decompose/plan_and_budget/cached_soft_budget.py +362 -0
- agentinputlab/algorithms/decompose/plan_and_budget/decomposer_component.py +107 -0
- agentinputlab/algorithms/decompose/plan_and_budget/full.py +205 -0
- agentinputlab/algorithms/decompose/plan_and_budget/hard_cap.py +246 -0
- agentinputlab/algorithms/decompose/plan_and_budget/regenerated_soft_budget.py +448 -0
- agentinputlab/algorithms/decompose/plan_and_solve/__init__.py +33 -0
- agentinputlab/algorithms/decompose/plan_and_solve/plan_component.py +192 -0
- agentinputlab/algorithms/decompose/plan_and_solve/ps_prompt.py +127 -0
- agentinputlab/algorithms/decompose/self_ask/__init__.py +27 -0
- agentinputlab/algorithms/decompose/self_ask/loop.py +362 -0
- agentinputlab/algorithms/decompose/self_ask/planner_component.py +117 -0
- agentinputlab/algorithms/reformulate/__init__.py +85 -0
- agentinputlab/algorithms/reformulate/infocqr/__init__.py +40 -0
- agentinputlab/algorithms/reformulate/infocqr/cache_replay.py +362 -0
- agentinputlab/algorithms/reformulate/infocqr/external.py +675 -0
- agentinputlab/algorithms/reformulate/infocqr/student.py +189 -0
- agentinputlab/algorithms/reformulate/query2doc/__init__.py +42 -0
- agentinputlab/algorithms/reformulate/query2doc/controlled_live.py +190 -0
- agentinputlab/algorithms/reformulate/query2doc/models.py +95 -0
- agentinputlab/algorithms/reformulate/query2doc/official_cached.py +335 -0
- agentinputlab/algorithms/reformulate/query2doc/querygym_bridge.py +223 -0
- agentinputlab/algorithms/reformulate/query2doc/querygym_runtime.py +326 -0
- agentinputlab/algorithms/reformulate/reformer/__init__.py +23 -0
- agentinputlab/algorithms/reformulate/reformer/external.py +417 -0
- agentinputlab/algorithms/reformulate/rrr.py +696 -0
- agentinputlab/annotation/__init__.py +5 -0
- agentinputlab/annotation/independent.py +428 -0
- agentinputlab/annotation/server.py +233 -0
- agentinputlab/annotation/static/app.js +121 -0
- agentinputlab/annotation/static/index.html +60 -0
- agentinputlab/annotation/static/review.css +2 -0
- agentinputlab/annotation/static/review.html +10 -0
- agentinputlab/annotation/static/review.js +11 -0
- agentinputlab/annotation/static/styles.css +55 -0
- agentinputlab/annotation/store.py +444 -0
- agentinputlab/api.py +368 -0
- agentinputlab/artifacts/__init__.py +15 -0
- agentinputlab/artifacts/json_file.py +104 -0
- agentinputlab/artifacts/manifest.py +98 -0
- agentinputlab/artifacts/store.py +157 -0
- agentinputlab/backends/__init__.py +98 -0
- agentinputlab/backends/embedding/__init__.py +25 -0
- agentinputlab/backends/embedding/base.py +145 -0
- agentinputlab/backends/embedding/callable.py +75 -0
- agentinputlab/backends/embedding/replay.py +129 -0
- agentinputlab/backends/llm/__init__.py +31 -0
- agentinputlab/backends/llm/base.py +149 -0
- agentinputlab/backends/llm/callable.py +108 -0
- agentinputlab/backends/llm/decomp_completion.py +88 -0
- agentinputlab/backends/llm/final_content_stop.py +72 -0
- agentinputlab/backends/llm/gpt_oss_final_channel.py +84 -0
- agentinputlab/backends/llm/openai_compatible.py +332 -0
- agentinputlab/backends/llm/transformers.py +171 -0
- agentinputlab/backends/retrieval/__init__.py +35 -0
- agentinputlab/backends/retrieval/base.py +50 -0
- agentinputlab/backends/retrieval/dense.py +777 -0
- agentinputlab/backends/retrieval/memory.py +119 -0
- agentinputlab/backends/retrieval/sqlite_fts.py +155 -0
- agentinputlab/backends/user/__init__.py +15 -0
- agentinputlab/backends/user/base.py +66 -0
- agentinputlab/backends/user/callable.py +54 -0
- agentinputlab/backends/user/replay.py +89 -0
- agentinputlab/backends/vector.py +38 -0
- agentinputlab/benchmarks/__init__.py +246 -0
- agentinputlab/benchmarks/acquisition.py +106 -0
- agentinputlab/benchmarks/ale.py +540 -0
- agentinputlab/benchmarks/appworld_ul.py +669 -0
- agentinputlab/benchmarks/assets.py +226 -0
- agentinputlab/benchmarks/base.py +68 -0
- agentinputlab/benchmarks/beir_archive.py +235 -0
- agentinputlab/benchmarks/beir_sample.py +269 -0
- agentinputlab/benchmarks/bfcl_native.py +245 -0
- agentinputlab/benchmarks/bfcl_v3.py +607 -0
- agentinputlab/benchmarks/cache.py +218 -0
- agentinputlab/benchmarks/clariq.py +203 -0
- agentinputlab/benchmarks/data/registry.json +321 -0
- agentinputlab/benchmarks/data/smoke_assets.json +511 -0
- agentinputlab/benchmarks/data/statebench_v0_8_1_task_ids.json +475 -0
- agentinputlab/benchmarks/data/toolsandbox_165848b_scenario_names.json +1 -0
- agentinputlab/benchmarks/execution.py +1111 -0
- agentinputlab/benchmarks/interactive.py +257 -0
- agentinputlab/benchmarks/ir_query.py +204 -0
- agentinputlab/benchmarks/longbench.py +203 -0
- agentinputlab/benchmarks/paired_replay.py +479 -0
- agentinputlab/benchmarks/preprocess.py +2313 -0
- agentinputlab/benchmarks/production.py +236 -0
- agentinputlab/benchmarks/qa.py +963 -0
- agentinputlab/benchmarks/registry.py +110 -0
- agentinputlab/benchmarks/retrieval.py +459 -0
- agentinputlab/benchmarks/runtime.py +509 -0
- agentinputlab/benchmarks/statebench.py +938 -0
- agentinputlab/benchmarks/tau3.py +725 -0
- agentinputlab/benchmarks/toolsandbox.py +626 -0
- agentinputlab/benchmarks/trec_dl_sample.py +263 -0
- agentinputlab/campaign.py +654 -0
- agentinputlab/campaign_runner.py +953 -0
- agentinputlab/cli/__init__.py +6 -0
- agentinputlab/cli/main.py +933 -0
- agentinputlab/connectors/__init__.py +46 -0
- agentinputlab/connectors/knowledge/__init__.py +4 -0
- agentinputlab/connectors/tools/__init__.py +30 -0
- agentinputlab/connectors/tools/base.py +57 -0
- agentinputlab/connectors/tools/mcp.py +222 -0
- agentinputlab/connectors/tools/memory.py +41 -0
- agentinputlab/connectors/tools/openapi.py +695 -0
- agentinputlab/connectors/tools/schema.py +102 -0
- agentinputlab/contracts/__init__.py +91 -0
- agentinputlab/contracts/availability.py +24 -0
- agentinputlab/contracts/dependency.py +58 -0
- agentinputlab/contracts/enums.py +82 -0
- agentinputlab/contracts/errors.py +63 -0
- agentinputlab/contracts/evidence.py +43 -0
- agentinputlab/contracts/intent.py +110 -0
- agentinputlab/contracts/lifecycle.py +164 -0
- agentinputlab/contracts/patch.py +167 -0
- agentinputlab/contracts/serialization.py +93 -0
- agentinputlab/contracts/state.py +127 -0
- agentinputlab/contracts/trace.py +59 -0
- agentinputlab/contracts/worker.py +42 -0
- agentinputlab/core/__init__.py +7 -0
- agentinputlab/data/__init__.py +32 -0
- agentinputlab/data/checksums.py +9 -0
- agentinputlab/data/downloader.py +235 -0
- agentinputlab/data/http.py +171 -0
- agentinputlab/data/index.py +21 -0
- agentinputlab/data/overlap.py +28 -0
- agentinputlab/data/preprocess.py +32 -0
- agentinputlab/data/storage.py +57 -0
- agentinputlab/data/trec_sample.py +427 -0
- agentinputlab/evaluation/__init__.py +5 -0
- agentinputlab/evaluation/outcomes.py +32 -0
- agentinputlab/experiments/__init__.py +278 -0
- agentinputlab/experiments/analysis.py +981 -0
- agentinputlab/experiments/artifacts.py +687 -0
- agentinputlab/experiments/automatic_measurement.py +678 -0
- agentinputlab/experiments/auxiliary_matrix.py +203 -0
- agentinputlab/experiments/bindings.py +430 -0
- agentinputlab/experiments/blueprints.py +147 -0
- agentinputlab/experiments/builders/__init__.py +45 -0
- agentinputlab/experiments/builders/executable_external_payload.py +423 -0
- agentinputlab/experiments/builders/external_process_arena.py +1018 -0
- agentinputlab/experiments/builders/registry_intervention.py +463 -0
- agentinputlab/experiments/cli.py +158 -0
- agentinputlab/experiments/controls.py +340 -0
- agentinputlab/experiments/driver_guard.py +205 -0
- agentinputlab/experiments/drivers/__init__.py +89 -0
- agentinputlab/experiments/drivers/external_arena.py +2908 -0
- agentinputlab/experiments/drivers/method_process.py +662 -0
- agentinputlab/experiments/drivers/parity_process.py +1484 -0
- agentinputlab/experiments/drivers/static_production.py +1574 -0
- agentinputlab/experiments/executor.py +448 -0
- agentinputlab/experiments/loader.py +605 -0
- agentinputlab/experiments/measurement.py +675 -0
- agentinputlab/experiments/paper_analysis.py +1818 -0
- agentinputlab/experiments/parity_identity.py +605 -0
- agentinputlab/experiments/plan.py +277 -0
- agentinputlab/experiments/production_builders.py +3739 -0
- agentinputlab/experiments/readiness.py +432 -0
- agentinputlab/experiments/schema.py +514 -0
- agentinputlab/experiments/secrets.py +177 -0
- agentinputlab/experiments/stateful_analysis.py +1738 -0
- agentinputlab/experiments/transfer_analysis.py +1316 -0
- agentinputlab/experiments/unseen_analysis.py +1025 -0
- agentinputlab/integrations/__init__.py +1 -0
- agentinputlab/integrations/action_space/__init__.py +1 -0
- agentinputlab/integrations/external_workers/__init__.py +76 -0
- agentinputlab/integrations/external_workers/ace.py +431 -0
- agentinputlab/integrations/external_workers/base.py +181 -0
- agentinputlab/integrations/external_workers/decomp_live.py +249 -0
- agentinputlab/integrations/external_workers/decomp_planner.py +30 -0
- agentinputlab/integrations/external_workers/infocqr_student.py +280 -0
- agentinputlab/integrations/external_workers/ircot.py +576 -0
- agentinputlab/integrations/external_workers/llmcompiler_live.py +360 -0
- agentinputlab/integrations/external_workers/longllmlingua.py +162 -0
- agentinputlab/integrations/external_workers/method.py +314 -0
- agentinputlab/integrations/external_workers/process.py +678 -0
- agentinputlab/integrations/external_workers/supervisor.py +1071 -0
- agentinputlab/integrations/full_architecture/__init__.py +1 -0
- agentinputlab/integrations/full_architecture/hyde.py +212 -0
- agentinputlab/integrations/full_architecture/plan_and_budget.py +188 -0
- agentinputlab/integrations/live_history.py +222 -0
- agentinputlab/integrations/prompt_optimization/__init__.py +1 -0
- agentinputlab/integrations/training/__init__.py +1 -0
- agentinputlab/metrics/__init__.py +50 -0
- agentinputlab/metrics/catalog.py +313 -0
- agentinputlab/metrics/harm.py +95 -0
- agentinputlab/metrics/paired.py +40 -0
- agentinputlab/metrics/qa.py +76 -0
- agentinputlab/metrics/text.py +33 -0
- agentinputlab/metrics/third_party.py +161 -0
- agentinputlab/models/__init__.py +4 -0
- agentinputlab/observability/__init__.py +56 -0
- agentinputlab/observability/budget.py +316 -0
- agentinputlab/observability/instrumentation.py +754 -0
- agentinputlab/observability/models.py +403 -0
- agentinputlab/observability/monitor.py +380 -0
- agentinputlab/optimizers/__init__.py +1 -0
- agentinputlab/optimizers/clarify/__init__.py +12 -0
- agentinputlab/optimizers/context/__init__.py +22 -0
- agentinputlab/optimizers/decompose/__init__.py +64 -0
- agentinputlab/optimizers/evidence/__init__.py +22 -0
- agentinputlab/optimizers/rewrite/__init__.py +30 -0
- agentinputlab/parity/__init__.py +5 -0
- agentinputlab/parity/base.py +206 -0
- agentinputlab/production_campaign.py +763 -0
- agentinputlab/py.typed +1 -0
- agentinputlab/recipes/__init__.py +333 -0
- agentinputlab/recipes/ace.py +418 -0
- agentinputlab/recipes/ace_bfcl.py +273 -0
- agentinputlab/recipes/ace_common_support.py +169 -0
- agentinputlab/recipes/ace_native_plan.py +322 -0
- agentinputlab/recipes/ace_online.py +711 -0
- agentinputlab/recipes/ace_retrieval.py +330 -0
- agentinputlab/recipes/ace_statebench.py +382 -0
- agentinputlab/recipes/ace_tau.py +463 -0
- agentinputlab/recipes/ace_toolsandbox.py +432 -0
- agentinputlab/recipes/acon_guideline.py +22 -0
- agentinputlab/recipes/adaptive_rag.py +822 -0
- agentinputlab/recipes/llmlingua2_inference.py +146 -0
- agentinputlab/recipes/llmlingua2_training.py +796 -0
- agentinputlab/recipes/mce.py +1486 -0
- agentinputlab/recipes/mce_final.py +335 -0
- agentinputlab/recipes/mce_text_fit.py +499 -0
- agentinputlab/recipes/mce_text_only.py +298 -0
- agentinputlab/recipes/qrag.py +729 -0
- agentinputlab/recipes/qrag_inference.py +364 -0
- agentinputlab/recipes/r2ag_features.py +236 -0
- agentinputlab/recipes/r2ag_inference.py +503 -0
- agentinputlab/recipes/r2ag_training.py +948 -0
- agentinputlab/recipes/recomp_abstractive.py +592 -0
- agentinputlab/recipes/recomp_extractive.py +525 -0
- agentinputlab/recipes/reformer.py +17 -0
- agentinputlab/recipes/rqrag.py +1516 -0
- agentinputlab/recipes/rqrag_common_support.py +193 -0
- agentinputlab/recipes/rqrag_completion.py +147 -0
- agentinputlab/recipes/skillopt.py +462 -0
- agentinputlab/recipes/skillopt_common_support.py +154 -0
- agentinputlab/recipes/skillopt_environment.py +165 -0
- agentinputlab/recipes/skillopt_environment_cli.py +5 -0
- agentinputlab/recipes/skillopt_retrieval.py +188 -0
- agentinputlab/recipes/smartsearch.py +897 -0
- agentinputlab/recipes/smartsearch_curriculum.py +283 -0
- agentinputlab/recipes/smartsearch_grpo_contracts.py +58 -0
- agentinputlab/recipes/smartsearch_grpo_data.py +314 -0
- agentinputlab/recipes/smartsearch_grpo_meter.py +324 -0
- agentinputlab/recipes/smartsearch_grpo_prefix.py +144 -0
- agentinputlab/recipes/smartsearch_grpo_rollout.py +119 -0
- agentinputlab/recipes/smartsearch_grpo_tools.py +364 -0
- agentinputlab/recipes/smartsearch_grpo_training.py +305 -0
- agentinputlab/recipes/smartsearch_grpo_transport.py +248 -0
- agentinputlab/recipes/smartsearch_grpo_worker.py +364 -0
- agentinputlab/recipes/smartsearch_inference.py +524 -0
- agentinputlab/recipes/smartsearch_native_data.py +172 -0
- agentinputlab/recipes/smartsearch_policy.py +616 -0
- agentinputlab/recipes/smartsearch_policy_worker.py +9 -0
- agentinputlab/recipes/smartsearch_sampling.py +232 -0
- agentinputlab/recipes/smartsearch_teacher.py +203 -0
- agentinputlab/recipes/smartsearch_training.py +265 -0
- agentinputlab/registry/__init__.py +46 -0
- agentinputlab/registry/beta.py +727 -0
- agentinputlab/registry/data/algorithms.json +1959 -0
- agentinputlab/registry/data/auxiliary_method_profiles.json +34 -0
- agentinputlab/registry/loader.py +100 -0
- agentinputlab/registry/result_manifest.py +184 -0
- agentinputlab/registry/runtime.py +195 -0
- agentinputlab/registry/schema.py +449 -0
- agentinputlab/runner/__init__.py +30 -0
- agentinputlab/runner/balance.py +248 -0
- agentinputlab/runner/cache.py +135 -0
- agentinputlab/runner/composite.py +130 -0
- agentinputlab/runner/engine.py +285 -0
- agentinputlab/runner/runtime_guard.py +268 -0
- agentinputlab/runner/worker.py +80 -0
- agentinputlab/runtime_bootstrap.py +288 -0
- agentinputlab/selectors/__init__.py +196 -0
- agentinputlab/selectors/advanced.py +1514 -0
- agentinputlab/selectors/artifacts.py +1087 -0
- agentinputlab/selectors/base.py +112 -0
- agentinputlab/selectors/campaigns.py +1893 -0
- agentinputlab/selectors/catalog.py +86 -0
- agentinputlab/selectors/discovery.py +1860 -0
- agentinputlab/selectors/evaluation.py +1348 -0
- agentinputlab/selectors/features.py +383 -0
- agentinputlab/selectors/learned.py +502 -0
- agentinputlab/selectors/no_op.py +28 -0
- agentinputlab/selectors/outcome_table.py +475 -0
- agentinputlab/selectors/producer.py +853 -0
- agentinputlab/selectors/rules.py +109 -0
- agentinputlab/selectors/runtime.py +420 -0
- agentinputlab/studies/__init__.py +16 -0
- agentinputlab/studies/design.py +165 -0
- agentinputlab/studies/paired.py +322 -0
- agentinputlab/studio/__init__.py +6 -0
- agentinputlab/studio/campaigns.py +819 -0
- agentinputlab/studio/engine.py +1408 -0
- agentinputlab/studio/model_backends.py +425 -0
- agentinputlab/studio/prepared.py +410 -0
- agentinputlab/studio/server.py +613 -0
- agentinputlab/studio/static/app.js +397 -0
- agentinputlab/studio/static/campaigns.html +175 -0
- agentinputlab/studio/static/campaigns.js +426 -0
- agentinputlab/studio/static/index.html +192 -0
- agentinputlab/studio/static/result.html +103 -0
- agentinputlab/studio/static/result.js +273 -0
- agentinputlab/studio/static/styles.css +467 -0
- agentinputlab-0.1.0a1.data/data/share/agentinputlab/CITATION.cff +22 -0
- agentinputlab-0.1.0a1.dist-info/METADATA +392 -0
- agentinputlab-0.1.0a1.dist-info/RECORD +469 -0
- agentinputlab-0.1.0a1.dist-info/WHEEL +5 -0
- agentinputlab-0.1.0a1.dist-info/entry_points.txt +6 -0
- agentinputlab-0.1.0a1.dist-info/licenses/LICENSE +21 -0
- agentinputlab-0.1.0a1.dist-info/licenses/NOTICE +31 -0
- agentinputlab-0.1.0a1.dist-info/licenses/THIRD_PARTY_NOTICES.md +94 -0
- agentinputlab-0.1.0a1.dist-info/top_level.txt +2 -0
- workers/README.md +94 -0
- workers/ace/README.md +261 -0
- workers/ace/__init__.py +1 -0
- workers/ace/compatibility_manifest.json +39 -0
- workers/ace/driver.py +1682 -0
- workers/acon/README.md +215 -0
- workers/acon/__init__.py +5 -0
- workers/acon/compatibility_manifest.json +93 -0
- workers/acon/driver.py +1859 -0
- workers/acon/fitted.py +395 -0
- workers/acon/guideline.py +470 -0
- workers/acon/guideline_source_manifest.json +20 -0
- workers/adaptive_rag/README.md +96 -0
- workers/adaptive_rag/__init__.py +15 -0
- workers/adaptive_rag/compatibility_manifest.json +61 -0
- workers/adaptive_rag/driver.py +1470 -0
- workers/adaptive_rag/fixture_runtime.py +163 -0
- workers/adaptive_rag/runtime.py +252 -0
- workers/ask_or_assume/README.md +124 -0
- workers/ask_or_assume/__init__.py +1 -0
- workers/ask_or_assume/compatibility_manifest.json +36 -0
- workers/ask_or_assume/driver.py +1227 -0
- workers/decomp/README.md +133 -0
- workers/decomp/UPSTREAM_LICENSE.Apache-2.0.txt +201 -0
- workers/decomp/__init__.py +5 -0
- workers/decomp/compatibility_manifest.json +205 -0
- workers/decomp/driver.py +1015 -0
- workers/decomp/live.py +137 -0
- workers/decomp/runtime.py +231 -0
- workers/infocqr/README.md +106 -0
- workers/infocqr/__init__.py +1 -0
- workers/infocqr/compatibility_manifest.json +84 -0
- workers/infocqr/driver.py +1766 -0
- workers/ircot/README.md +167 -0
- workers/ircot/__init__.py +7 -0
- workers/ircot/compatibility_manifest.json +121 -0
- workers/ircot/controlled.py +457 -0
- workers/ircot/driver.py +1373 -0
- workers/ircot/fixture_bindings.json +39 -0
- workers/ircot/fixture_config.json +48 -0
- workers/ircot/fixture_prompt.txt +2 -0
- workers/ircot/native_qa.py +398 -0
- workers/ircot/qa_compatibility_manifest.json +57 -0
- workers/ircot/runtime.py +97 -0
- workers/llmcompiler/README.md +160 -0
- workers/llmcompiler/__init__.py +1 -0
- workers/llmcompiler/compatibility_manifest.json +45 -0
- workers/llmcompiler/driver.py +1514 -0
- workers/llmcompiler/export_prompts.py +74 -0
- workers/llmcompiler/live.py +174 -0
- workers/llmcompiler/patch_manifest.json +147 -0
- workers/mce/README.md +198 -0
- workers/mce/__init__.py +29 -0
- workers/mce/compatibility_manifest.json +115 -0
- workers/mce/driver.py +2585 -0
- workers/mce/final_eval.py +173 -0
- workers/official_arenas/README.md +25 -0
- workers/official_arenas/__init__.py +1 -0
- workers/official_arenas/capture_records.py +110 -0
- workers/official_arenas/driver.py +619 -0
- workers/official_arenas/protocol.py +334 -0
- workers/official_arenas/providers/__init__.py +1 -0
- workers/official_arenas/providers/_live_history.py +228 -0
- workers/official_arenas/providers/_message_compat.py +77 -0
- workers/official_arenas/providers/appworld_ul.py +196 -0
- workers/official_arenas/providers/fixture.py +112 -0
- workers/official_arenas/providers/statebench.py +960 -0
- workers/official_arenas/providers/tau3.py +1672 -0
- workers/official_arenas/providers/tau_retail.py +586 -0
- workers/official_arenas/providers/toolsandbox.py +1321 -0
- workers/official_arenas/providers/tooltalk.py +699 -0
- workers/official_arenas/retail_capture.py +188 -0
- workers/official_arenas/retail_driver.py +275 -0
- workers/qrag/README.md +172 -0
- workers/qrag/__init__.py +1 -0
- workers/qrag/compatibility_manifest.json +64 -0
- workers/qrag/driver.py +673 -0
- workers/r2ag/README.md +154 -0
- workers/r2ag/__init__.py +5 -0
- workers/r2ag/compatibility_manifest.json +55 -0
- workers/r2ag/driver.py +1287 -0
- workers/r2ag/feature_schema.json +8 -0
- workers/r2ag/runtime.py +589 -0
- workers/recomp/README.md +120 -0
- workers/recomp/__init__.py +1 -0
- workers/recomp/compatibility_manifest.json +43 -0
- workers/recomp/driver.py +1653 -0
- workers/reformer/README.md +125 -0
- workers/reformer/__init__.py +1 -0
- workers/reformer/compatibility_manifest.json +68 -0
- workers/reformer/driver.py +1459 -0
- workers/reformer/induction.py +457 -0
- workers/rrr/README.md +126 -0
- workers/rrr/__init__.py +1 -0
- workers/rrr/compatibility_manifest.json +57 -0
- workers/rrr/driver.py +1133 -0
- workers/smartsearch/README.md +331 -0
- workers/smartsearch/__init__.py +1 -0
- workers/smartsearch/compatibility_manifest.json +41 -0
- workers/smartsearch/driver.py +1643 -0
|
@@ -0,0 +1,383 @@
|
|
|
1
|
+
"""Deterministic, evaluator-blind features for selector inference.
|
|
2
|
+
|
|
3
|
+
The extractor in this module deliberately implements a small, transparent
|
|
4
|
+
feature schema. It reads only visible text plus conversation, context, and
|
|
5
|
+
tool-list shape from :class:`FrozenAgentState`. In particular,
|
|
6
|
+
``task_id`` and ``metadata`` are never consulted: both may contain benchmark
|
|
7
|
+
or source identities that are useful for audit but unsafe as selector inputs.
|
|
8
|
+
"""
|
|
9
|
+
|
|
10
|
+
from __future__ import annotations
|
|
11
|
+
|
|
12
|
+
import math
|
|
13
|
+
import re
|
|
14
|
+
from dataclasses import dataclass
|
|
15
|
+
|
|
16
|
+
from agentinputlab.contracts.errors import ValidationError
|
|
17
|
+
from agentinputlab.contracts.serialization import content_hash
|
|
18
|
+
from agentinputlab.contracts.state import FrozenAgentState
|
|
19
|
+
|
|
20
|
+
from .base import SelectionFeatures
|
|
21
|
+
from .runtime import FeatureAccessDomain, FeatureObservation
|
|
22
|
+
|
|
23
|
+
_TOKEN = re.compile(r"[A-Za-z0-9_]+|[\u3400-\u4dbf\u4e00-\u9fff]|[^\W_]+", re.UNICODE)
|
|
24
|
+
_SENTENCE_BOUNDARY = re.compile(r"[.!?\u3002\uff01\uff1f]+")
|
|
25
|
+
_AMBIGUITY_TERMS = frozenset(
|
|
26
|
+
{
|
|
27
|
+
"appropriate",
|
|
28
|
+
"best",
|
|
29
|
+
"better",
|
|
30
|
+
"few",
|
|
31
|
+
"it",
|
|
32
|
+
"maybe",
|
|
33
|
+
"nearby",
|
|
34
|
+
"optimal",
|
|
35
|
+
"reasonable",
|
|
36
|
+
"recent",
|
|
37
|
+
"several",
|
|
38
|
+
"some",
|
|
39
|
+
"something",
|
|
40
|
+
"that",
|
|
41
|
+
"them",
|
|
42
|
+
"there",
|
|
43
|
+
"they",
|
|
44
|
+
"this",
|
|
45
|
+
"those",
|
|
46
|
+
"various",
|
|
47
|
+
"合适",
|
|
48
|
+
"最好",
|
|
49
|
+
"最近",
|
|
50
|
+
"一些",
|
|
51
|
+
"这个",
|
|
52
|
+
"那个",
|
|
53
|
+
"它",
|
|
54
|
+
"他们",
|
|
55
|
+
}
|
|
56
|
+
)
|
|
57
|
+
_MULTI_STEP_TERMS = frozenset(
|
|
58
|
+
{
|
|
59
|
+
"after",
|
|
60
|
+
"and",
|
|
61
|
+
"before",
|
|
62
|
+
"compare",
|
|
63
|
+
"contrast",
|
|
64
|
+
"finally",
|
|
65
|
+
"first",
|
|
66
|
+
"next",
|
|
67
|
+
"then",
|
|
68
|
+
"versus",
|
|
69
|
+
"vs",
|
|
70
|
+
"之后",
|
|
71
|
+
"以及",
|
|
72
|
+
"并且",
|
|
73
|
+
"然后",
|
|
74
|
+
"比较",
|
|
75
|
+
"首先",
|
|
76
|
+
}
|
|
77
|
+
)
|
|
78
|
+
_RETRIEVAL_TERMS = frozenset(
|
|
79
|
+
{
|
|
80
|
+
"according",
|
|
81
|
+
"cite",
|
|
82
|
+
"citation",
|
|
83
|
+
"current",
|
|
84
|
+
"evidence",
|
|
85
|
+
"latest",
|
|
86
|
+
"paper",
|
|
87
|
+
"reference",
|
|
88
|
+
"report",
|
|
89
|
+
"research",
|
|
90
|
+
"source",
|
|
91
|
+
"statistics",
|
|
92
|
+
"study",
|
|
93
|
+
"today",
|
|
94
|
+
"verify",
|
|
95
|
+
"when",
|
|
96
|
+
"where",
|
|
97
|
+
"who",
|
|
98
|
+
"year",
|
|
99
|
+
"今天",
|
|
100
|
+
"何时",
|
|
101
|
+
"哪里",
|
|
102
|
+
"引用",
|
|
103
|
+
"数据",
|
|
104
|
+
"最新",
|
|
105
|
+
"来源",
|
|
106
|
+
"研究",
|
|
107
|
+
"论文",
|
|
108
|
+
"证据",
|
|
109
|
+
}
|
|
110
|
+
)
|
|
111
|
+
_QUESTION_TERMS = frozenset(
|
|
112
|
+
{
|
|
113
|
+
"how",
|
|
114
|
+
"what",
|
|
115
|
+
"when",
|
|
116
|
+
"where",
|
|
117
|
+
"which",
|
|
118
|
+
"who",
|
|
119
|
+
"why",
|
|
120
|
+
"什么",
|
|
121
|
+
"为什么",
|
|
122
|
+
"何时",
|
|
123
|
+
"哪里",
|
|
124
|
+
"如何",
|
|
125
|
+
"谁",
|
|
126
|
+
}
|
|
127
|
+
)
|
|
128
|
+
|
|
129
|
+
|
|
130
|
+
@dataclass(frozen=True, slots=True)
|
|
131
|
+
class VisibleFeatureConfig:
|
|
132
|
+
"""Frozen configuration for the five-family visible feature schema."""
|
|
133
|
+
|
|
134
|
+
schema_id: str = "family_defect_features.v1"
|
|
135
|
+
tokenization_revision: str = "unicode_word_cjk_v1"
|
|
136
|
+
context_soft_limit_tokens: int = 8192
|
|
137
|
+
|
|
138
|
+
def __post_init__(self) -> None:
|
|
139
|
+
for label, value in (
|
|
140
|
+
("schema_id", self.schema_id),
|
|
141
|
+
("tokenization_revision", self.tokenization_revision),
|
|
142
|
+
):
|
|
143
|
+
if not isinstance(value, str) or not value.strip():
|
|
144
|
+
raise ValidationError(f"visible feature {label} must be non-empty")
|
|
145
|
+
if (
|
|
146
|
+
isinstance(self.context_soft_limit_tokens, bool)
|
|
147
|
+
or not isinstance(self.context_soft_limit_tokens, int)
|
|
148
|
+
or self.context_soft_limit_tokens < 1
|
|
149
|
+
):
|
|
150
|
+
raise ValidationError("context_soft_limit_tokens must be a positive integer")
|
|
151
|
+
|
|
152
|
+
@property
|
|
153
|
+
def config_hash(self) -> str:
|
|
154
|
+
return content_hash(self)
|
|
155
|
+
|
|
156
|
+
|
|
157
|
+
@dataclass(frozen=True, slots=True)
|
|
158
|
+
class VisibleStateFeatureExtractor:
|
|
159
|
+
"""Extract five normalized scores without consulting identities or metadata."""
|
|
160
|
+
|
|
161
|
+
config: VisibleFeatureConfig = VisibleFeatureConfig()
|
|
162
|
+
extractor_id: str = "visible_state_heuristic"
|
|
163
|
+
extractor_revision: str = "v1"
|
|
164
|
+
|
|
165
|
+
def __post_init__(self) -> None:
|
|
166
|
+
if not isinstance(self.config, VisibleFeatureConfig):
|
|
167
|
+
raise ValidationError("visible feature extractor requires VisibleFeatureConfig")
|
|
168
|
+
for label, value in (
|
|
169
|
+
("extractor_id", self.extractor_id),
|
|
170
|
+
("extractor_revision", self.extractor_revision),
|
|
171
|
+
):
|
|
172
|
+
if not isinstance(value, str) or not value.strip():
|
|
173
|
+
raise ValidationError(f"feature extractor {label} must be non-empty")
|
|
174
|
+
|
|
175
|
+
@property
|
|
176
|
+
def extractor_hash(self) -> str:
|
|
177
|
+
return content_hash(
|
|
178
|
+
{
|
|
179
|
+
"extractor_id": self.extractor_id,
|
|
180
|
+
"extractor_revision": self.extractor_revision,
|
|
181
|
+
"config": self.config,
|
|
182
|
+
}
|
|
183
|
+
)
|
|
184
|
+
|
|
185
|
+
@property
|
|
186
|
+
def feature_schema_id(self) -> str:
|
|
187
|
+
return self.config.schema_id
|
|
188
|
+
|
|
189
|
+
@property
|
|
190
|
+
def feature_schema_hash(self) -> str:
|
|
191
|
+
return content_hash(
|
|
192
|
+
{
|
|
193
|
+
"schema_id": self.config.schema_id,
|
|
194
|
+
"ordered_features": (
|
|
195
|
+
"ambiguity_score",
|
|
196
|
+
"multi_step_score",
|
|
197
|
+
"retrieval_need_score",
|
|
198
|
+
"context_pressure",
|
|
199
|
+
"reformulation_need_score",
|
|
200
|
+
),
|
|
201
|
+
"tokenization_revision": self.config.tokenization_revision,
|
|
202
|
+
}
|
|
203
|
+
)
|
|
204
|
+
|
|
205
|
+
def extract(self, state: FrozenAgentState) -> FeatureObservation:
|
|
206
|
+
if not isinstance(state, FrozenAgentState):
|
|
207
|
+
raise TypeError("visible feature extraction requires FrozenAgentState")
|
|
208
|
+
|
|
209
|
+
query_tokens = _tokens(state.query)
|
|
210
|
+
instruction_tokens = _tokens(state.instruction)
|
|
211
|
+
message_tokens = tuple(
|
|
212
|
+
token for message in state.messages for token in _tokens(message.content)
|
|
213
|
+
)
|
|
214
|
+
context_tokens = tuple(token for block in state.context for token in _tokens(block.text))
|
|
215
|
+
visible_tokens = (*query_tokens, *instruction_tokens, *message_tokens, *context_tokens)
|
|
216
|
+
request_text = "\n".join(part for part in (state.instruction, state.query) if part)
|
|
217
|
+
request_lower = request_text.casefold()
|
|
218
|
+
request_tokens = (*instruction_tokens, *query_tokens)
|
|
219
|
+
request_term_set = frozenset(request_tokens)
|
|
220
|
+
|
|
221
|
+
ambiguity_hits = _term_hits(request_lower, request_term_set, _AMBIGUITY_TERMS)
|
|
222
|
+
dangling_reference = int(
|
|
223
|
+
bool(request_term_set & {"it", "that", "them", "there", "they", "this", "those"})
|
|
224
|
+
and not state.messages
|
|
225
|
+
and not state.context
|
|
226
|
+
)
|
|
227
|
+
alternative_signal = int(
|
|
228
|
+
" or " in f" {request_lower} "
|
|
229
|
+
or "either" in request_term_set
|
|
230
|
+
or "或者" in request_text
|
|
231
|
+
or "还是" in request_text
|
|
232
|
+
)
|
|
233
|
+
ambiguity = _saturating(
|
|
234
|
+
ambiguity_hits + dangling_reference + 0.5 * alternative_signal,
|
|
235
|
+
scale=3.0,
|
|
236
|
+
)
|
|
237
|
+
|
|
238
|
+
multi_step_hits = _term_hits(
|
|
239
|
+
request_lower,
|
|
240
|
+
request_term_set,
|
|
241
|
+
_MULTI_STEP_TERMS,
|
|
242
|
+
)
|
|
243
|
+
sentence_count = len(_SENTENCE_BOUNDARY.findall(request_text))
|
|
244
|
+
list_signal = int(bool(re.search(r"(?:^|\s)(?:\d+[.)]|[-*])\s", request_text)))
|
|
245
|
+
multi_step = _saturating(
|
|
246
|
+
multi_step_hits + max(0, sentence_count - 1) + list_signal,
|
|
247
|
+
scale=4.0,
|
|
248
|
+
)
|
|
249
|
+
|
|
250
|
+
retrieval_hits = _term_hits(
|
|
251
|
+
request_lower,
|
|
252
|
+
request_term_set,
|
|
253
|
+
_RETRIEVAL_TERMS,
|
|
254
|
+
)
|
|
255
|
+
interrogative = int(
|
|
256
|
+
bool(_term_hits(request_lower, request_term_set, _QUESTION_TERMS))
|
|
257
|
+
or "?" in request_text
|
|
258
|
+
or "\uff1f" in request_text
|
|
259
|
+
)
|
|
260
|
+
context_relief = min(0.6, len(context_tokens) / 256.0)
|
|
261
|
+
retrieval = _clamp(
|
|
262
|
+
_saturating(retrieval_hits + 0.5 * interrogative, scale=3.0) - context_relief
|
|
263
|
+
)
|
|
264
|
+
|
|
265
|
+
visible_token_count = len(visible_tokens)
|
|
266
|
+
context_pressure = _clamp(
|
|
267
|
+
visible_token_count / float(self.config.context_soft_limit_tokens)
|
|
268
|
+
)
|
|
269
|
+
|
|
270
|
+
nonspace = [character for character in request_text if not character.isspace()]
|
|
271
|
+
noisy = sum(
|
|
272
|
+
not character.isalnum()
|
|
273
|
+
and character
|
|
274
|
+
not in {
|
|
275
|
+
"?",
|
|
276
|
+
"!",
|
|
277
|
+
".",
|
|
278
|
+
",",
|
|
279
|
+
":",
|
|
280
|
+
";",
|
|
281
|
+
"-",
|
|
282
|
+
"_",
|
|
283
|
+
"'",
|
|
284
|
+
'"',
|
|
285
|
+
"\u3001",
|
|
286
|
+
"\u3002",
|
|
287
|
+
"\uff01",
|
|
288
|
+
"\uff08",
|
|
289
|
+
"\uff09",
|
|
290
|
+
"\uff0c",
|
|
291
|
+
"\uff1a",
|
|
292
|
+
"\uff1b",
|
|
293
|
+
"\uff1f",
|
|
294
|
+
}
|
|
295
|
+
for character in nonspace
|
|
296
|
+
)
|
|
297
|
+
noise_ratio = noisy / max(1, len(nonspace))
|
|
298
|
+
fragment = int(0 < len(request_tokens) <= 3 and not _SENTENCE_BOUNDARY.search(request_text))
|
|
299
|
+
extreme_length = int(len(request_tokens) >= 64)
|
|
300
|
+
keyword_style = int(
|
|
301
|
+
len(request_tokens) >= 4
|
|
302
|
+
and not _SENTENCE_BOUNDARY.search(request_text)
|
|
303
|
+
and not _term_hits(request_lower, request_term_set, _QUESTION_TERMS)
|
|
304
|
+
)
|
|
305
|
+
reformulation = _clamp(
|
|
306
|
+
0.45 * fragment
|
|
307
|
+
+ 0.35 * keyword_style
|
|
308
|
+
+ 0.35 * extreme_length
|
|
309
|
+
+ min(0.5, noise_ratio * 2.0)
|
|
310
|
+
)
|
|
311
|
+
|
|
312
|
+
features = SelectionFeatures(
|
|
313
|
+
ambiguity_score=ambiguity,
|
|
314
|
+
multi_step_score=multi_step,
|
|
315
|
+
retrieval_need_score=retrieval,
|
|
316
|
+
context_pressure=context_pressure,
|
|
317
|
+
reformulation_need_score=reformulation,
|
|
318
|
+
)
|
|
319
|
+
provenance = {
|
|
320
|
+
"schema_id": self.config.schema_id,
|
|
321
|
+
"schema_hash": self.feature_schema_hash,
|
|
322
|
+
"extractor_hash": self.extractor_hash,
|
|
323
|
+
"tokenization_revision": self.config.tokenization_revision,
|
|
324
|
+
"input_fields": (
|
|
325
|
+
"query",
|
|
326
|
+
"instruction",
|
|
327
|
+
"context.text",
|
|
328
|
+
"messages.content",
|
|
329
|
+
"tool_names.count",
|
|
330
|
+
),
|
|
331
|
+
"ignored_fields": ("task_id", "metadata"),
|
|
332
|
+
"counts": {
|
|
333
|
+
"query_tokens": len(query_tokens),
|
|
334
|
+
"instruction_tokens": len(instruction_tokens),
|
|
335
|
+
"message_tokens": len(message_tokens),
|
|
336
|
+
"context_tokens": len(context_tokens),
|
|
337
|
+
"messages": len(state.messages),
|
|
338
|
+
"context_blocks": len(state.context),
|
|
339
|
+
"tools": len(state.tool_names),
|
|
340
|
+
},
|
|
341
|
+
"signals": {
|
|
342
|
+
"ambiguity": ambiguity_hits + dangling_reference + alternative_signal,
|
|
343
|
+
"multi_step": multi_step_hits + max(0, sentence_count - 1) + list_signal,
|
|
344
|
+
"retrieval": retrieval_hits + interrogative,
|
|
345
|
+
"reformulation": fragment + keyword_style + extreme_length + int(noisy > 0),
|
|
346
|
+
},
|
|
347
|
+
}
|
|
348
|
+
return FeatureObservation(
|
|
349
|
+
features=features,
|
|
350
|
+
state_hash=state.state_hash,
|
|
351
|
+
extractor_id=self.extractor_id,
|
|
352
|
+
extractor_revision=self.extractor_revision,
|
|
353
|
+
access_domains=(FeatureAccessDomain.VISIBLE_PRE_ACTION,),
|
|
354
|
+
provenance=provenance,
|
|
355
|
+
)
|
|
356
|
+
|
|
357
|
+
|
|
358
|
+
def _tokens(text: str) -> tuple[str, ...]:
|
|
359
|
+
return tuple(match.group(0).casefold() for match in _TOKEN.finditer(text))
|
|
360
|
+
|
|
361
|
+
|
|
362
|
+
def _term_hits(
|
|
363
|
+
text: str,
|
|
364
|
+
tokens: frozenset[str],
|
|
365
|
+
terms: frozenset[str],
|
|
366
|
+
) -> int:
|
|
367
|
+
return sum((term in tokens) if term.isascii() else (term in text) for term in terms)
|
|
368
|
+
|
|
369
|
+
|
|
370
|
+
def _saturating(value: float, *, scale: float) -> float:
|
|
371
|
+
if not math.isfinite(value) or value < 0:
|
|
372
|
+
raise ValueError("feature signal counts must be finite and non-negative")
|
|
373
|
+
return _clamp(value / scale)
|
|
374
|
+
|
|
375
|
+
|
|
376
|
+
def _clamp(value: float) -> float:
|
|
377
|
+
return max(0.0, min(1.0, float(value)))
|
|
378
|
+
|
|
379
|
+
|
|
380
|
+
__all__ = [
|
|
381
|
+
"VisibleFeatureConfig",
|
|
382
|
+
"VisibleStateFeatureExtractor",
|
|
383
|
+
]
|