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 +1,80 @@
|
|
|
1
|
+
from typing import Any, Union
|
|
2
|
+
from fi.simulate.agent.wrapper import AgentWrapper, AgentInput, AgentResponse
|
|
3
|
+
|
|
4
|
+
try:
|
|
5
|
+
from langchain_core.messages import HumanMessage, AIMessage, SystemMessage
|
|
6
|
+
except ImportError:
|
|
7
|
+
# LangChain is optional. Fallback message classes keep wrappers and tests
|
|
8
|
+
# usable for mocked/local simulations without installing langchain-core.
|
|
9
|
+
class _FallbackMessage:
|
|
10
|
+
def __init__(self, content: str):
|
|
11
|
+
self.content = content
|
|
12
|
+
|
|
13
|
+
class HumanMessage(_FallbackMessage):
|
|
14
|
+
pass
|
|
15
|
+
|
|
16
|
+
class AIMessage(_FallbackMessage):
|
|
17
|
+
pass
|
|
18
|
+
|
|
19
|
+
class SystemMessage(_FallbackMessage):
|
|
20
|
+
pass
|
|
21
|
+
|
|
22
|
+
LANGCHAIN_INSTALLED = False
|
|
23
|
+
else:
|
|
24
|
+
LANGCHAIN_INSTALLED = True
|
|
25
|
+
|
|
26
|
+
class LangChainAgentWrapper(AgentWrapper):
|
|
27
|
+
"""
|
|
28
|
+
Wrapper for LangChain Runnable or Chain agents.
|
|
29
|
+
"""
|
|
30
|
+
def __init__(self, agent: Any, system_prompt: str = None):
|
|
31
|
+
"""
|
|
32
|
+
Args:
|
|
33
|
+
agent: A LangChain Runnable (chain, agent executor) that accepts input.
|
|
34
|
+
It is expected to accept a dictionary with "messages" or "input".
|
|
35
|
+
system_prompt: Optional system prompt to prepend to message history.
|
|
36
|
+
"""
|
|
37
|
+
self.agent = agent
|
|
38
|
+
self.system_prompt = system_prompt
|
|
39
|
+
|
|
40
|
+
async def call(self, input: AgentInput) -> Union[str, AgentResponse]:
|
|
41
|
+
|
|
42
|
+
# Convert history to LangChain messages
|
|
43
|
+
lc_messages = []
|
|
44
|
+
|
|
45
|
+
if self.system_prompt:
|
|
46
|
+
lc_messages.append(SystemMessage(content=self.system_prompt))
|
|
47
|
+
|
|
48
|
+
for msg in input.messages:
|
|
49
|
+
role = msg.get("role")
|
|
50
|
+
content = msg.get("content")
|
|
51
|
+
if role == "user":
|
|
52
|
+
lc_messages.append(HumanMessage(content=content))
|
|
53
|
+
elif role == "assistant":
|
|
54
|
+
lc_messages.append(AIMessage(content=content))
|
|
55
|
+
elif role == "system":
|
|
56
|
+
lc_messages.append(SystemMessage(content=content))
|
|
57
|
+
|
|
58
|
+
# Invoke the agent
|
|
59
|
+
# We try standard keys used in LC agents
|
|
60
|
+
inputs = {
|
|
61
|
+
"messages": lc_messages,
|
|
62
|
+
"input": input.new_message.get("content") if input.new_message else "",
|
|
63
|
+
"chat_history": lc_messages[:-1] if lc_messages else []
|
|
64
|
+
}
|
|
65
|
+
|
|
66
|
+
# Support both ainvoke and invoke
|
|
67
|
+
if hasattr(self.agent, "ainvoke"):
|
|
68
|
+
response = await self.agent.ainvoke(inputs)
|
|
69
|
+
else:
|
|
70
|
+
response = self.agent.invoke(inputs)
|
|
71
|
+
|
|
72
|
+
# Parse response
|
|
73
|
+
if isinstance(response, str):
|
|
74
|
+
return response
|
|
75
|
+
elif hasattr(response, "content"):
|
|
76
|
+
return response.content
|
|
77
|
+
elif isinstance(response, dict) and "output" in response:
|
|
78
|
+
return response["output"]
|
|
79
|
+
|
|
80
|
+
return str(response)
|
|
@@ -0,0 +1,75 @@
|
|
|
1
|
+
from typing import Any, Union
|
|
2
|
+
from fi.simulate.agent.wrapper import AgentWrapper, AgentInput, AgentResponse
|
|
3
|
+
|
|
4
|
+
class OpenAIAgentWrapper(AgentWrapper):
|
|
5
|
+
"""
|
|
6
|
+
Wrapper for OpenAI-based agents.
|
|
7
|
+
Automatically handles message conversion to OpenAI format.
|
|
8
|
+
"""
|
|
9
|
+
def __init__(self, client: Any, model: str = "gpt-4-turbo", system_prompt: str = None):
|
|
10
|
+
"""
|
|
11
|
+
Args:
|
|
12
|
+
client: The OpenAI client instance (AsyncOpenAI or OpenAI).
|
|
13
|
+
model: The model name to use (e.g., "gpt-4-turbo").
|
|
14
|
+
system_prompt: Optional system instructions for the agent.
|
|
15
|
+
"""
|
|
16
|
+
self.client = client
|
|
17
|
+
self.model = model
|
|
18
|
+
self.system_prompt = system_prompt
|
|
19
|
+
|
|
20
|
+
async def call(self, input: AgentInput) -> Union[str, AgentResponse]:
|
|
21
|
+
# Convert internal message format to OpenAI format
|
|
22
|
+
# Input messages are already in [{"role": "...", "content": "..."}] format which OpenAI accepts
|
|
23
|
+
messages = []
|
|
24
|
+
if self.system_prompt:
|
|
25
|
+
messages.append({"role": "system", "content": self.system_prompt})
|
|
26
|
+
|
|
27
|
+
messages.extend(input.messages)
|
|
28
|
+
|
|
29
|
+
# Check if client is async or sync
|
|
30
|
+
if hasattr(self.client, "chat") and hasattr(self.client.chat, "completions"):
|
|
31
|
+
# Handle AsyncOpenAI vs OpenAI
|
|
32
|
+
if hasattr(self.client.chat.completions, "create"):
|
|
33
|
+
# Modern OpenAI SDK (v1+)
|
|
34
|
+
if is_async_client(self.client):
|
|
35
|
+
completion = await self.client.chat.completions.create(
|
|
36
|
+
model=self.model,
|
|
37
|
+
messages=messages
|
|
38
|
+
)
|
|
39
|
+
else:
|
|
40
|
+
completion = self.client.chat.completions.create(
|
|
41
|
+
model=self.model,
|
|
42
|
+
messages=messages
|
|
43
|
+
)
|
|
44
|
+
|
|
45
|
+
message = completion.choices[0].message
|
|
46
|
+
content = message.content or ""
|
|
47
|
+
|
|
48
|
+
# Extract tool_calls if present
|
|
49
|
+
tool_calls = None
|
|
50
|
+
if hasattr(message, 'tool_calls') and message.tool_calls:
|
|
51
|
+
# Convert tool_calls to dict format for serialization
|
|
52
|
+
tool_calls = []
|
|
53
|
+
for tc in message.tool_calls:
|
|
54
|
+
tool_call_dict = {
|
|
55
|
+
"id": tc.id,
|
|
56
|
+
"type": tc.type,
|
|
57
|
+
"function": {
|
|
58
|
+
"name": tc.function.name,
|
|
59
|
+
"arguments": tc.function.arguments
|
|
60
|
+
}
|
|
61
|
+
}
|
|
62
|
+
tool_calls.append(tool_call_dict)
|
|
63
|
+
|
|
64
|
+
# Return AgentResponse if tool_calls exist, otherwise just content string
|
|
65
|
+
if tool_calls:
|
|
66
|
+
return AgentResponse(content=content, tool_calls=tool_calls)
|
|
67
|
+
return content
|
|
68
|
+
|
|
69
|
+
raise ValueError("Unsupported OpenAI client. Please provide a valid OpenAI or AsyncOpenAI client.")
|
|
70
|
+
|
|
71
|
+
def is_async_client(client: Any) -> bool:
|
|
72
|
+
"""Check if the client is an async client."""
|
|
73
|
+
# Heuristic check for async client
|
|
74
|
+
return type(client).__name__ == "AsyncOpenAI"
|
|
75
|
+
|
|
@@ -0,0 +1,326 @@
|
|
|
1
|
+
from __future__ import annotations
|
|
2
|
+
|
|
3
|
+
import asyncio
|
|
4
|
+
import base64
|
|
5
|
+
import hashlib
|
|
6
|
+
import json
|
|
7
|
+
import os
|
|
8
|
+
import socket
|
|
9
|
+
import time
|
|
10
|
+
from typing import Any, Mapping, Optional
|
|
11
|
+
from urllib.parse import urlparse
|
|
12
|
+
|
|
13
|
+
from fi.simulate.agent.wrapper import (
|
|
14
|
+
AgentInput,
|
|
15
|
+
AgentResponse,
|
|
16
|
+
SimulationArtifact,
|
|
17
|
+
SimulationEvent,
|
|
18
|
+
)
|
|
19
|
+
from fi.simulate.agent.wrapper import AgentWrapper
|
|
20
|
+
from fi.simulate.agent.wrappers.http import (
|
|
21
|
+
_artifact_list,
|
|
22
|
+
_content_text,
|
|
23
|
+
_event_list,
|
|
24
|
+
_optional_mapping,
|
|
25
|
+
_redacted_endpoint,
|
|
26
|
+
_tool_call_list,
|
|
27
|
+
_tool_response_list,
|
|
28
|
+
)
|
|
29
|
+
|
|
30
|
+
|
|
31
|
+
class WebSocketAgentWrapper(AgentWrapper):
|
|
32
|
+
"""WebSocket target adapter for local framework transport simulation."""
|
|
33
|
+
|
|
34
|
+
def __init__(
|
|
35
|
+
self,
|
|
36
|
+
*,
|
|
37
|
+
endpoint: str,
|
|
38
|
+
protocol: str = "fi.alk",
|
|
39
|
+
model: Optional[str] = None,
|
|
40
|
+
api_key: Optional[str] = None,
|
|
41
|
+
api_key_env: Optional[str] = None,
|
|
42
|
+
headers: Optional[Mapping[str, str]] = None,
|
|
43
|
+
timeout: float = 30.0,
|
|
44
|
+
include_tools: bool = True,
|
|
45
|
+
system_prompt: Optional[str] = None,
|
|
46
|
+
metadata: Optional[Mapping[str, Any]] = None,
|
|
47
|
+
) -> None:
|
|
48
|
+
if not endpoint:
|
|
49
|
+
raise ValueError("endpoint is required")
|
|
50
|
+
self.endpoint = endpoint
|
|
51
|
+
self.protocol = _normalize_protocol(protocol)
|
|
52
|
+
self.model = model
|
|
53
|
+
self.api_key = api_key
|
|
54
|
+
self.api_key_env = api_key_env
|
|
55
|
+
self.headers = {str(k): str(v) for k, v in dict(headers or {}).items()}
|
|
56
|
+
self.timeout = float(timeout)
|
|
57
|
+
self.include_tools = bool(include_tools)
|
|
58
|
+
self.system_prompt = system_prompt
|
|
59
|
+
self.metadata = dict(metadata or {})
|
|
60
|
+
|
|
61
|
+
async def call(self, input: AgentInput) -> AgentResponse:
|
|
62
|
+
started = time.time()
|
|
63
|
+
request_payload = self._request_payload(input)
|
|
64
|
+
headers = self._request_headers()
|
|
65
|
+
status_code = 0
|
|
66
|
+
response_payload: dict[str, Any] = {}
|
|
67
|
+
error: Optional[str] = None
|
|
68
|
+
try:
|
|
69
|
+
status_code, response_payload = await asyncio.to_thread(
|
|
70
|
+
self._send_json,
|
|
71
|
+
request_payload,
|
|
72
|
+
headers,
|
|
73
|
+
)
|
|
74
|
+
response = self._agent_response_from_payload(response_payload)
|
|
75
|
+
except Exception as exc:
|
|
76
|
+
error = str(exc)
|
|
77
|
+
response = AgentResponse(content=f"WebSocket target failed: {exc}")
|
|
78
|
+
|
|
79
|
+
latency_ms = round((time.time() - started) * 1000, 4)
|
|
80
|
+
parsed = urlparse(self.endpoint)
|
|
81
|
+
trace = {
|
|
82
|
+
"kind": "external_agent_websocket_trace",
|
|
83
|
+
"protocol": self.protocol,
|
|
84
|
+
"endpoint": _redacted_endpoint(self.endpoint),
|
|
85
|
+
"endpoint_host": parsed.netloc,
|
|
86
|
+
"model": self.model,
|
|
87
|
+
"status_code": status_code,
|
|
88
|
+
"latency_ms": latency_ms,
|
|
89
|
+
"request_message_count": len(input.messages),
|
|
90
|
+
"request_tool_count": len(input.tools) if self.include_tools else 0,
|
|
91
|
+
"response_tool_call_count": len(response.tool_calls or []),
|
|
92
|
+
"success": error is None and status_code == 101,
|
|
93
|
+
"request_header_names": sorted(headers),
|
|
94
|
+
"auth": {
|
|
95
|
+
"mode": "bearer" if self._resolved_api_key() else "none",
|
|
96
|
+
"api_key_env": self.api_key_env,
|
|
97
|
+
"redacted": bool(self._resolved_api_key()),
|
|
98
|
+
},
|
|
99
|
+
"error": error,
|
|
100
|
+
**self.metadata,
|
|
101
|
+
}
|
|
102
|
+
response.events.append(
|
|
103
|
+
SimulationEvent(
|
|
104
|
+
type="external_agent",
|
|
105
|
+
name="external_agent_websocket_call",
|
|
106
|
+
payload=trace,
|
|
107
|
+
)
|
|
108
|
+
)
|
|
109
|
+
response.artifacts.append(
|
|
110
|
+
SimulationArtifact(
|
|
111
|
+
type="trace",
|
|
112
|
+
role="agent",
|
|
113
|
+
data=trace,
|
|
114
|
+
metadata={"kind": "external_agent_websocket_trace"},
|
|
115
|
+
)
|
|
116
|
+
)
|
|
117
|
+
state = dict(response.state or {})
|
|
118
|
+
state["external_agent"] = trace
|
|
119
|
+
state["external_agent_trace"] = trace
|
|
120
|
+
response.state = state
|
|
121
|
+
metadata = dict(response.metadata or {})
|
|
122
|
+
metadata["external_agent"] = trace
|
|
123
|
+
metadata["external_agent_trace"] = trace
|
|
124
|
+
response.metadata = metadata
|
|
125
|
+
return response
|
|
126
|
+
|
|
127
|
+
def _request_payload(self, input: AgentInput) -> dict[str, Any]:
|
|
128
|
+
messages = list(input.messages)
|
|
129
|
+
if self.system_prompt:
|
|
130
|
+
messages = [{"role": "system", "content": self.system_prompt}, *messages]
|
|
131
|
+
return {
|
|
132
|
+
"thread_id": input.thread_id,
|
|
133
|
+
"execution_id": input.execution_id,
|
|
134
|
+
"turn_index": input.turn_index,
|
|
135
|
+
"scenario_name": input.scenario_name,
|
|
136
|
+
"persona": input.persona,
|
|
137
|
+
"situation": input.situation,
|
|
138
|
+
"expected_outcome": input.expected_outcome,
|
|
139
|
+
"messages": messages,
|
|
140
|
+
"new_message": input.new_message,
|
|
141
|
+
"tools": list(input.tools) if self.include_tools else [],
|
|
142
|
+
"metadata": input.metadata,
|
|
143
|
+
}
|
|
144
|
+
|
|
145
|
+
def _request_headers(self) -> dict[str, str]:
|
|
146
|
+
headers = dict(self.headers)
|
|
147
|
+
api_key = self._resolved_api_key()
|
|
148
|
+
if api_key and not any(key.lower() == "authorization" for key in headers):
|
|
149
|
+
headers["Authorization"] = f"Bearer {api_key}"
|
|
150
|
+
return headers
|
|
151
|
+
|
|
152
|
+
def _resolved_api_key(self) -> str:
|
|
153
|
+
if self.api_key not in (None, ""):
|
|
154
|
+
return str(self.api_key)
|
|
155
|
+
if self.api_key_env:
|
|
156
|
+
return os.environ.get(self.api_key_env, "")
|
|
157
|
+
return ""
|
|
158
|
+
|
|
159
|
+
def _send_json(
|
|
160
|
+
self,
|
|
161
|
+
payload: Mapping[str, Any],
|
|
162
|
+
headers: Mapping[str, str],
|
|
163
|
+
) -> tuple[int, dict[str, Any]]:
|
|
164
|
+
parsed = urlparse(self.endpoint)
|
|
165
|
+
if parsed.scheme != "ws":
|
|
166
|
+
raise ValueError("WebSocketAgentWrapper currently supports ws:// endpoints")
|
|
167
|
+
host = parsed.hostname or ""
|
|
168
|
+
port = int(parsed.port or 80)
|
|
169
|
+
path = parsed.path or "/"
|
|
170
|
+
if parsed.query:
|
|
171
|
+
path = f"{path}?{parsed.query}"
|
|
172
|
+
key = base64.b64encode(os.urandom(16)).decode("ascii")
|
|
173
|
+
with socket.create_connection((host, port), timeout=self.timeout) as sock:
|
|
174
|
+
sock.settimeout(self.timeout)
|
|
175
|
+
status = self._handshake(
|
|
176
|
+
sock,
|
|
177
|
+
host=host,
|
|
178
|
+
port=port,
|
|
179
|
+
path=path,
|
|
180
|
+
key=key,
|
|
181
|
+
headers=headers,
|
|
182
|
+
)
|
|
183
|
+
_send_text_frame(sock, json.dumps(payload, default=str), mask=True)
|
|
184
|
+
opcode, text = _read_frame(sock)
|
|
185
|
+
if opcode == 8:
|
|
186
|
+
raise ValueError("WebSocket target closed before returning JSON")
|
|
187
|
+
try:
|
|
188
|
+
_send_close_frame(sock)
|
|
189
|
+
except OSError:
|
|
190
|
+
pass
|
|
191
|
+
try:
|
|
192
|
+
response = json.loads(text or "{}")
|
|
193
|
+
except json.JSONDecodeError as exc:
|
|
194
|
+
raise ValueError(f"WebSocket target returned non-JSON response: {exc}") from exc
|
|
195
|
+
if not isinstance(response, dict):
|
|
196
|
+
raise ValueError("WebSocket target response must be a JSON object")
|
|
197
|
+
return status, response
|
|
198
|
+
|
|
199
|
+
def _handshake(
|
|
200
|
+
self,
|
|
201
|
+
sock: socket.socket,
|
|
202
|
+
*,
|
|
203
|
+
host: str,
|
|
204
|
+
port: int,
|
|
205
|
+
path: str,
|
|
206
|
+
key: str,
|
|
207
|
+
headers: Mapping[str, str],
|
|
208
|
+
) -> int:
|
|
209
|
+
request_headers = {
|
|
210
|
+
"Host": f"{host}:{port}",
|
|
211
|
+
"Upgrade": "websocket",
|
|
212
|
+
"Connection": "Upgrade",
|
|
213
|
+
"Sec-WebSocket-Key": key,
|
|
214
|
+
"Sec-WebSocket-Version": "13",
|
|
215
|
+
**dict(headers),
|
|
216
|
+
}
|
|
217
|
+
request = "GET " + path + " HTTP/1.1\r\n" + "\r\n".join(
|
|
218
|
+
f"{name}: {value}" for name, value in request_headers.items()
|
|
219
|
+
) + "\r\n\r\n"
|
|
220
|
+
sock.sendall(request.encode("utf-8"))
|
|
221
|
+
response = _read_until(sock, b"\r\n\r\n").decode("utf-8", errors="replace")
|
|
222
|
+
lines = response.split("\r\n")
|
|
223
|
+
status_line = lines[0] if lines else ""
|
|
224
|
+
parts = status_line.split()
|
|
225
|
+
status = int(parts[1]) if len(parts) >= 2 and parts[1].isdigit() else 0
|
|
226
|
+
response_headers: dict[str, str] = {}
|
|
227
|
+
for line in lines[1:]:
|
|
228
|
+
if ":" in line:
|
|
229
|
+
name, value = line.split(":", 1)
|
|
230
|
+
response_headers[name.strip().lower()] = value.strip()
|
|
231
|
+
expected_accept = base64.b64encode(
|
|
232
|
+
hashlib.sha1(
|
|
233
|
+
(key + "258EAFA5-E914-47DA-95CA-C5AB0DC85B11").encode("ascii")
|
|
234
|
+
).digest()
|
|
235
|
+
).decode("ascii")
|
|
236
|
+
if status != 101:
|
|
237
|
+
raise ValueError(f"WebSocket target returned status {status}")
|
|
238
|
+
if response_headers.get("sec-websocket-accept") != expected_accept:
|
|
239
|
+
raise ValueError("WebSocket target returned invalid accept key")
|
|
240
|
+
return status
|
|
241
|
+
|
|
242
|
+
def _agent_response_from_payload(self, payload: Mapping[str, Any]) -> AgentResponse:
|
|
243
|
+
return AgentResponse(
|
|
244
|
+
content=_content_text(payload.get("content") or payload.get("message")),
|
|
245
|
+
tool_calls=_tool_call_list(payload.get("tool_calls")),
|
|
246
|
+
tool_responses=_tool_response_list(payload.get("tool_responses")),
|
|
247
|
+
artifacts=_artifact_list(payload.get("artifacts")),
|
|
248
|
+
events=_event_list(payload.get("events")),
|
|
249
|
+
memory_updates=_optional_mapping(payload.get("memory_updates")),
|
|
250
|
+
state=_optional_mapping(payload.get("state")),
|
|
251
|
+
metadata=_optional_mapping(payload.get("metadata")),
|
|
252
|
+
)
|
|
253
|
+
|
|
254
|
+
|
|
255
|
+
def _normalize_protocol(value: str) -> str:
|
|
256
|
+
protocol = str(value or "fi.alk").lower().replace("-", "_")
|
|
257
|
+
aliases = {
|
|
258
|
+
"agent_learning_websocket": "fi.alk",
|
|
259
|
+
"websocket": "fi.alk",
|
|
260
|
+
"ws": "fi.alk",
|
|
261
|
+
}
|
|
262
|
+
protocol = aliases.get(protocol, protocol)
|
|
263
|
+
if protocol != "fi.alk":
|
|
264
|
+
raise ValueError("protocol must be fi.alk")
|
|
265
|
+
return protocol
|
|
266
|
+
|
|
267
|
+
|
|
268
|
+
def _read_until(sock: socket.socket, marker: bytes) -> bytes:
|
|
269
|
+
chunks: list[bytes] = []
|
|
270
|
+
data = b""
|
|
271
|
+
while marker not in data:
|
|
272
|
+
chunk = sock.recv(4096)
|
|
273
|
+
if not chunk:
|
|
274
|
+
break
|
|
275
|
+
chunks.append(chunk)
|
|
276
|
+
data = b"".join(chunks)
|
|
277
|
+
return data
|
|
278
|
+
|
|
279
|
+
|
|
280
|
+
def _send_text_frame(sock: socket.socket, text: str, *, mask: bool) -> None:
|
|
281
|
+
payload = text.encode("utf-8")
|
|
282
|
+
header = bytearray([0x81])
|
|
283
|
+
length = len(payload)
|
|
284
|
+
mask_bit = 0x80 if mask else 0
|
|
285
|
+
if length < 126:
|
|
286
|
+
header.append(mask_bit | length)
|
|
287
|
+
elif length <= 0xFFFF:
|
|
288
|
+
header.extend([mask_bit | 126, *length.to_bytes(2, "big")])
|
|
289
|
+
else:
|
|
290
|
+
header.extend([mask_bit | 127, *length.to_bytes(8, "big")])
|
|
291
|
+
if mask:
|
|
292
|
+
key = os.urandom(4)
|
|
293
|
+
masked = bytes(byte ^ key[index % 4] for index, byte in enumerate(payload))
|
|
294
|
+
sock.sendall(bytes(header) + key + masked)
|
|
295
|
+
else:
|
|
296
|
+
sock.sendall(bytes(header) + payload)
|
|
297
|
+
|
|
298
|
+
|
|
299
|
+
def _send_close_frame(sock: socket.socket) -> None:
|
|
300
|
+
sock.sendall(b"\x88\x80" + os.urandom(4))
|
|
301
|
+
|
|
302
|
+
|
|
303
|
+
def _read_frame(sock: socket.socket) -> tuple[int, str]:
|
|
304
|
+
first = _read_exact(sock, 2)
|
|
305
|
+
opcode = first[0] & 0x0F
|
|
306
|
+
masked = bool(first[1] & 0x80)
|
|
307
|
+
length = first[1] & 0x7F
|
|
308
|
+
if length == 126:
|
|
309
|
+
length = int.from_bytes(_read_exact(sock, 2), "big")
|
|
310
|
+
elif length == 127:
|
|
311
|
+
length = int.from_bytes(_read_exact(sock, 8), "big")
|
|
312
|
+
mask_key = _read_exact(sock, 4) if masked else b""
|
|
313
|
+
payload = _read_exact(sock, length) if length else b""
|
|
314
|
+
if masked:
|
|
315
|
+
payload = bytes(byte ^ mask_key[index % 4] for index, byte in enumerate(payload))
|
|
316
|
+
return opcode, payload.decode("utf-8")
|
|
317
|
+
|
|
318
|
+
|
|
319
|
+
def _read_exact(sock: socket.socket, size: int) -> bytes:
|
|
320
|
+
data = bytearray()
|
|
321
|
+
while len(data) < size:
|
|
322
|
+
chunk = sock.recv(size - len(data))
|
|
323
|
+
if not chunk:
|
|
324
|
+
raise ValueError("WebSocket connection closed unexpectedly")
|
|
325
|
+
data.extend(chunk)
|
|
326
|
+
return bytes(data)
|
|
@@ -0,0 +1,62 @@
|
|
|
1
|
+
from __future__ import annotations
|
|
2
|
+
|
|
3
|
+
from pydantic import BaseModel, Field, JsonValue, model_validator
|
|
4
|
+
|
|
5
|
+
from fi.simulate.agent.wrapper import ArtifactType
|
|
6
|
+
from fi.simulate.evidence import EvidenceClass
|
|
7
|
+
from fi.simulate._hashing import content_hash
|
|
8
|
+
|
|
9
|
+
ARTIFACT_MANIFEST_SCHEMA_VERSION = "futureagi.artifact-manifest.v1"
|
|
10
|
+
|
|
11
|
+
|
|
12
|
+
class ArtifactManifestEntry(BaseModel):
|
|
13
|
+
artifact_id: str
|
|
14
|
+
test_case_id: str | None = None
|
|
15
|
+
type: ArtifactType
|
|
16
|
+
path: str | None = None
|
|
17
|
+
uri: str | None = None
|
|
18
|
+
checksum: str
|
|
19
|
+
size_bytes: int = Field(ge=0)
|
|
20
|
+
mime_type: str | None = None
|
|
21
|
+
codec: str | None = None
|
|
22
|
+
sample_rate: int | None = Field(default=None, gt=0)
|
|
23
|
+
channels: int | None = Field(default=None, gt=0)
|
|
24
|
+
participant_id: str | None = None
|
|
25
|
+
track_id: str | None = None
|
|
26
|
+
leg_id: str | None = None
|
|
27
|
+
start_time_ns: int | None = None
|
|
28
|
+
end_time_ns: int | None = None
|
|
29
|
+
evidence_class: EvidenceClass
|
|
30
|
+
evidence_source_id: str
|
|
31
|
+
redacted: bool = False
|
|
32
|
+
metadata: dict[str, JsonValue] = Field(default_factory=dict)
|
|
33
|
+
|
|
34
|
+
@model_validator(mode="after")
|
|
35
|
+
def _validate_location(self) -> "ArtifactManifestEntry":
|
|
36
|
+
if not self.path and not self.uri:
|
|
37
|
+
raise ValueError("artifact_location_missing: path or uri is required")
|
|
38
|
+
if self.path and self.uri:
|
|
39
|
+
raise ValueError("artifact_location_ambiguous: provide path or uri, not both")
|
|
40
|
+
if not self.checksum.startswith("sha256:"):
|
|
41
|
+
raise ValueError("artifact_checksum_invalid: checksum must use sha256:<digest>")
|
|
42
|
+
return self
|
|
43
|
+
|
|
44
|
+
|
|
45
|
+
class ArtifactManifest(BaseModel):
|
|
46
|
+
schema_version: str = ARTIFACT_MANIFEST_SCHEMA_VERSION
|
|
47
|
+
run_id: str
|
|
48
|
+
entries: list[ArtifactManifestEntry] = Field(default_factory=list)
|
|
49
|
+
manifest_hash: str | None = None
|
|
50
|
+
|
|
51
|
+
def content_hash(self) -> str:
|
|
52
|
+
payload = self.model_dump(exclude={"manifest_hash"}, exclude_none=True)
|
|
53
|
+
payload["entries"] = sorted(payload["entries"], key=lambda item: item["artifact_id"])
|
|
54
|
+
return content_hash(payload)
|
|
55
|
+
|
|
56
|
+
@model_validator(mode="after")
|
|
57
|
+
def _stamp_hash(self) -> "ArtifactManifest":
|
|
58
|
+
expected = self.content_hash()
|
|
59
|
+
if self.manifest_hash is not None and self.manifest_hash != expected:
|
|
60
|
+
raise ValueError("artifact_manifest_hash_mismatch")
|
|
61
|
+
object.__setattr__(self, "manifest_hash", expected)
|
|
62
|
+
return self
|