agent-learning-kit 0.1.0__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.
- agent_learning_kit-0.1.0.dist-info/METADATA +381 -0
- agent_learning_kit-0.1.0.dist-info/RECORD +642 -0
- agent_learning_kit-0.1.0.dist-info/WHEEL +4 -0
- agent_learning_kit-0.1.0.dist-info/entry_points.txt +5 -0
- agent_learning_kit-0.1.0.dist-info/licenses/LICENSE +173 -0
- agent_learning_kit-0.1.0.dist-info/licenses/NOTICE +7 -0
- fi/__init__.py +5 -0
- fi/alk/__init__.py +57 -0
- fi/alk/_facade.py +31 -0
- fi/alk/_module_alias.py +68 -0
- fi/alk/_paths.py +14 -0
- fi/alk/_schema.py +522 -0
- fi/alk/actions.py +727 -0
- fi/alk/bench/__init__.py +517 -0
- fi/alk/bench/_codeexec.py +213 -0
- fi/alk/bench/_coding.py +215 -0
- fi/alk/bench/_docker.py +237 -0
- fi/alk/bench/_grader.py +286 -0
- fi/alk/bench/_pull.py +212 -0
- fi/alk/bench/_voice.py +147 -0
- fi/alk/capabilities.py +627 -0
- fi/alk/cli.py +6396 -0
- fi/alk/config.py +130 -0
- fi/alk/cua_loop.py +562 -0
- fi/alk/evals.py +2351 -0
- fi/alk/extensions.py +163 -0
- fi/alk/harness/ARCHITECTURE.md +231 -0
- fi/alk/harness/DESIGN.md +246 -0
- fi/alk/harness/ENVIRONMENT_CONFORMANCE.md +127 -0
- fi/alk/harness/HOW-IT-WORKS.md +297 -0
- fi/alk/harness/IMPLEMENTATION_AND_VALIDATION_STATUS.md +229 -0
- fi/alk/harness/README.md +417 -0
- fi/alk/harness/__init__.py +77 -0
- fi/alk/harness/__main__.py +3 -0
- fi/alk/harness/amend.py +312 -0
- fi/alk/harness/artifacts.py +319 -0
- fi/alk/harness/authoring_entrypoint.py +189 -0
- fi/alk/harness/authoring_runtime_validation.py +267 -0
- fi/alk/harness/backends/README.md +43 -0
- fi/alk/harness/backends/__init__.py +122 -0
- fi/alk/harness/backends/base.py +241 -0
- fi/alk/harness/backends/claude.py +211 -0
- fi/alk/harness/backends/files.py +182 -0
- fi/alk/harness/backends/vertex_gemini.py +457 -0
- fi/alk/harness/background_noise.py +95 -0
- fi/alk/harness/build.py +385 -0
- fi/alk/harness/bundle.py +593 -0
- fi/alk/harness/bundle_author_v2.py +1831 -0
- fi/alk/harness/bundle_v2.py +719 -0
- fi/alk/harness/call_runner.py +1440 -0
- fi/alk/harness/callback_http_adapter.py +111 -0
- fi/alk/harness/catalogue.py +287 -0
- fi/alk/harness/chat.py +428 -0
- fi/alk/harness/chat_call_runner.py +506 -0
- fi/alk/harness/checks.py +136 -0
- fi/alk/harness/cli.py +1354 -0
- fi/alk/harness/config.py +338 -0
- fi/alk/harness/contract.py +718 -0
- fi/alk/harness/credentials.py +674 -0
- fi/alk/harness/data/persona_vocabulary.json +111 -0
- fi/alk/harness/environment.py +99 -0
- fi/alk/harness/environment_plan.py +168 -0
- fi/alk/harness/events.py +125 -0
- fi/alk/harness/executor.py +304 -0
- fi/alk/harness/folder.py +234 -0
- fi/alk/harness/generated_runtime.py +815 -0
- fi/alk/harness/github.py +72 -0
- fi/alk/harness/hosted_authoring_entrypoint.py +183 -0
- fi/alk/harness/hosted_entrypoint.py +2402 -0
- fi/alk/harness/hosted_scheduler.py +2218 -0
- fi/alk/harness/job.py +426 -0
- fi/alk/harness/judge.py +184 -0
- fi/alk/harness/livekit_source.py +50 -0
- fi/alk/harness/livekit_tool_trace_bootstrap.py +71 -0
- fi/alk/harness/observability.py +208 -0
- fi/alk/harness/outbound.py +3252 -0
- fi/alk/harness/packaging.py +515 -0
- fi/alk/harness/persona_guides.py +157 -0
- fi/alk/harness/platform.py +692 -0
- fi/alk/harness/process_preflight.py +764 -0
- fi/alk/harness/process_runtime.py +5670 -0
- fi/alk/harness/prove.py +425 -0
- fi/alk/harness/provider_import.py +703 -0
- fi/alk/harness/provider_lifecycle.py +392 -0
- fi/alk/harness/provision.py +2896 -0
- fi/alk/harness/reception.py +147 -0
- fi/alk/harness/retell_chat_call_runner.py +373 -0
- fi/alk/harness/run/__init__.py +296 -0
- fi/alk/harness/run/alk.py +184 -0
- fi/alk/harness/run/call.py +162 -0
- fi/alk/harness/run/conversation.py +264 -0
- fi/alk/harness/run/data/voices_by_language_and_gender.json +693 -0
- fi/alk/harness/run/evidence.py +195 -0
- fi/alk/harness/run/grade.py +598 -0
- fi/alk/harness/run/live.py +297 -0
- fi/alk/harness/run/models.py +56 -0
- fi/alk/harness/run/platform_evals.py +227 -0
- fi/alk/harness/run/sdk_voice.py +130 -0
- fi/alk/harness/run/simulation.py +1209 -0
- fi/alk/harness/run/stage.py +91 -0
- fi/alk/harness/run/targets.py +508 -0
- fi/alk/harness/run/tools.py +601 -0
- fi/alk/harness/run/voice.py +340 -0
- fi/alk/harness/runtime.py +172 -0
- fi/alk/harness/sandbox_server.py +2011 -0
- fi/alk/harness/sandbox_worker.py +44 -0
- fi/alk/harness/scenario.py +1048 -0
- fi/alk/harness/scenario_source.py +879 -0
- fi/alk/harness/scenario_tools.py +1143 -0
- fi/alk/harness/scenarios.py +915 -0
- fi/alk/harness/secrets.py +168 -0
- fi/alk/harness/service_catalog.py +97 -0
- fi/alk/harness/session.py +391 -0
- fi/alk/harness/sessions.py +372 -0
- fi/alk/harness/simulator.py +76 -0
- fi/alk/harness/simulator_voice.py +928 -0
- fi/alk/harness/skills/build-environment/SKILL.md +538 -0
- fi/alk/harness/skills/harness.md +131 -0
- fi/alk/harness/skills/kinds/chat.md +48 -0
- fi/alk/harness/skills/kinds/voice-voicemail.md +63 -0
- fi/alk/harness/skills/kinds/voice.md +59 -0
- fi/alk/harness/skills/plan-suite/SKILL.md +103 -0
- fi/alk/harness/skills/provision-environment/SKILL.md +136 -0
- fi/alk/harness/skills/run-scenarios/SKILL.md +112 -0
- fi/alk/harness/skills/understand-agent/SKILL.md +251 -0
- fi/alk/harness/skills/write-scenarios/SKILL.md +606 -0
- fi/alk/harness/skills/write-scenarios/references/refusals.md +28 -0
- fi/alk/harness/skills/write-scenarios/references/world-api.md +92 -0
- fi/alk/harness/source_data_invariants.py +444 -0
- fi/alk/harness/source_tool_evidence.py +79 -0
- fi/alk/harness/sources.py +253 -0
- fi/alk/harness/spend.py +140 -0
- fi/alk/harness/tool_trace_proxy.py +104 -0
- fi/alk/harness/tools.py +1018 -0
- fi/alk/harness/understand.py +169 -0
- fi/alk/harness/voicemail_audio.py +74 -0
- fi/alk/harness/world/__init__.py +33 -0
- fi/alk/harness/world/errors.py +68 -0
- fi/alk/harness/world/expectations.py +91 -0
- fi/alk/harness/world/handle.py +538 -0
- fi/alk/harness/world/kinds.py +196 -0
- fi/alk/harness/world/mutate.py +186 -0
- fi/alk/harness/world/probe.py +413 -0
- fi/alk/harness/world/provision.py +511 -0
- fi/alk/harness/world/provisioned.py +191 -0
- fi/alk/harness/world/runtime.py +616 -0
- fi/alk/harness/world/snapshot.py +288 -0
- fi/alk/harness/world/stores/__init__.py +305 -0
- fi/alk/harness/world/stores/container.py +215 -0
- fi/alk/harness/world/stores/inprocess.py +346 -0
- fi/alk/harness/world/stores/postgres.py +481 -0
- fi/alk/harness/world/stores/prove.py +202 -0
- fi/alk/harness/world/stores/sqlite.py +245 -0
- fi/alk/harness/world/stores/written.py +182 -0
- fi/alk/harness/world/tools.py +1516 -0
- fi/alk/harness/world/workspace.py +144 -0
- fi/alk/image_loop.py +453 -0
- fi/alk/image_perturb.py +241 -0
- fi/alk/improve.py +274 -0
- fi/alk/live/__init__.py +154 -0
- fi/alk/live/_attribution.py +184 -0
- fi/alk/live/_capture.py +264 -0
- fi/alk/live/_codec.py +391 -0
- fi/alk/live/_contract.py +134 -0
- fi/alk/live/_loopback.py +316 -0
- fi/alk/live/_perturb.py +449 -0
- fi/alk/live/_runner.py +386 -0
- fi/alk/live/_stats.py +561 -0
- fi/alk/live/_transcript.py +240 -0
- fi/alk/live/_workers/__init__.py +9 -0
- fi/alk/live/_workers/a2a_worker.py +316 -0
- fi/alk/live/_workers/langgraph_worker.py +217 -0
- fi/alk/live/_workers/livekit_worker.py +207 -0
- fi/alk/live/_workers/mcp_loopback_server.py +46 -0
- fi/alk/live/_workers/mcp_worker.py +158 -0
- fi/alk/live/_workers/pipecat_worker.py +189 -0
- fi/alk/live/a2a_lane.py +138 -0
- fi/alk/live/langgraph_lane.py +339 -0
- fi/alk/live/livekit_lane.py +376 -0
- fi/alk/live/mcp_lane.py +172 -0
- fi/alk/live/pipecat_lane.py +341 -0
- fi/alk/live/voice_redteam.py +494 -0
- fi/alk/loss.py +306 -0
- fi/alk/optimize.py +36260 -0
- fi/alk/practice/__init__.py +51 -0
- fi/alk/practice/_assess.py +103 -0
- fi/alk/practice/_budget.py +81 -0
- fi/alk/practice/_calibrate.py +69 -0
- fi/alk/practice/_capstone.py +86 -0
- fi/alk/practice/_contract.py +91 -0
- fi/alk/practice/_diagnose.py +79 -0
- fi/alk/practice/_drill.py +196 -0
- fi/alk/practice/_experiment.py +720 -0
- fi/alk/practice/_schedule.py +102 -0
- fi/alk/practice/_store.py +194 -0
- fi/alk/practice/_trainer.py +245 -0
- fi/alk/practice/_update.py +125 -0
- fi/alk/redteam.py +2621 -0
- fi/alk/rewardhack.py +237 -0
- fi/alk/simulate.py +10351 -0
- fi/alk/studio/__init__.py +82 -0
- fi/alk/studio/_bias.py +314 -0
- fi/alk/studio/_calibration.py +522 -0
- fi/alk/studio/_coverage.py +262 -0
- fi/alk/studio/_download.py +665 -0
- fi/alk/studio/_fidelity_attack.py +114 -0
- fi/alk/studio/_generate.py +652 -0
- fi/alk/studio/_library.py +370 -0
- fi/alk/studio/_scan.py +134 -0
- fi/alk/studio/_upgrade.py +42 -0
- fi/alk/studio/_vendor.py +172 -0
- fi/alk/suite.py +4200 -0
- fi/alk/tasks.py +828 -0
- fi/alk/telemetry/__init__.py +149 -0
- fi/alk/telemetry/_contract.py +141 -0
- fi/alk/telemetry/_emit.py +182 -0
- fi/alk/telemetry/_ledger.py +296 -0
- fi/alk/telemetry/_queue.py +127 -0
- fi/alk/telemetry/_row.py +294 -0
- fi/alk/telemetry/_run.py +233 -0
- fi/alk/telemetry/_sync.py +193 -0
- fi/alk/telemetry/_url.py +119 -0
- fi/alk/trinity.py +49397 -0
- fi/alk/voice_loop.py +174 -0
- fi/api/__init__.py +1 -0
- fi/api/auth.py +137 -0
- fi/api/types.py +29 -0
- fi/cli/__init__.py +9 -0
- fi/cli/assertions/__init__.py +25 -0
- fi/cli/assertions/conditions.py +76 -0
- fi/cli/assertions/evaluator.py +286 -0
- fi/cli/assertions/exit_codes.py +20 -0
- fi/cli/assertions/parser.py +131 -0
- fi/cli/assertions/reporter.py +194 -0
- fi/cli/commands/__init__.py +9 -0
- fi/cli/commands/config.py +165 -0
- fi/cli/commands/export.py +208 -0
- fi/cli/commands/init.py +112 -0
- fi/cli/commands/list_cmd.py +213 -0
- fi/cli/commands/run.py +486 -0
- fi/cli/commands/validate.py +173 -0
- fi/cli/commands/view.py +424 -0
- fi/cli/config/__init__.py +6 -0
- fi/cli/config/defaults.py +206 -0
- fi/cli/config/loader.py +155 -0
- fi/cli/config/schema.py +174 -0
- fi/cli/main.py +78 -0
- fi/cli/output/__init__.py +6 -0
- fi/cli/output/formatters.py +106 -0
- fi/cli/output/reporters.py +46 -0
- fi/cli/storage/__init__.py +5 -0
- fi/cli/storage/run_history.py +249 -0
- fi/cli/utils/__init__.py +5 -0
- fi/cli/utils/console.py +44 -0
- fi/evals/__init__.py +131 -0
- fi/evals/autoeval/__init__.py +137 -0
- fi/evals/autoeval/analyzer.py +211 -0
- fi/evals/autoeval/config.py +244 -0
- fi/evals/autoeval/export.py +213 -0
- fi/evals/autoeval/interactive.py +283 -0
- fi/evals/autoeval/pipeline.py +625 -0
- fi/evals/autoeval/prompts.py +139 -0
- fi/evals/autoeval/recommender.py +242 -0
- fi/evals/autoeval/rules.py +589 -0
- fi/evals/autoeval/templates.py +299 -0
- fi/evals/autoeval/types.py +232 -0
- fi/evals/core/__init__.py +16 -0
- fi/evals/core/cloud_registry.py +184 -0
- fi/evals/core/engines.py +368 -0
- fi/evals/core/evaluate.py +319 -0
- fi/evals/core/judge_prompt.py +90 -0
- fi/evals/core/prompt_generator.py +83 -0
- fi/evals/core/registry.py +57 -0
- fi/evals/core/result.py +55 -0
- fi/evals/evaluator.py +721 -0
- fi/evals/execution.py +168 -0
- fi/evals/feedback/__init__.py +32 -0
- fi/evals/feedback/calibrator.py +160 -0
- fi/evals/feedback/collector.py +214 -0
- fi/evals/feedback/hooks.py +81 -0
- fi/evals/feedback/retriever.py +128 -0
- fi/evals/feedback/store.py +272 -0
- fi/evals/feedback/types.py +99 -0
- fi/evals/framework/README.md +79 -0
- fi/evals/framework/__init__.py +267 -0
- fi/evals/framework/backends/Dockerfile.eval-runner +33 -0
- fi/evals/framework/backends/__init__.py +99 -0
- fi/evals/framework/backends/_container.py +141 -0
- fi/evals/framework/backends/_utils.py +145 -0
- fi/evals/framework/backends/base.py +223 -0
- fi/evals/framework/backends/celery_backend.py +417 -0
- fi/evals/framework/backends/celery_worker.py +78 -0
- fi/evals/framework/backends/kubernetes_backend.py +665 -0
- fi/evals/framework/backends/ray_backend.py +521 -0
- fi/evals/framework/backends/temporal.py +350 -0
- fi/evals/framework/backends/temporal_worker.py +126 -0
- fi/evals/framework/backends/thread_pool.py +286 -0
- fi/evals/framework/context.py +258 -0
- fi/evals/framework/enrichment.py +306 -0
- fi/evals/framework/evals/__init__.py +68 -0
- fi/evals/framework/evals/agentic.py +399 -0
- fi/evals/framework/evals/builder.py +609 -0
- fi/evals/framework/evals/semantic.py +142 -0
- fi/evals/framework/evaluator.py +647 -0
- fi/evals/framework/evaluators/__init__.py +22 -0
- fi/evals/framework/evaluators/blocking.py +347 -0
- fi/evals/framework/evaluators/non_blocking.py +577 -0
- fi/evals/framework/propagation.py +421 -0
- fi/evals/framework/protocols.py +385 -0
- fi/evals/framework/registry.py +370 -0
- fi/evals/framework/resilience/__init__.py +150 -0
- fi/evals/framework/resilience/circuit_breaker.py +309 -0
- fi/evals/framework/resilience/degradation.py +355 -0
- fi/evals/framework/resilience/health.py +505 -0
- fi/evals/framework/resilience/rate_limiter.py +228 -0
- fi/evals/framework/resilience/retry.py +274 -0
- fi/evals/framework/resilience/types.py +288 -0
- fi/evals/framework/resilience/wrapper.py +433 -0
- fi/evals/framework/types.py +218 -0
- fi/evals/guardrails/README.md +915 -0
- fi/evals/guardrails/__init__.py +96 -0
- fi/evals/guardrails/backends/__init__.py +43 -0
- fi/evals/guardrails/backends/azure.py +361 -0
- fi/evals/guardrails/backends/base.py +88 -0
- fi/evals/guardrails/backends/generic_llm.py +163 -0
- fi/evals/guardrails/backends/granite.py +216 -0
- fi/evals/guardrails/backends/llamaguard.py +221 -0
- fi/evals/guardrails/backends/local_base.py +479 -0
- fi/evals/guardrails/backends/openai.py +365 -0
- fi/evals/guardrails/backends/qwen.py +170 -0
- fi/evals/guardrails/backends/shieldgemma.py +154 -0
- fi/evals/guardrails/backends/turing.py +235 -0
- fi/evals/guardrails/backends/vllm_client.py +321 -0
- fi/evals/guardrails/backends/wildguard.py +188 -0
- fi/evals/guardrails/base.py +888 -0
- fi/evals/guardrails/config.py +221 -0
- fi/evals/guardrails/discovery.py +243 -0
- fi/evals/guardrails/gateway.py +437 -0
- fi/evals/guardrails/registry.py +231 -0
- fi/evals/guardrails/scanners/__init__.py +127 -0
- fi/evals/guardrails/scanners/base.py +191 -0
- fi/evals/guardrails/scanners/code_injection.py +243 -0
- fi/evals/guardrails/scanners/eval_delegate.py +574 -0
- fi/evals/guardrails/scanners/invisible_chars.py +351 -0
- fi/evals/guardrails/scanners/jailbreak.py +412 -0
- fi/evals/guardrails/scanners/language.py +288 -0
- fi/evals/guardrails/scanners/pipeline.py +260 -0
- fi/evals/guardrails/scanners/regex.py +311 -0
- fi/evals/guardrails/scanners/secrets.py +274 -0
- fi/evals/guardrails/scanners/topics.py +649 -0
- fi/evals/guardrails/scanners/urls.py +341 -0
- fi/evals/guardrails/types.py +96 -0
- fi/evals/llm/__init__.py +3 -0
- fi/evals/llm/base_llm_provider.py +35 -0
- fi/evals/llm/providers/litellm.py +70 -0
- fi/evals/local/__init__.py +90 -0
- fi/evals/local/evaluator.py +690 -0
- fi/evals/local/execution_mode.py +121 -0
- fi/evals/local/llm.py +489 -0
- fi/evals/local/metrics/__init__.py +19 -0
- fi/evals/local/registry.py +360 -0
- fi/evals/manager.py +1018 -0
- fi/evals/manager_types.py +362 -0
- fi/evals/metrics/__init__.py +185 -0
- fi/evals/metrics/agents/__init__.py +74 -0
- fi/evals/metrics/agents/metrics.py +693 -0
- fi/evals/metrics/agents/report.py +36463 -0
- fi/evals/metrics/agents/types.py +160 -0
- fi/evals/metrics/base_llm_metric.py +111 -0
- fi/evals/metrics/base_metric.py +138 -0
- fi/evals/metrics/code_security/__init__.py +305 -0
- fi/evals/metrics/code_security/analyzer.py +985 -0
- fi/evals/metrics/code_security/benchmarks/__init__.py +73 -0
- fi/evals/metrics/code_security/benchmarks/builtin.py +750 -0
- fi/evals/metrics/code_security/benchmarks/loader.py +580 -0
- fi/evals/metrics/code_security/benchmarks/types.py +308 -0
- fi/evals/metrics/code_security/detectors/__init__.py +186 -0
- fi/evals/metrics/code_security/detectors/base.py +394 -0
- fi/evals/metrics/code_security/detectors/cryptography.py +345 -0
- fi/evals/metrics/code_security/detectors/injection.py +744 -0
- fi/evals/metrics/code_security/detectors/secrets.py +287 -0
- fi/evals/metrics/code_security/detectors/serialization.py +192 -0
- fi/evals/metrics/code_security/joint_metrics.py +588 -0
- fi/evals/metrics/code_security/judges/__init__.py +83 -0
- fi/evals/metrics/code_security/judges/base.py +238 -0
- fi/evals/metrics/code_security/judges/dual_judge.py +534 -0
- fi/evals/metrics/code_security/judges/llm_judge.py +301 -0
- fi/evals/metrics/code_security/judges/pattern_judge.py +515 -0
- fi/evals/metrics/code_security/metrics.py +388 -0
- fi/evals/metrics/code_security/modes/__init__.py +63 -0
- fi/evals/metrics/code_security/modes/adversarial.py +284 -0
- fi/evals/metrics/code_security/modes/autocomplete.py +198 -0
- fi/evals/metrics/code_security/modes/base.py +283 -0
- fi/evals/metrics/code_security/modes/instruct.py +253 -0
- fi/evals/metrics/code_security/modes/repair.py +230 -0
- fi/evals/metrics/code_security/reports/__init__.py +57 -0
- fi/evals/metrics/code_security/reports/generator.py +404 -0
- fi/evals/metrics/code_security/reports/leaderboard.py +509 -0
- fi/evals/metrics/code_security/types.py +534 -0
- fi/evals/metrics/function_calling/__init__.py +34 -0
- fi/evals/metrics/function_calling/metrics.py +573 -0
- fi/evals/metrics/function_calling/types.py +87 -0
- fi/evals/metrics/hallucination/__init__.py +54 -0
- fi/evals/metrics/hallucination/detector.py +149 -0
- fi/evals/metrics/hallucination/metrics.py +390 -0
- fi/evals/metrics/hallucination/nli.py +253 -0
- fi/evals/metrics/hallucination/sentinel.py +106 -0
- fi/evals/metrics/hallucination/types.py +132 -0
- fi/evals/metrics/heuristics/aggregation_metrics.py +85 -0
- fi/evals/metrics/heuristics/json_metrics.py +87 -0
- fi/evals/metrics/heuristics/similarity_metrics.py +375 -0
- fi/evals/metrics/heuristics/string_metrics.py +391 -0
- fi/evals/metrics/llm_as_judges/__init__.py +17 -0
- fi/evals/metrics/llm_as_judges/custom_judge/metric.py +112 -0
- fi/evals/metrics/llm_as_judges/custom_judge/prompts.py +26 -0
- fi/evals/metrics/llm_as_judges/types.py +48 -0
- fi/evals/metrics/rag/__init__.py +111 -0
- fi/evals/metrics/rag/advanced/__init__.py +14 -0
- fi/evals/metrics/rag/advanced/multi_hop.py +283 -0
- fi/evals/metrics/rag/advanced/source_attribution.py +344 -0
- fi/evals/metrics/rag/generation/__init__.py +17 -0
- fi/evals/metrics/rag/generation/answer_relevancy.py +176 -0
- fi/evals/metrics/rag/generation/context_utilization.py +245 -0
- fi/evals/metrics/rag/generation/faithfulness.py +241 -0
- fi/evals/metrics/rag/generation/groundedness.py +131 -0
- fi/evals/metrics/rag/rag_score.py +277 -0
- fi/evals/metrics/rag/retrieval/__init__.py +20 -0
- fi/evals/metrics/rag/retrieval/context_entity_recall.py +124 -0
- fi/evals/metrics/rag/retrieval/context_precision.py +158 -0
- fi/evals/metrics/rag/retrieval/context_recall.py +106 -0
- fi/evals/metrics/rag/retrieval/noise_sensitivity.py +163 -0
- fi/evals/metrics/rag/retrieval/ranking.py +261 -0
- fi/evals/metrics/rag/types.py +100 -0
- fi/evals/metrics/rag/utils/__init__.py +62 -0
- fi/evals/metrics/rag/utils/claims.py +189 -0
- fi/evals/metrics/rag/utils/entities.py +244 -0
- fi/evals/metrics/rag/utils/nli.py +92 -0
- fi/evals/metrics/rag/utils/similarity.py +345 -0
- fi/evals/metrics/structured/__init__.py +114 -0
- fi/evals/metrics/structured/field_completeness.py +313 -0
- fi/evals/metrics/structured/hierarchy_score.py +366 -0
- fi/evals/metrics/structured/json_validation.py +190 -0
- fi/evals/metrics/structured/schema_compliance.py +280 -0
- fi/evals/metrics/structured/structured_output_score.py +298 -0
- fi/evals/metrics/structured/types.py +108 -0
- fi/evals/metrics/structured/validators/__init__.py +30 -0
- fi/evals/metrics/structured/validators/base.py +189 -0
- fi/evals/metrics/structured/validators/json_validator.py +196 -0
- fi/evals/metrics/structured/validators/pydantic_validator.py +178 -0
- fi/evals/metrics/structured/validators/yaml_validator.py +248 -0
- fi/evals/otel/__init__.py +266 -0
- fi/evals/otel/config.py +400 -0
- fi/evals/otel/conventions.py +463 -0
- fi/evals/otel/enrichment.py +371 -0
- fi/evals/otel/instrumentors/__init__.py +140 -0
- fi/evals/otel/instrumentors/anthropic.py +517 -0
- fi/evals/otel/instrumentors/base.py +382 -0
- fi/evals/otel/instrumentors/openai.py +673 -0
- fi/evals/otel/processors/__init__.py +36 -0
- fi/evals/otel/processors/base.py +473 -0
- fi/evals/otel/processors/cost.py +445 -0
- fi/evals/otel/processors/evaluation.py +559 -0
- fi/evals/otel/processors/llm.py +462 -0
- fi/evals/otel/tracer.py +506 -0
- fi/evals/otel/types.py +232 -0
- fi/evals/otel_utils.py +23 -0
- fi/evals/protect.py +671 -0
- fi/evals/protect_input_adapter.py +154 -0
- fi/evals/streaming/__init__.py +88 -0
- fi/evals/streaming/buffer.py +213 -0
- fi/evals/streaming/evaluator.py +551 -0
- fi/evals/streaming/policy.py +307 -0
- fi/evals/streaming/scorers.py +368 -0
- fi/evals/streaming/types.py +238 -0
- fi/evals/templates.py +472 -0
- fi/evals/types.py +156 -0
- fi/opt/__init__.py +221 -0
- fi/opt/_objective_scoring.py +85 -0
- fi/opt/base/__init__.py +11 -0
- fi/opt/base/base_generator.py +33 -0
- fi/opt/base/base_mapper.py +26 -0
- fi/opt/base/base_optimizer.py +45 -0
- fi/opt/base/evaluator.py +211 -0
- fi/opt/components.py +3095 -0
- fi/opt/datamappers/__init__.py +3 -0
- fi/opt/datamappers/basic_mapper.py +40 -0
- fi/opt/deployment.py +1021 -0
- fi/opt/evidence.py +4332 -0
- fi/opt/generators/__init__.py +3 -0
- fi/opt/generators/litellm.py +66 -0
- fi/opt/integrations/__init__.py +23 -0
- fi/opt/integrations/generative_suite.py +410 -0
- fi/opt/integrations/simulate.py +1313 -0
- fi/opt/mutations.py +771 -0
- fi/opt/observability.py +4639 -0
- fi/opt/optimizer_trace.py +889 -0
- fi/opt/optimizers/__init__.py +80 -0
- fi/opt/optimizers/agent.py +331 -0
- fi/opt/optimizers/agent_bandit.py +392 -0
- fi/opt/optimizers/agent_curriculum.py +635 -0
- fi/opt/optimizers/agent_evolution.py +894 -0
- fi/opt/optimizers/agent_feedback.py +1863 -0
- fi/opt/optimizers/agent_pareto.py +547 -0
- fi/opt/optimizers/agent_social_memory.py +1113 -0
- fi/opt/optimizers/agent_tpe.py +321 -0
- fi/opt/optimizers/bayesian_search.py +449 -0
- fi/opt/optimizers/council.py +2075 -0
- fi/opt/optimizers/futureagi_replay.py +799 -0
- fi/opt/optimizers/gepa.py +322 -0
- fi/opt/optimizers/metaprompt.py +243 -0
- fi/opt/optimizers/promptwizard.py +417 -0
- fi/opt/optimizers/protegi.py +329 -0
- fi/opt/optimizers/random_search.py +224 -0
- fi/opt/research.py +518 -0
- fi/opt/simulation.py +260 -0
- fi/opt/targets.py +232 -0
- fi/opt/types.py +66 -0
- fi/opt/utils/__init__.py +4 -0
- fi/opt/utils/early_stopping.py +266 -0
- fi/opt/utils/setup_logging.py +82 -0
- fi/simulate/__init__.py +540 -0
- fi/simulate/_hashing.py +35 -0
- fi/simulate/_logging.py +10 -0
- fi/simulate/adapters.py +87 -0
- fi/simulate/agent/__init__.py +120 -0
- fi/simulate/agent/browser.py +658 -0
- fi/simulate/agent/definition.py +587 -0
- fi/simulate/agent/frameworks.py +3528 -0
- fi/simulate/agent/generic.py +8286 -0
- fi/simulate/agent/import_probe.py +227 -0
- fi/simulate/agent/memory.py +905 -0
- fi/simulate/agent/mocks.py +101 -0
- fi/simulate/agent/multi_agent.py +361 -0
- fi/simulate/agent/orchestration.py +903 -0
- fi/simulate/agent/realtime.py +665 -0
- fi/simulate/agent/wrapper.py +99 -0
- fi/simulate/agent/wrappers/__init__.py +18 -0
- fi/simulate/agent/wrappers/anthropic.py +62 -0
- fi/simulate/agent/wrappers/gemini.py +65 -0
- fi/simulate/agent/wrappers/http.py +404 -0
- fi/simulate/agent/wrappers/langchain.py +80 -0
- fi/simulate/agent/wrappers/openai.py +75 -0
- fi/simulate/agent/wrappers/websocket.py +326 -0
- fi/simulate/artifacts/__init__.py +11 -0
- fi/simulate/artifacts/manifest.py +62 -0
- fi/simulate/cli.py +20560 -0
- fi/simulate/endpoints/__init__.py +45 -0
- fi/simulate/endpoints/_http_actor.py +73 -0
- fi/simulate/endpoints/actor_sources.py +243 -0
- fi/simulate/endpoints/base.py +107 -0
- fi/simulate/endpoints/builtins.py +10 -0
- fi/simulate/endpoints/callable.py +95 -0
- fi/simulate/endpoints/http.py +76 -0
- fi/simulate/endpoints/livekit.py +138 -0
- fi/simulate/endpoints/originators.py +132 -0
- fi/simulate/endpoints/profiles.py +348 -0
- fi/simulate/endpoints/retell.py +633 -0
- fi/simulate/endpoints/vapi.py +205 -0
- fi/simulate/endpoints/websocket.py +76 -0
- fi/simulate/environment.py +33026 -0
- fi/simulate/environments/__init__.py +11 -0
- fi/simulate/environments/base.py +73 -0
- fi/simulate/environments/chat.py +697 -0
- fi/simulate/environments/voice.py +212 -0
- fi/simulate/evaluation/__init__.py +4 -0
- fi/simulate/evaluation/ai_eval.py +227 -0
- fi/simulate/evidence/__init__.py +35 -0
- fi/simulate/evidence/base.py +59 -0
- fi/simulate/evidence/caller_observed.py +50 -0
- fi/simulate/evidence/livekit_instrumentation.py +51 -0
- fi/simulate/evidence/livekit_room.py +50 -0
- fi/simulate/evidence/otel.py +49 -0
- fi/simulate/evidence/providers/__init__.py +24 -0
- fi/simulate/evidence/providers/base.py +61 -0
- fi/simulate/evidence/providers/retell.py +376 -0
- fi/simulate/evidence/providers/vapi.py +426 -0
- fi/simulate/hosted/__init__.py +32 -0
- fi/simulate/hosted/child_entrypoint.py +306 -0
- fi/simulate/hosted/job.py +150 -0
- fi/simulate/hosted/targets.py +53 -0
- fi/simulate/instrumentation/__init__.py +5 -0
- fi/simulate/instrumentation/livekit/__init__.py +122 -0
- fi/simulate/manifest.py +1033 -0
- fi/simulate/matrix_cli.py +165 -0
- fi/simulate/realtime/__init__.py +40 -0
- fi/simulate/realtime/events.py +107 -0
- fi/simulate/realtime/media.py +61 -0
- fi/simulate/realtime/session.py +91 -0
- fi/simulate/recording/__init__.py +5 -0
- fi/simulate/recording/room_recorder.py +326 -0
- fi/simulate/registry.py +185 -0
- fi/simulate/results/__init__.py +9 -0
- fi/simulate/results/base.py +18 -0
- fi/simulate/results/filesystem.py +71 -0
- fi/simulate/results/futureagi.py +1340 -0
- fi/simulate/runtime/__init__.py +85 -0
- fi/simulate/runtime/capabilities.py +40 -0
- fi/simulate/runtime/events.py +63 -0
- fi/simulate/runtime/failures.py +25 -0
- fi/simulate/runtime/ids.py +34 -0
- fi/simulate/runtime/plan.py +70 -0
- fi/simulate/runtime/planner.py +102 -0
- fi/simulate/runtime/report.py +174 -0
- fi/simulate/runtime/run.py +75 -0
- fi/simulate/runtime/runner.py +333 -0
- fi/simulate/runtime/spec.py +186 -0
- fi/simulate/simulation/__init__.py +30 -0
- fi/simulate/simulation/behavior_policy.py +425 -0
- fi/simulate/simulation/bridge/__init__.py +9 -0
- fi/simulate/simulation/bridge/audio.py +29 -0
- fi/simulate/simulation/bridge/connector.py +46 -0
- fi/simulate/simulation/bridge/livekit.py +252 -0
- fi/simulate/simulation/bridge/retell.py +188 -0
- fi/simulate/simulation/bridge/vapi.py +177 -0
- fi/simulate/simulation/contract.py +419 -0
- fi/simulate/simulation/engines/__init__.py +12 -0
- fi/simulate/simulation/engines/base.py +21 -0
- fi/simulate/simulation/engines/cloud.py +517 -0
- fi/simulate/simulation/engines/livekit.py +4167 -0
- fi/simulate/simulation/engines/local_text.py +89 -0
- fi/simulate/simulation/fidelity.py +374 -0
- fi/simulate/simulation/gemini_tts_stream.py +110 -0
- fi/simulate/simulation/generator.py +91 -0
- fi/simulate/simulation/goal_machine.py +185 -0
- fi/simulate/simulation/livekit_models.py +467 -0
- fi/simulate/simulation/matrix.py +170 -0
- fi/simulate/simulation/models.py +279 -0
- fi/simulate/simulation/runner.py +153 -0
- fi/simulate/simulation/synthetic.py +880 -0
- fi/simulate/simulation/voice_prompt.py +502 -0
- fi/simulate/simulator/__init__.py +55 -0
- fi/simulate/simulator/builtins.py +53 -0
- fi/simulate/suite.py +1288 -0
- fi/simulate/utils/routes.py +164 -0
- fi/simulate/voice.py +225 -0
- fi/simulate/voice_cli.py +182 -0
- fi/utils/__init__.py +1 -0
- fi/utils/constants.py +14 -0
- fi/utils/errors.py +200 -0
- fi/utils/executor.py +26 -0
- fi/utils/routes.py +119 -0
- fi/utils/utils.py +17 -0
|
@@ -0,0 +1,341 @@
|
|
|
1
|
+
"""Pipecat live lane (3C) — real Pipecat ``Pipeline`` with frame injection.
|
|
2
|
+
|
|
3
|
+
Framework imports: NONE at module top (P3-D1); execution happens in
|
|
4
|
+
``_workers/pipecat_worker.py``. Rung 1 (default, implemented) injects
|
|
5
|
+
``TranscriptionFrame``s — bypassing STT/TTS, Pipecat's own documented eval
|
|
6
|
+
technique — and collects output frames + TTFB/processing timing. Same
|
|
7
|
+
dual-channel/perturbation/variance contract as 3B (PRD §4.3); rung-1
|
|
8
|
+
honesty: timing-only voice metrics, no ``channels`` block.
|
|
9
|
+
"""
|
|
10
|
+
|
|
11
|
+
from __future__ import annotations
|
|
12
|
+
|
|
13
|
+
import tempfile
|
|
14
|
+
import uuid
|
|
15
|
+
from pathlib import Path
|
|
16
|
+
from typing import Any, Mapping, Optional, Sequence
|
|
17
|
+
|
|
18
|
+
from ._contract import lane_budget_s, require_lane_enabled
|
|
19
|
+
from ._perturb import apply_text_perturbations, perturbations_stanza
|
|
20
|
+
from ._runner import run_worker_once
|
|
21
|
+
from ._stats import (
|
|
22
|
+
derive_channel_evidence,
|
|
23
|
+
lane_run_payload,
|
|
24
|
+
primary_transcript_events,
|
|
25
|
+
run_repeated,
|
|
26
|
+
)
|
|
27
|
+
|
|
28
|
+
_WORKERS = Path(__file__).resolve().parent / "_workers"
|
|
29
|
+
_RUNG_LABELS = {1: "frame_injection", 2: "loopback_transport", 3: "credentialed_providers"}
|
|
30
|
+
|
|
31
|
+
_DEFAULT_TURNS = (
|
|
32
|
+
{"user": "Hello there."},
|
|
33
|
+
{"user": "What can you help me with today?"},
|
|
34
|
+
)
|
|
35
|
+
|
|
36
|
+
|
|
37
|
+
def _scenario_turns(scenario: Mapping[str, Any]) -> list[dict[str, Any]]:
|
|
38
|
+
raw = scenario.get("turns") or scenario.get("user_messages")
|
|
39
|
+
if not raw:
|
|
40
|
+
return [dict(turn) for turn in _DEFAULT_TURNS]
|
|
41
|
+
turns: list[dict[str, Any]] = []
|
|
42
|
+
for item in raw:
|
|
43
|
+
if isinstance(item, str):
|
|
44
|
+
turns.append({"user": item})
|
|
45
|
+
elif isinstance(item, Mapping):
|
|
46
|
+
turns.append(dict(item))
|
|
47
|
+
return turns or [dict(turn) for turn in _DEFAULT_TURNS]
|
|
48
|
+
|
|
49
|
+
|
|
50
|
+
def _frame_timing(events: Sequence[Mapping[str, Any]]) -> dict[str, Any]:
|
|
51
|
+
"""TTFB/processing timing evidence reported by the worker (PRD §4.3)."""
|
|
52
|
+
|
|
53
|
+
ttfb: list[float] = []
|
|
54
|
+
processing: list[float] = []
|
|
55
|
+
for event in events:
|
|
56
|
+
if event.get("channel") == "lane" and event.get("type") == "timing":
|
|
57
|
+
payload = event.get("payload")
|
|
58
|
+
payload = payload if isinstance(payload, Mapping) else {}
|
|
59
|
+
if isinstance(payload.get("ttfb_ms"), (int, float)):
|
|
60
|
+
ttfb.append(float(payload["ttfb_ms"]))
|
|
61
|
+
if isinstance(payload.get("processing_ms"), (int, float)):
|
|
62
|
+
processing.append(float(payload["processing_ms"]))
|
|
63
|
+
return {
|
|
64
|
+
"ttfb_ms": ttfb,
|
|
65
|
+
"mean_ttfb_ms": round(sum(ttfb) / len(ttfb), 3) if ttfb else None,
|
|
66
|
+
"processing_ms": processing,
|
|
67
|
+
}
|
|
68
|
+
|
|
69
|
+
|
|
70
|
+
def _realtime_state(
|
|
71
|
+
events: Sequence[Mapping[str, Any]], *, rung_label: str
|
|
72
|
+
) -> dict[str, Any]:
|
|
73
|
+
items = []
|
|
74
|
+
for index, event in enumerate(events, start=1):
|
|
75
|
+
if event.get("channel") in ("user", "agent", "tool"):
|
|
76
|
+
payload = event.get("payload")
|
|
77
|
+
payload = payload if isinstance(payload, Mapping) else {}
|
|
78
|
+
items.append(
|
|
79
|
+
{
|
|
80
|
+
"index": index,
|
|
81
|
+
"channel": event.get("channel"),
|
|
82
|
+
"item_type": event.get("type"),
|
|
83
|
+
"text": payload.get("text"),
|
|
84
|
+
}
|
|
85
|
+
)
|
|
86
|
+
return {
|
|
87
|
+
"engine": "live_lane_pipecat",
|
|
88
|
+
"rung": rung_label,
|
|
89
|
+
"item_count": len(items),
|
|
90
|
+
"items": items[:200],
|
|
91
|
+
}
|
|
92
|
+
|
|
93
|
+
|
|
94
|
+
def _rung2_loopback_channels(
|
|
95
|
+
turns: Sequence[Mapping[str, Any]],
|
|
96
|
+
*,
|
|
97
|
+
loopback: Optional[Mapping[str, Any]],
|
|
98
|
+
codec_profile: str,
|
|
99
|
+
seed: int,
|
|
100
|
+
acoustic_operators: Sequence[str] = (),
|
|
101
|
+
) -> tuple[dict[str, Any], str, list[dict[str, Any]]]:
|
|
102
|
+
"""Phase 9A unit 2 + Phase-12 12C rung-2 — the rung-2 loopback dispatch
|
|
103
|
+
(§2.1 / §2.5 + ARCH §2c), byte-parallel to the LiveKit lane. The rung-2
|
|
104
|
+
``loopback_transport`` label is byte-identical across both lanes (the seam 9A
|
|
105
|
+
grows). Produces the two PCM streams via the deterministic ``_loopback``
|
|
106
|
+
round-trip, applies the rung-2 ACOUSTIC operators (``mix_noise`` /
|
|
107
|
+
``mix_interference`` / ``reverb_blend`` over the user PCM) BEFORE the codec
|
|
108
|
+
stage, applies the default-ON codec round-trip (9A-A11) unless
|
|
109
|
+
``codec_profile == "none"``, feeds the ALREADY-BUILT
|
|
110
|
+
``derive_channel_evidence`` (REUSED), and returns ``channels`` + the
|
|
111
|
+
``fidelity_tier`` marker + the applied acoustic operator records. The
|
|
112
|
+
codec-survival score is computed on the perturbed-then-channel signal so
|
|
113
|
+
``phone_survival`` honestly reflects channel reproduction (P12-D2)."""
|
|
114
|
+
|
|
115
|
+
from fi.alk import live # sanctioned facade idiom (cli.py)
|
|
116
|
+
|
|
117
|
+
cfg = dict(loopback or {})
|
|
118
|
+
tick_ms = float(cfg.get("tick_ms", live._loopback.DEFAULT_TICK_MS))
|
|
119
|
+
sample_rate = int(cfg.get("sample_rate", live._loopback.DEFAULT_SAMPLE_RATE))
|
|
120
|
+
loop_seed = int(cfg.get("seed", seed))
|
|
121
|
+
profile = str(cfg.get("codec_profile", codec_profile))
|
|
122
|
+
|
|
123
|
+
loop = live._loopback.run_loopback_roundtrip(
|
|
124
|
+
list(turns),
|
|
125
|
+
user_wav=cfg.get("user_wav"),
|
|
126
|
+
agent_wav=cfg.get("agent_wav"),
|
|
127
|
+
tick_ms=tick_ms,
|
|
128
|
+
sample_rate=sample_rate,
|
|
129
|
+
seed=loop_seed,
|
|
130
|
+
)
|
|
131
|
+
user_pcm, agent_pcm = loop["user_pcm"], loop["agent_pcm"]
|
|
132
|
+
|
|
133
|
+
# Phase-12 12C rung-2: the acoustic attack rides the USER channel, applied to
|
|
134
|
+
# the CLEAN loopback PCM before the codec stage; deterministic under loop_seed.
|
|
135
|
+
acoustic_applied: list[dict[str, Any]] = []
|
|
136
|
+
attacked_user_pcm = user_pcm
|
|
137
|
+
if acoustic_operators:
|
|
138
|
+
attacked_user_pcm, acoustic_applied = live._perturb.apply_acoustic_perturbations(
|
|
139
|
+
user_pcm,
|
|
140
|
+
list(acoustic_operators),
|
|
141
|
+
seed=loop_seed,
|
|
142
|
+
sample_rate=sample_rate,
|
|
143
|
+
)
|
|
144
|
+
user_pcm = attacked_user_pcm
|
|
145
|
+
|
|
146
|
+
codec_record: dict[str, Any] | None = None
|
|
147
|
+
phone_survival: dict[str, Any] | None = None
|
|
148
|
+
if profile != "none":
|
|
149
|
+
user_pcm, agent_pcm, codec_record = live._codec.apply_codec_profile(
|
|
150
|
+
user_pcm, agent_pcm, profile=profile, seed=loop_seed, sample_rate=sample_rate
|
|
151
|
+
)
|
|
152
|
+
codec, packet_loss = live._codec._PROFILE_BUNDLE[profile]
|
|
153
|
+
phone_survival = live._codec.score_codec_survival(
|
|
154
|
+
loop["user_pcm"],
|
|
155
|
+
attacked_user_pcm,
|
|
156
|
+
codec=codec,
|
|
157
|
+
packet_loss=packet_loss,
|
|
158
|
+
seed=loop_seed,
|
|
159
|
+
sample_rate=sample_rate,
|
|
160
|
+
)
|
|
161
|
+
|
|
162
|
+
derived = derive_channel_evidence(
|
|
163
|
+
user_pcm, agent_pcm, sample_rate=(8000 if profile != "none" else sample_rate)
|
|
164
|
+
)
|
|
165
|
+
channels: dict[str, Any] = {
|
|
166
|
+
"derived": derived,
|
|
167
|
+
"source": "derive_channel_evidence",
|
|
168
|
+
"rung": _RUNG_LABELS[2],
|
|
169
|
+
"fidelity_tier": "deterministic_loopback",
|
|
170
|
+
"seed": loop_seed,
|
|
171
|
+
"loopback_provenance": loop["provenance"],
|
|
172
|
+
}
|
|
173
|
+
if codec_record is not None:
|
|
174
|
+
channels["codec_round_trip"] = codec_record
|
|
175
|
+
if phone_survival is not None:
|
|
176
|
+
channels["phone_survival"] = phone_survival
|
|
177
|
+
if acoustic_applied:
|
|
178
|
+
channels["acoustic_operators"] = acoustic_applied
|
|
179
|
+
return channels, "deterministic_loopback", acoustic_applied
|
|
180
|
+
|
|
181
|
+
|
|
182
|
+
def run_pipecat_lane(
|
|
183
|
+
pipeline_factory_path: str | None,
|
|
184
|
+
scenario: Mapping[str, Any],
|
|
185
|
+
*,
|
|
186
|
+
rung: int = 1,
|
|
187
|
+
repeats: int = 8,
|
|
188
|
+
stressed: bool = False,
|
|
189
|
+
perturbations: Optional[Sequence[str]] = None,
|
|
190
|
+
seed: int = 0,
|
|
191
|
+
required_env: Optional[Sequence[str]] = None,
|
|
192
|
+
version_requirement: str | None = None,
|
|
193
|
+
budget_s: float | None = None,
|
|
194
|
+
artifacts_dir: str | Path | None = None,
|
|
195
|
+
# Phase 9A (BBG A2): additive optional loopback config consumed ONLY on the
|
|
196
|
+
# rung==2 branch; rung-1/rung-3 callers are unaffected.
|
|
197
|
+
loopback: Optional[Mapping[str, Any]] = None,
|
|
198
|
+
codec_profile: str = "g711_ulaw_8k_ge",
|
|
199
|
+
) -> dict[str, Any]:
|
|
200
|
+
require_lane_enabled("pipecat")
|
|
201
|
+
if rung >= 3:
|
|
202
|
+
require_lane_enabled("credentialed")
|
|
203
|
+
if rung not in _RUNG_LABELS:
|
|
204
|
+
raise ValueError(f"rung must be one of {sorted(_RUNG_LABELS)}, got {rung}")
|
|
205
|
+
|
|
206
|
+
required = tuple(required_env) if required_env is not None else ()
|
|
207
|
+
operators = list(perturbations or (["asr_error"] if stressed else []))
|
|
208
|
+
turns = _scenario_turns(scenario)
|
|
209
|
+
# Phase-12 12C rung-2: split text-rung operators from acoustic operators
|
|
210
|
+
# (applied to the rung-2 loopback PCM). The rung wall is unchanged at rung-1.
|
|
211
|
+
from ._perturb import ACOUSTIC_RUNG_OPERATORS
|
|
212
|
+
|
|
213
|
+
acoustic_operators = [op for op in operators if op in ACOUSTIC_RUNG_OPERATORS]
|
|
214
|
+
text_operators = [op for op in operators if op not in ACOUSTIC_RUNG_OPERATORS]
|
|
215
|
+
if rung != 2 and acoustic_operators:
|
|
216
|
+
raise ValueError(
|
|
217
|
+
f"acoustic operators {acoustic_operators} need a real audio channel "
|
|
218
|
+
"(rung 2 loopback transport or above); rung "
|
|
219
|
+
f"{rung} ({_RUNG_LABELS[rung]}) is a text-rung tier"
|
|
220
|
+
)
|
|
221
|
+
applied: list[dict[str, Any]] = []
|
|
222
|
+
if text_operators:
|
|
223
|
+
turns, applied = apply_text_perturbations(turns, text_operators, seed=seed)
|
|
224
|
+
|
|
225
|
+
# Phase 9A unit 2: the rung wall narrows — rung-2 dispatches into the
|
|
226
|
+
# deterministic loopback (§2.1); rung-3 still raises (the owner live-proof,
|
|
227
|
+
# unit 7). rung-1 is completely untouched (timing-only, NO channels block).
|
|
228
|
+
channels: dict[str, Any] | None = None
|
|
229
|
+
fidelity_tier: str | None = None
|
|
230
|
+
acoustic_applied: list[dict[str, Any]] = []
|
|
231
|
+
if rung == 2:
|
|
232
|
+
channels, fidelity_tier, acoustic_applied = _rung2_loopback_channels(
|
|
233
|
+
turns,
|
|
234
|
+
loopback=loopback,
|
|
235
|
+
codec_profile=codec_profile,
|
|
236
|
+
seed=seed,
|
|
237
|
+
acoustic_operators=acoustic_operators,
|
|
238
|
+
)
|
|
239
|
+
# §2.5 binding correction: a deterministic in-process loopback is
|
|
240
|
+
# NEVER live_lane.
|
|
241
|
+
evidence_class = "live_stressed"
|
|
242
|
+
elif rung != 1:
|
|
243
|
+
# rung == 3: unchanged keyed path; still requires the credentialed flag
|
|
244
|
+
# + RUNG3_REQUIRED_ENV; rung-3 lands as the owner live-proof (unit 7).
|
|
245
|
+
raise NotImplementedError(
|
|
246
|
+
f"pipecat lane rung {rung} ({_RUNG_LABELS[rung]}) is not "
|
|
247
|
+
"implemented yet; rung 1 (frame_injection) and rung 2 "
|
|
248
|
+
"(loopback_transport) are the supported tiers — rung 3 "
|
|
249
|
+
"(credentialed_providers) is the owner-keyed live-proof lane"
|
|
250
|
+
)
|
|
251
|
+
else:
|
|
252
|
+
evidence_class = "live_stressed" if operators else "live_lane"
|
|
253
|
+
|
|
254
|
+
base_dir = (
|
|
255
|
+
Path(artifacts_dir)
|
|
256
|
+
if artifacts_dir is not None
|
|
257
|
+
else Path(tempfile.mkdtemp(prefix="agent-learning-live-pipecat-"))
|
|
258
|
+
)
|
|
259
|
+
run_id = uuid.uuid4().hex
|
|
260
|
+
resolved_budget = float(budget_s) if budget_s is not None else lane_budget_s("pipecat")
|
|
261
|
+
boot = {
|
|
262
|
+
"type": "boot",
|
|
263
|
+
"lane": "pipecat",
|
|
264
|
+
"rung": rung,
|
|
265
|
+
"scenario": {"name": str(scenario.get("name") or "pipecat-smoke")},
|
|
266
|
+
"turns": turns,
|
|
267
|
+
"config": {
|
|
268
|
+
"pipeline_factory": pipeline_factory_path,
|
|
269
|
+
"responses": scenario.get("responses"),
|
|
270
|
+
"expect": scenario.get("expect"),
|
|
271
|
+
},
|
|
272
|
+
}
|
|
273
|
+
worker = _WORKERS / "pipecat_worker.py"
|
|
274
|
+
|
|
275
|
+
def _run_once(index: int, transcript: Any) -> dict[str, Any]:
|
|
276
|
+
return run_worker_once(
|
|
277
|
+
worker,
|
|
278
|
+
boot,
|
|
279
|
+
lane="pipecat",
|
|
280
|
+
required_env=required,
|
|
281
|
+
cwd=base_dir,
|
|
282
|
+
timeout_s=resolved_budget,
|
|
283
|
+
transcript=transcript,
|
|
284
|
+
version_requirement=version_requirement,
|
|
285
|
+
)
|
|
286
|
+
|
|
287
|
+
result = run_repeated(
|
|
288
|
+
_run_once,
|
|
289
|
+
lane="pipecat",
|
|
290
|
+
evidence_class=evidence_class,
|
|
291
|
+
repeats=repeats,
|
|
292
|
+
budget_s=budget_s,
|
|
293
|
+
required_env=required,
|
|
294
|
+
artifacts_dir=base_dir,
|
|
295
|
+
run_id=run_id,
|
|
296
|
+
rung=_RUNG_LABELS[rung],
|
|
297
|
+
framework="pipecat-ai",
|
|
298
|
+
version_requirement=version_requirement,
|
|
299
|
+
)
|
|
300
|
+
|
|
301
|
+
events = primary_transcript_events(result)
|
|
302
|
+
from .. import simulate as _simulate
|
|
303
|
+
|
|
304
|
+
manifest = _simulate.build_realtime_run_manifest(
|
|
305
|
+
name=f"live-pipecat-{run_id[:8]}",
|
|
306
|
+
framework="pipecat",
|
|
307
|
+
required_env=required,
|
|
308
|
+
min_turns=1,
|
|
309
|
+
max_turns=max(len(turns), 1),
|
|
310
|
+
metadata={
|
|
311
|
+
"simulation_engine": "live_lane_pipecat",
|
|
312
|
+
"live_lane": {"lane": "pipecat", "rung": _RUNG_LABELS[rung]},
|
|
313
|
+
},
|
|
314
|
+
)
|
|
315
|
+
|
|
316
|
+
payload = lane_run_payload(
|
|
317
|
+
result,
|
|
318
|
+
name=f"live-pipecat-{run_id[:8]}",
|
|
319
|
+
scenario=scenario,
|
|
320
|
+
manifest=manifest,
|
|
321
|
+
states={"realtime_trace": _realtime_state(events, rung_label=_RUNG_LABELS[rung])},
|
|
322
|
+
metadata={
|
|
323
|
+
"execution_model": "subprocess",
|
|
324
|
+
"rung": _RUNG_LABELS[rung],
|
|
325
|
+
# rung-1 honesty: timing-only voice metrics, NO channels block
|
|
326
|
+
"voice_timing": _frame_timing(events),
|
|
327
|
+
},
|
|
328
|
+
)
|
|
329
|
+
all_applied = list(applied) + list(acoustic_applied)
|
|
330
|
+
if all_applied:
|
|
331
|
+
payload["live_lane"]["perturbations"] = perturbations_stanza(
|
|
332
|
+
all_applied, seed=seed, paired_clean_run=None
|
|
333
|
+
)
|
|
334
|
+
if channels is not None:
|
|
335
|
+
# rung-2: attach the dual-channel evidence + the fidelity marker (§2.5 /
|
|
336
|
+
# 9A-A10). fidelity_tier is a MARKER FIELD, not a new evidence class.
|
|
337
|
+
payload["channels"] = channels
|
|
338
|
+
if isinstance(payload.get("live_lane"), dict):
|
|
339
|
+
payload["live_lane"]["fidelity_tier"] = fidelity_tier
|
|
340
|
+
payload["fidelity_tier"] = fidelity_tier
|
|
341
|
+
return payload
|