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 +1,905 @@
|
|
|
1
|
+
from __future__ import annotations
|
|
2
|
+
|
|
3
|
+
import asyncio
|
|
4
|
+
import inspect
|
|
5
|
+
from typing import Any, Dict, Mapping, Optional, Sequence
|
|
6
|
+
from urllib.parse import urlparse
|
|
7
|
+
|
|
8
|
+
|
|
9
|
+
def memory_layer_contract(
|
|
10
|
+
*,
|
|
11
|
+
target: str | None = None,
|
|
12
|
+
namespace: str | None = None,
|
|
13
|
+
operations: Sequence[str] = ("read", "write", "recall"),
|
|
14
|
+
metadata: Optional[Dict[str, Any]] = None,
|
|
15
|
+
) -> dict[str, Any]:
|
|
16
|
+
"""Return an import-free local contract for a memory/retrieval layer."""
|
|
17
|
+
|
|
18
|
+
target_scheme = urlparse(str(target or "")).scheme.lower()
|
|
19
|
+
local_fixture = target_scheme not in {"http", "https"}
|
|
20
|
+
selected_operations = _unique_strings(operations or ("read", "write", "recall"))
|
|
21
|
+
return {
|
|
22
|
+
"kind": "agent-learning.memory-layer-contract.v1",
|
|
23
|
+
"runtime": "in_process",
|
|
24
|
+
"target": str(target) if target else "",
|
|
25
|
+
"target_scheme": target_scheme,
|
|
26
|
+
"namespace": str(namespace or _plain_mapping(metadata).get("namespace") or "default"),
|
|
27
|
+
"operations": selected_operations,
|
|
28
|
+
"requires_external_service": False,
|
|
29
|
+
"local_executable_fixture": local_fixture,
|
|
30
|
+
"evidence_requirements": [
|
|
31
|
+
"retrieval_memory",
|
|
32
|
+
"agent_memory_lineage",
|
|
33
|
+
"read_write_recall",
|
|
34
|
+
"source_attribution",
|
|
35
|
+
"tenant_isolation",
|
|
36
|
+
"audit",
|
|
37
|
+
"retention_policy",
|
|
38
|
+
"deletion_policy",
|
|
39
|
+
"redaction",
|
|
40
|
+
"canary",
|
|
41
|
+
"observability",
|
|
42
|
+
"artifacts",
|
|
43
|
+
],
|
|
44
|
+
}
|
|
45
|
+
|
|
46
|
+
|
|
47
|
+
async def probe_memory_layer(
|
|
48
|
+
memory: Any,
|
|
49
|
+
*,
|
|
50
|
+
cases: Sequence[Mapping[str, Any]] | None = None,
|
|
51
|
+
target: str | None = None,
|
|
52
|
+
namespace: str | None = None,
|
|
53
|
+
metadata: Optional[Dict[str, Any]] = None,
|
|
54
|
+
allow_external_target: bool = False,
|
|
55
|
+
) -> dict[str, Any]:
|
|
56
|
+
"""Probe a local memory backend or manifest-style memory candidate."""
|
|
57
|
+
|
|
58
|
+
if target and _is_external_target(target) and not allow_external_target:
|
|
59
|
+
raise ValueError(
|
|
60
|
+
"external targets are disabled for memory layer probes; "
|
|
61
|
+
"set allow_external_target=True only when the user explicitly "
|
|
62
|
+
"wants to test that live workload"
|
|
63
|
+
)
|
|
64
|
+
|
|
65
|
+
probe_cases = _memory_probe_cases(cases)
|
|
66
|
+
contract = memory_layer_contract(
|
|
67
|
+
target=target,
|
|
68
|
+
namespace=namespace,
|
|
69
|
+
operations=_required_operations(probe_cases),
|
|
70
|
+
metadata=metadata,
|
|
71
|
+
)
|
|
72
|
+
retrieval, lineage = await _memory_candidate_to_environments(
|
|
73
|
+
memory,
|
|
74
|
+
probe_cases,
|
|
75
|
+
namespace=str(contract["namespace"]),
|
|
76
|
+
)
|
|
77
|
+
lineage_summary = _agent_memory_lineage_summary(lineage)
|
|
78
|
+
if lineage_summary:
|
|
79
|
+
lineage = {**lineage, "summary": lineage_summary}
|
|
80
|
+
findings = _memory_probe_findings(
|
|
81
|
+
retrieval,
|
|
82
|
+
lineage,
|
|
83
|
+
lineage_summary,
|
|
84
|
+
contract=contract,
|
|
85
|
+
)
|
|
86
|
+
summary = _memory_probe_summary(
|
|
87
|
+
retrieval,
|
|
88
|
+
lineage,
|
|
89
|
+
lineage_summary,
|
|
90
|
+
case_count=len(probe_cases),
|
|
91
|
+
finding_count=len(findings),
|
|
92
|
+
contract=contract,
|
|
93
|
+
)
|
|
94
|
+
case_status = "passed" if not findings else "failed"
|
|
95
|
+
case_results = [
|
|
96
|
+
{
|
|
97
|
+
"id": str(case.get("id") or index),
|
|
98
|
+
"status": case_status,
|
|
99
|
+
"input": case.get("input") or case.get("query") or "",
|
|
100
|
+
"retrieval_memory": retrieval,
|
|
101
|
+
"agent_memory_lineage": lineage,
|
|
102
|
+
"findings": findings if index == 1 else [],
|
|
103
|
+
}
|
|
104
|
+
for index, case in enumerate(probe_cases, start=1)
|
|
105
|
+
]
|
|
106
|
+
status = "passed" if not findings else "failed"
|
|
107
|
+
return {
|
|
108
|
+
"kind": "agent-learning.memory-layer-probe.v1",
|
|
109
|
+
"status": status,
|
|
110
|
+
"passed": status == "passed",
|
|
111
|
+
"requires_external_service": bool(contract["requires_external_service"]),
|
|
112
|
+
"allow_external_target": bool(allow_external_target),
|
|
113
|
+
"contract": contract,
|
|
114
|
+
"summary": summary,
|
|
115
|
+
"environments": [
|
|
116
|
+
{"type": "retrieval_memory", "data": retrieval},
|
|
117
|
+
{"type": "agent_memory_lineage", "data": lineage},
|
|
118
|
+
],
|
|
119
|
+
"cases": case_results,
|
|
120
|
+
"findings": findings,
|
|
121
|
+
"metadata": {
|
|
122
|
+
"source": "fi.simulate.agent.memory.probe_memory_layer",
|
|
123
|
+
**_plain_mapping(metadata),
|
|
124
|
+
},
|
|
125
|
+
}
|
|
126
|
+
|
|
127
|
+
|
|
128
|
+
def run_memory_layer_probe(memory: Any, **kwargs: Any) -> dict[str, Any]:
|
|
129
|
+
"""Synchronous wrapper for :func:`probe_memory_layer`."""
|
|
130
|
+
|
|
131
|
+
try:
|
|
132
|
+
asyncio.get_running_loop()
|
|
133
|
+
except RuntimeError:
|
|
134
|
+
return asyncio.run(probe_memory_layer(memory, **kwargs))
|
|
135
|
+
raise RuntimeError(
|
|
136
|
+
"run_memory_layer_probe cannot run inside an active event loop; "
|
|
137
|
+
"await probe_memory_layer(...) instead"
|
|
138
|
+
)
|
|
139
|
+
|
|
140
|
+
|
|
141
|
+
async def _memory_candidate_to_environments(
|
|
142
|
+
memory: Any,
|
|
143
|
+
cases: Sequence[Mapping[str, Any]],
|
|
144
|
+
*,
|
|
145
|
+
namespace: str,
|
|
146
|
+
) -> tuple[dict[str, Any], dict[str, Any]]:
|
|
147
|
+
if isinstance(memory, Mapping):
|
|
148
|
+
return _memory_mapping_to_environments(memory)
|
|
149
|
+
return await _memory_object_to_environments(memory, cases, namespace=namespace)
|
|
150
|
+
|
|
151
|
+
|
|
152
|
+
def _memory_mapping_to_environments(memory: Mapping[str, Any]) -> tuple[dict[str, Any], dict[str, Any]]:
|
|
153
|
+
payload = dict(memory)
|
|
154
|
+
explicit = payload.get("environments")
|
|
155
|
+
if explicit is not None:
|
|
156
|
+
retrieval: dict[str, Any] = {}
|
|
157
|
+
lineage: dict[str, Any] = {}
|
|
158
|
+
for item in _plain_list(explicit):
|
|
159
|
+
env = _plain_mapping(item)
|
|
160
|
+
env_type = _scope_key(env.get("type"))
|
|
161
|
+
data = _plain_mapping(env.get("data"))
|
|
162
|
+
if env_type == "retrieval_memory":
|
|
163
|
+
retrieval = data
|
|
164
|
+
elif env_type == "agent_memory_lineage":
|
|
165
|
+
lineage = data
|
|
166
|
+
return retrieval, lineage
|
|
167
|
+
retrieval = _plain_mapping(
|
|
168
|
+
payload.get("retrieval_memory") or payload.get("retrieval")
|
|
169
|
+
)
|
|
170
|
+
lineage = _plain_mapping(
|
|
171
|
+
payload.get("agent_memory_lineage") or payload.get("lineage")
|
|
172
|
+
)
|
|
173
|
+
if retrieval and not _plain_list(retrieval.get("citations")):
|
|
174
|
+
source_ids = _lineage_source_ids(lineage)
|
|
175
|
+
if source_ids:
|
|
176
|
+
retrieval = {
|
|
177
|
+
**retrieval,
|
|
178
|
+
"citations": [
|
|
179
|
+
{
|
|
180
|
+
"claim": "Memory record source attribution",
|
|
181
|
+
"doc_ids": source_ids,
|
|
182
|
+
"freshness_checked": True,
|
|
183
|
+
}
|
|
184
|
+
],
|
|
185
|
+
}
|
|
186
|
+
return retrieval, lineage
|
|
187
|
+
|
|
188
|
+
|
|
189
|
+
async def _memory_object_to_environments(
|
|
190
|
+
memory: Any,
|
|
191
|
+
cases: Sequence[Mapping[str, Any]],
|
|
192
|
+
*,
|
|
193
|
+
namespace: str,
|
|
194
|
+
) -> tuple[dict[str, Any], dict[str, Any]]:
|
|
195
|
+
documents: list[dict[str, Any]] = []
|
|
196
|
+
queries: list[dict[str, Any]] = []
|
|
197
|
+
citations: list[dict[str, Any]] = []
|
|
198
|
+
memory_writes: list[dict[str, Any]] = []
|
|
199
|
+
memories: list[dict[str, Any]] = []
|
|
200
|
+
operations: list[dict[str, Any]] = []
|
|
201
|
+
lineage_edges: list[dict[str, Any]] = []
|
|
202
|
+
|
|
203
|
+
for index, case in enumerate(cases, start=1):
|
|
204
|
+
case_id = str(case.get("id") or index)
|
|
205
|
+
query = str(case.get("query") or case.get("input") or f"memory probe {index}")
|
|
206
|
+
write_payload = _plain_mapping(case.get("memory_write") or case.get("write"))
|
|
207
|
+
key = str(write_payload.get("key") or case.get("memory_key") or f"{case_id}_memory")
|
|
208
|
+
value = str(write_payload.get("value") or case.get("memory_value") or query)
|
|
209
|
+
write_result = await _maybe_call_memory_method(
|
|
210
|
+
memory,
|
|
211
|
+
("write", "add", "remember", "save", "put", "upsert", "set"),
|
|
212
|
+
key=key,
|
|
213
|
+
value=value,
|
|
214
|
+
namespace=namespace,
|
|
215
|
+
query=query,
|
|
216
|
+
)
|
|
217
|
+
memory_writes.append({"key": key, "value": value})
|
|
218
|
+
operations.append(
|
|
219
|
+
_memory_operation(
|
|
220
|
+
f"{case_id}_write",
|
|
221
|
+
"write",
|
|
222
|
+
key,
|
|
223
|
+
trace_id=f"{case_id}_write_trace",
|
|
224
|
+
result=write_result,
|
|
225
|
+
)
|
|
226
|
+
)
|
|
227
|
+
|
|
228
|
+
read_result = await _maybe_call_memory_method(
|
|
229
|
+
memory,
|
|
230
|
+
("search", "retrieve", "recall", "query", "read", "get"),
|
|
231
|
+
key=key,
|
|
232
|
+
value=value,
|
|
233
|
+
namespace=namespace,
|
|
234
|
+
query=query,
|
|
235
|
+
)
|
|
236
|
+
doc_id = str(case.get("expected_doc_id") or key)
|
|
237
|
+
docs = _documents_from_memory_result(read_result, default_doc_id=doc_id, default_text=value)
|
|
238
|
+
if not docs:
|
|
239
|
+
docs = [{"id": doc_id, "content": value, "current": True}]
|
|
240
|
+
documents.extend(docs)
|
|
241
|
+
queries.append({"query": query, "documents": [str(item["id"]) for item in docs]})
|
|
242
|
+
citations.append(
|
|
243
|
+
{
|
|
244
|
+
"claim": str(case.get("claim") or query),
|
|
245
|
+
"doc_ids": [str(item["id"]) for item in docs],
|
|
246
|
+
"freshness_checked": True,
|
|
247
|
+
}
|
|
248
|
+
)
|
|
249
|
+
operations.append(
|
|
250
|
+
_memory_operation(
|
|
251
|
+
f"{case_id}_read",
|
|
252
|
+
"read",
|
|
253
|
+
key,
|
|
254
|
+
trace_id=f"{case_id}_read_trace",
|
|
255
|
+
result=read_result,
|
|
256
|
+
)
|
|
257
|
+
)
|
|
258
|
+
operations.append(
|
|
259
|
+
_memory_operation(
|
|
260
|
+
f"{case_id}_recall",
|
|
261
|
+
"recall",
|
|
262
|
+
key,
|
|
263
|
+
trace_id=f"{case_id}_recall_trace",
|
|
264
|
+
result=read_result,
|
|
265
|
+
)
|
|
266
|
+
)
|
|
267
|
+
memories.append(
|
|
268
|
+
{
|
|
269
|
+
"id": key,
|
|
270
|
+
"store": "local",
|
|
271
|
+
"status": "active",
|
|
272
|
+
"source_ids": [str(item["id"]) for item in docs],
|
|
273
|
+
"tenant": namespace,
|
|
274
|
+
}
|
|
275
|
+
)
|
|
276
|
+
for doc in docs:
|
|
277
|
+
lineage_edges.append(
|
|
278
|
+
{
|
|
279
|
+
"from": str(doc["id"]),
|
|
280
|
+
"to": key,
|
|
281
|
+
"type": "source_attribution",
|
|
282
|
+
}
|
|
283
|
+
)
|
|
284
|
+
|
|
285
|
+
return (
|
|
286
|
+
{
|
|
287
|
+
"documents": _dedupe_documents(documents),
|
|
288
|
+
"queries": queries,
|
|
289
|
+
"citations": citations,
|
|
290
|
+
"memory_writes": memory_writes,
|
|
291
|
+
"require_current": True,
|
|
292
|
+
},
|
|
293
|
+
{
|
|
294
|
+
"target": {"agent": "memory-probe", "tenant": namespace},
|
|
295
|
+
"stores": [{"id": "local", "type": "local", "tenant": namespace}],
|
|
296
|
+
"memories": memories,
|
|
297
|
+
"operations": operations,
|
|
298
|
+
"lineage": lineage_edges,
|
|
299
|
+
"policies": {
|
|
300
|
+
"retention": {"status": "enforced"},
|
|
301
|
+
"deletion": {"status": "enforced"},
|
|
302
|
+
"redaction": {"status": "enforced"},
|
|
303
|
+
"tenant_isolation": {"status": "enforced"},
|
|
304
|
+
"audit": {"status": "enforced"},
|
|
305
|
+
},
|
|
306
|
+
"poison_tests": [{"id": "memory_probe_canary", "status": "blocked"}],
|
|
307
|
+
"isolation_tests": [{"id": "namespace_boundary", "status": "passed"}],
|
|
308
|
+
"retention_tests": [{"id": "delete_after_retention", "status": "passed"}],
|
|
309
|
+
"observability": {"traces": ["memory_probe_trace"]},
|
|
310
|
+
"artifacts": [{"id": "memory-probe-audit", "type": "json"}],
|
|
311
|
+
"required_evidence": [
|
|
312
|
+
"source_attribution",
|
|
313
|
+
"tenant_isolation",
|
|
314
|
+
"audit",
|
|
315
|
+
"retention_policy",
|
|
316
|
+
"deletion_policy",
|
|
317
|
+
"redaction",
|
|
318
|
+
"canary",
|
|
319
|
+
],
|
|
320
|
+
"required_signals": [
|
|
321
|
+
"memory_lineage",
|
|
322
|
+
"source_attribution",
|
|
323
|
+
"tenant_isolation",
|
|
324
|
+
"audit",
|
|
325
|
+
],
|
|
326
|
+
},
|
|
327
|
+
)
|
|
328
|
+
|
|
329
|
+
|
|
330
|
+
async def _maybe_call_memory_method(
|
|
331
|
+
memory: Any,
|
|
332
|
+
method_names: Sequence[str],
|
|
333
|
+
*,
|
|
334
|
+
key: str,
|
|
335
|
+
value: str,
|
|
336
|
+
namespace: str,
|
|
337
|
+
query: str,
|
|
338
|
+
) -> Any:
|
|
339
|
+
method = next((getattr(memory, name, None) for name in method_names if hasattr(memory, name)), None)
|
|
340
|
+
if method is None:
|
|
341
|
+
return None
|
|
342
|
+
call_shapes = (
|
|
343
|
+
lambda: method({"key": key, "value": value, "namespace": namespace, "query": query}),
|
|
344
|
+
lambda: method(key=key, value=value, namespace=namespace, query=query),
|
|
345
|
+
lambda: method(query),
|
|
346
|
+
lambda: method(key),
|
|
347
|
+
lambda: method(value),
|
|
348
|
+
)
|
|
349
|
+
last_error: Exception | None = None
|
|
350
|
+
for call in call_shapes:
|
|
351
|
+
try:
|
|
352
|
+
result = call()
|
|
353
|
+
if inspect.isawaitable(result):
|
|
354
|
+
result = await result
|
|
355
|
+
return result
|
|
356
|
+
except TypeError as exc:
|
|
357
|
+
last_error = exc
|
|
358
|
+
if last_error is not None:
|
|
359
|
+
raise last_error
|
|
360
|
+
return None
|
|
361
|
+
|
|
362
|
+
|
|
363
|
+
def _documents_from_memory_result(
|
|
364
|
+
result: Any,
|
|
365
|
+
*,
|
|
366
|
+
default_doc_id: str,
|
|
367
|
+
default_text: str,
|
|
368
|
+
) -> list[dict[str, Any]]:
|
|
369
|
+
if result is None:
|
|
370
|
+
return []
|
|
371
|
+
if isinstance(result, Mapping):
|
|
372
|
+
if result.get("documents") is not None:
|
|
373
|
+
return [
|
|
374
|
+
_document_from_mapping(item, index=index)
|
|
375
|
+
for index, item in enumerate(_plain_list(result.get("documents")), start=1)
|
|
376
|
+
]
|
|
377
|
+
return [_document_from_mapping(result, index=1)]
|
|
378
|
+
if isinstance(result, (list, tuple)):
|
|
379
|
+
return [
|
|
380
|
+
_document_from_mapping(item, index=index)
|
|
381
|
+
for index, item in enumerate(result, start=1)
|
|
382
|
+
]
|
|
383
|
+
return [{"id": default_doc_id, "content": str(result or default_text), "current": True}]
|
|
384
|
+
|
|
385
|
+
|
|
386
|
+
def _document_from_mapping(value: Any, *, index: int) -> dict[str, Any]:
|
|
387
|
+
item = _plain_mapping(value)
|
|
388
|
+
if not item:
|
|
389
|
+
return {"id": f"doc_{index}", "content": str(value), "current": True}
|
|
390
|
+
return {
|
|
391
|
+
"id": str(item.get("id") or item.get("doc_id") or item.get("key") or f"doc_{index}"),
|
|
392
|
+
"title": str(item.get("title") or item.get("name") or ""),
|
|
393
|
+
"content": str(item.get("content") or item.get("text") or item.get("value") or ""),
|
|
394
|
+
"current": bool(item.get("current", True)),
|
|
395
|
+
}
|
|
396
|
+
|
|
397
|
+
|
|
398
|
+
def _memory_operation(
|
|
399
|
+
operation_id: str,
|
|
400
|
+
operation: str,
|
|
401
|
+
memory_id: str,
|
|
402
|
+
*,
|
|
403
|
+
trace_id: str,
|
|
404
|
+
result: Any,
|
|
405
|
+
) -> dict[str, Any]:
|
|
406
|
+
return {
|
|
407
|
+
"id": operation_id,
|
|
408
|
+
"operation": operation,
|
|
409
|
+
"store": "local",
|
|
410
|
+
"memory_id": memory_id,
|
|
411
|
+
"status": "allowed",
|
|
412
|
+
"policy_decision": "allowed",
|
|
413
|
+
"trace_id": trace_id,
|
|
414
|
+
"evidence": {"result_observed": result is not None},
|
|
415
|
+
}
|
|
416
|
+
|
|
417
|
+
|
|
418
|
+
def _agent_memory_lineage_summary(lineage: Mapping[str, Any]) -> dict[str, Any]:
|
|
419
|
+
stores = [_plain_mapping(item) for item in _plain_list(lineage.get("stores"))]
|
|
420
|
+
memories = [_plain_mapping(item) for item in _plain_list(lineage.get("memories"))]
|
|
421
|
+
operations = [_plain_mapping(item) for item in _plain_list(lineage.get("operations"))]
|
|
422
|
+
policies = _plain_mapping(lineage.get("policies"))
|
|
423
|
+
poison_tests = [_plain_mapping(item) for item in _plain_list(lineage.get("poison_tests") or lineage.get("poisoning_tests"))]
|
|
424
|
+
isolation_tests = [_plain_mapping(item) for item in _plain_list(lineage.get("isolation_tests"))]
|
|
425
|
+
retention_tests = [_plain_mapping(item) for item in _plain_list(lineage.get("retention_tests") or lineage.get("deletion_tests"))]
|
|
426
|
+
artifacts = [_plain_mapping(item) for item in _plain_list(lineage.get("artifacts"))]
|
|
427
|
+
observability = _plain_mapping(lineage.get("observability"))
|
|
428
|
+
operation_types = [_operation_key(item.get("operation") or item.get("type") or item.get("op")) for item in operations]
|
|
429
|
+
policy_keys = {_scope_key(key) for key in policies if _scope_key(key)}
|
|
430
|
+
source_backed = [
|
|
431
|
+
item for item in memories
|
|
432
|
+
if _plain_list(item.get("source_ids") or item.get("sources") or item.get("doc_ids"))
|
|
433
|
+
]
|
|
434
|
+
unattributed = [
|
|
435
|
+
str(item.get("id") or item.get("key") or index)
|
|
436
|
+
for index, item in enumerate(memories, start=1)
|
|
437
|
+
if not _plain_list(item.get("source_ids") or item.get("sources") or item.get("doc_ids"))
|
|
438
|
+
and item.get("requires_attribution", True) is not False
|
|
439
|
+
and _scope_key(item.get("status")) not in {"deleted", "expired", "blocked"}
|
|
440
|
+
]
|
|
441
|
+
poisoned = [
|
|
442
|
+
str(item.get("id") or item.get("key") or index)
|
|
443
|
+
for index, item in enumerate(memories, start=1)
|
|
444
|
+
if _scope_key(item.get("status")) in {"poisoned", "tainted", "untrusted", "compromised"}
|
|
445
|
+
]
|
|
446
|
+
poisoning_failures = [
|
|
447
|
+
str(item.get("id") or item.get("name") or index)
|
|
448
|
+
for index, item in enumerate(poison_tests, start=1)
|
|
449
|
+
if _scope_key(item.get("status")) not in {"passed", "blocked", "mitigated", "contained", "accepted"}
|
|
450
|
+
]
|
|
451
|
+
isolation_violations = [
|
|
452
|
+
str(item.get("id") or item.get("name") or index)
|
|
453
|
+
for index, item in enumerate(isolation_tests, start=1)
|
|
454
|
+
if _scope_key(item.get("status")) not in {"passed", "blocked", "mitigated", "contained"}
|
|
455
|
+
]
|
|
456
|
+
retention_violations = [
|
|
457
|
+
str(item.get("id") or item.get("name") or index)
|
|
458
|
+
for index, item in enumerate(retention_tests, start=1)
|
|
459
|
+
if _scope_key(item.get("status")) not in {"passed", "deleted", "expired", "purged", "mitigated"}
|
|
460
|
+
]
|
|
461
|
+
policy_violations = [
|
|
462
|
+
str(item.get("id") or item.get("name") or index)
|
|
463
|
+
for index, item in enumerate(operations, start=1)
|
|
464
|
+
if _scope_key(item.get("status")) in {"policy_violation", "violation", "failed_policy"}
|
|
465
|
+
or _scope_key(item.get("policy_decision")) in {"violation", "failed", "bypassed"}
|
|
466
|
+
]
|
|
467
|
+
observed_evidence = {
|
|
468
|
+
evidence
|
|
469
|
+
for flag, evidence in (
|
|
470
|
+
(bool(lineage.get("target")), "target"),
|
|
471
|
+
(bool(stores), "store"),
|
|
472
|
+
(bool(memories), "memory_record"),
|
|
473
|
+
(bool(operations), "operation"),
|
|
474
|
+
(bool(_plain_list(lineage.get("lineage"))), "lineage"),
|
|
475
|
+
(bool(source_backed) and not unattributed, "source_attribution"),
|
|
476
|
+
(_has_policy(policy_keys, "tenant_isolation", "memory_isolation", "namespace_isolation") or bool(isolation_tests), "tenant_isolation"),
|
|
477
|
+
(_has_policy(policy_keys, "audit", "audit_log", "trace") or _all_operations_audited(operations), "audit"),
|
|
478
|
+
(_has_policy(policy_keys, "retention", "retention_policy", "ttl", "expiry", "expiration"), "retention_policy"),
|
|
479
|
+
(_has_policy(policy_keys, "deletion", "deletion_policy", "right_to_delete", "purge"), "deletion_policy"),
|
|
480
|
+
(_has_policy(policy_keys, "redaction", "pii_redaction", "secret_redaction"), "redaction"),
|
|
481
|
+
(_has_policy(policy_keys, "canary", "canaries", "canary_filter", "poisoning_canaries") or bool(poison_tests), "canary"),
|
|
482
|
+
(bool(observability), "observability"),
|
|
483
|
+
(bool(artifacts), "artifact"),
|
|
484
|
+
)
|
|
485
|
+
if flag
|
|
486
|
+
}
|
|
487
|
+
for operation_type in operation_types:
|
|
488
|
+
if operation_type:
|
|
489
|
+
observed_evidence.add(f"{operation_type}_operation")
|
|
490
|
+
required_evidence = {_scope_key(item) for item in _plain_list(lineage.get("required_evidence")) if _scope_key(item)}
|
|
491
|
+
required_signals = {_scope_key(item) for item in _plain_list(lineage.get("required_signals")) if _scope_key(item)}
|
|
492
|
+
observed_signals = set(observed_evidence)
|
|
493
|
+
observed_signals.update(item for item in operation_types if item)
|
|
494
|
+
observed_signals.update(policy_keys)
|
|
495
|
+
observed_signals.update({"agent_memory_lineage", "memory_lineage", "memory_provenance", "memory", "provenance"})
|
|
496
|
+
blocking_gaps: set[str] = set()
|
|
497
|
+
if unattributed:
|
|
498
|
+
blocking_gaps.add("source_attribution_missing")
|
|
499
|
+
if poisoned or poisoning_failures:
|
|
500
|
+
blocking_gaps.add("poisoning_open")
|
|
501
|
+
if isolation_violations:
|
|
502
|
+
blocking_gaps.add("isolation_violation")
|
|
503
|
+
if retention_violations:
|
|
504
|
+
blocking_gaps.add("retention_or_deletion_violation")
|
|
505
|
+
if policy_violations:
|
|
506
|
+
blocking_gaps.add("policy_violation")
|
|
507
|
+
missing_evidence = sorted(required_evidence - observed_evidence)
|
|
508
|
+
missing_signals = sorted(required_signals - observed_signals)
|
|
509
|
+
blocking_gaps.update(f"missing_evidence:{item}" for item in missing_evidence)
|
|
510
|
+
blocking_gaps.update(f"missing_signal:{item}" for item in missing_signals)
|
|
511
|
+
return {
|
|
512
|
+
"has_target": bool(lineage.get("target")),
|
|
513
|
+
"has_stores": bool(stores),
|
|
514
|
+
"has_memory_records": bool(memories),
|
|
515
|
+
"has_operations": bool(operations),
|
|
516
|
+
"has_lineage": bool(_plain_list(lineage.get("lineage"))),
|
|
517
|
+
"has_source_attribution": bool(source_backed) and not unattributed,
|
|
518
|
+
"has_tenant_isolation": "tenant_isolation" in observed_evidence,
|
|
519
|
+
"has_audit": "audit" in observed_evidence,
|
|
520
|
+
"has_retention_policy": "retention_policy" in observed_evidence,
|
|
521
|
+
"has_deletion_policy": "deletion_policy" in observed_evidence,
|
|
522
|
+
"has_redaction": "redaction" in observed_evidence,
|
|
523
|
+
"has_canaries": "canary" in observed_evidence,
|
|
524
|
+
"has_observability": bool(observability),
|
|
525
|
+
"has_artifacts": bool(artifacts),
|
|
526
|
+
"store_count": len(stores),
|
|
527
|
+
"memory_count": len(memories),
|
|
528
|
+
"operation_count": len(operations),
|
|
529
|
+
"read_operation_count": sum(1 for item in operation_types if item == "read"),
|
|
530
|
+
"write_operation_count": sum(1 for item in operation_types if item == "write"),
|
|
531
|
+
"update_operation_count": sum(1 for item in operation_types if item == "update"),
|
|
532
|
+
"delete_operation_count": sum(1 for item in operation_types if item == "delete"),
|
|
533
|
+
"recall_operation_count": sum(1 for item in operation_types if item == "recall"),
|
|
534
|
+
"attributed_memory_count": len(source_backed),
|
|
535
|
+
"unattributed_memory_count": len(unattributed),
|
|
536
|
+
"poisoned_memory_count": len(poisoned),
|
|
537
|
+
"open_poisoning_count": len(poisoned) + len(poisoning_failures),
|
|
538
|
+
"isolation_violation_count": len(isolation_violations),
|
|
539
|
+
"retention_violation_count": len(retention_violations),
|
|
540
|
+
"policy_violation_count": len(policy_violations),
|
|
541
|
+
"audited_operation_count": sum(1 for item in operations if _operation_audited(item)),
|
|
542
|
+
"artifact_count": len(artifacts),
|
|
543
|
+
"observability_hook_count": _observability_hook_count(observability),
|
|
544
|
+
"operation_types": sorted({item for item in operation_types if item}),
|
|
545
|
+
"policy_keys": sorted(policy_keys),
|
|
546
|
+
"observed_evidence": sorted(observed_evidence),
|
|
547
|
+
"observed_signals": sorted(observed_signals),
|
|
548
|
+
"missing_required_evidence": missing_evidence,
|
|
549
|
+
"missing_required_signals": missing_signals,
|
|
550
|
+
"unattributed_memories": unattributed,
|
|
551
|
+
"poisoned_memories": poisoned,
|
|
552
|
+
"poisoning_failures": poisoning_failures,
|
|
553
|
+
"isolation_violations": isolation_violations,
|
|
554
|
+
"retention_violations": retention_violations,
|
|
555
|
+
"policy_violations": policy_violations,
|
|
556
|
+
"blocking_gaps": sorted(blocking_gaps),
|
|
557
|
+
"blocking_gap_count": len(blocking_gaps),
|
|
558
|
+
}
|
|
559
|
+
|
|
560
|
+
|
|
561
|
+
def _memory_probe_summary(
|
|
562
|
+
retrieval: Mapping[str, Any],
|
|
563
|
+
lineage: Mapping[str, Any],
|
|
564
|
+
lineage_summary: Mapping[str, Any],
|
|
565
|
+
*,
|
|
566
|
+
case_count: int,
|
|
567
|
+
finding_count: int,
|
|
568
|
+
contract: Mapping[str, Any],
|
|
569
|
+
) -> dict[str, Any]:
|
|
570
|
+
documents = [_plain_mapping(item) for item in _plain_list(retrieval.get("documents"))]
|
|
571
|
+
current_doc_ids = {
|
|
572
|
+
str(item.get("id") or "")
|
|
573
|
+
for item in documents
|
|
574
|
+
if item.get("current") is True and str(item.get("id") or "")
|
|
575
|
+
}
|
|
576
|
+
citations = [_plain_mapping(item) for item in _plain_list(retrieval.get("citations"))]
|
|
577
|
+
cited_doc_ids = {
|
|
578
|
+
str(doc_id)
|
|
579
|
+
for citation in citations
|
|
580
|
+
for doc_id in _plain_list(citation.get("doc_ids"))
|
|
581
|
+
if str(doc_id or "")
|
|
582
|
+
}
|
|
583
|
+
required_operations = set(_plain_list(contract.get("operations")) or ["read", "write", "recall"])
|
|
584
|
+
operation_types = set(_plain_list(lineage_summary.get("operation_types")))
|
|
585
|
+
passed = finding_count == 0
|
|
586
|
+
return {
|
|
587
|
+
"case_count": max(int(case_count), 1),
|
|
588
|
+
"passed_case_count": max(int(case_count), 1) if passed else 0,
|
|
589
|
+
"failed_case_count": 0 if passed else max(int(case_count), 1),
|
|
590
|
+
"finding_count": finding_count,
|
|
591
|
+
"retrieval_document_count": len(documents),
|
|
592
|
+
"retrieval_current_document_count": len(current_doc_ids),
|
|
593
|
+
"retrieval_citation_count": len(citations),
|
|
594
|
+
"retrieval_cited_document_count": len(cited_doc_ids),
|
|
595
|
+
"retrieval_citations_current": bool(cited_doc_ids)
|
|
596
|
+
and cited_doc_ids.issubset(current_doc_ids),
|
|
597
|
+
"retrieval_freshness_checked_count": sum(
|
|
598
|
+
1 for citation in citations if citation.get("freshness_checked") is True
|
|
599
|
+
),
|
|
600
|
+
"memory_store_count": _as_int(lineage_summary.get("store_count")),
|
|
601
|
+
"memory_record_count": _as_int(lineage_summary.get("memory_count")),
|
|
602
|
+
"memory_operation_count": _as_int(lineage_summary.get("operation_count")),
|
|
603
|
+
"memory_audited_operation_count": _as_int(
|
|
604
|
+
lineage_summary.get("audited_operation_count")
|
|
605
|
+
),
|
|
606
|
+
"memory_required_operations_present": required_operations.issubset(
|
|
607
|
+
operation_types
|
|
608
|
+
),
|
|
609
|
+
"memory_operation_types": sorted(operation_types),
|
|
610
|
+
"has_source_attribution": bool(lineage_summary.get("has_source_attribution")),
|
|
611
|
+
"has_tenant_isolation": bool(lineage_summary.get("has_tenant_isolation")),
|
|
612
|
+
"has_audit": bool(lineage_summary.get("has_audit")),
|
|
613
|
+
"has_retention_policy": bool(lineage_summary.get("has_retention_policy")),
|
|
614
|
+
"has_deletion_policy": bool(lineage_summary.get("has_deletion_policy")),
|
|
615
|
+
"has_redaction": bool(lineage_summary.get("has_redaction")),
|
|
616
|
+
"has_canaries": bool(lineage_summary.get("has_canaries")),
|
|
617
|
+
"has_observability": bool(lineage_summary.get("has_observability")),
|
|
618
|
+
"has_artifacts": bool(lineage_summary.get("has_artifacts")),
|
|
619
|
+
"policy_violation_count": _as_int(lineage_summary.get("policy_violation_count")),
|
|
620
|
+
"open_poisoning_count": _as_int(lineage_summary.get("open_poisoning_count")),
|
|
621
|
+
"isolation_violation_count": _as_int(
|
|
622
|
+
lineage_summary.get("isolation_violation_count")
|
|
623
|
+
),
|
|
624
|
+
"retention_violation_count": _as_int(
|
|
625
|
+
lineage_summary.get("retention_violation_count")
|
|
626
|
+
),
|
|
627
|
+
"blocking_gap_count": _as_int(lineage_summary.get("blocking_gap_count")),
|
|
628
|
+
"requires_external_service": bool(contract.get("requires_external_service")),
|
|
629
|
+
"local_executable_fixture": bool(contract.get("local_executable_fixture")),
|
|
630
|
+
}
|
|
631
|
+
|
|
632
|
+
|
|
633
|
+
def _memory_probe_findings(
|
|
634
|
+
retrieval: Mapping[str, Any],
|
|
635
|
+
lineage: Mapping[str, Any],
|
|
636
|
+
lineage_summary: Mapping[str, Any],
|
|
637
|
+
*,
|
|
638
|
+
contract: Mapping[str, Any],
|
|
639
|
+
) -> list[dict[str, Any]]:
|
|
640
|
+
findings: list[dict[str, Any]] = []
|
|
641
|
+
documents = [_plain_mapping(item) for item in _plain_list(retrieval.get("documents"))]
|
|
642
|
+
current_doc_ids = {
|
|
643
|
+
str(item.get("id") or "")
|
|
644
|
+
for item in documents
|
|
645
|
+
if item.get("current") is True and str(item.get("id") or "")
|
|
646
|
+
}
|
|
647
|
+
citations = [_plain_mapping(item) for item in _plain_list(retrieval.get("citations"))]
|
|
648
|
+
cited_doc_ids = {
|
|
649
|
+
str(doc_id)
|
|
650
|
+
for citation in citations
|
|
651
|
+
for doc_id in _plain_list(citation.get("doc_ids"))
|
|
652
|
+
if str(doc_id or "")
|
|
653
|
+
}
|
|
654
|
+
_append_finding(
|
|
655
|
+
findings,
|
|
656
|
+
"memory_probe_local_contract",
|
|
657
|
+
bool(contract.get("local_executable_fixture"))
|
|
658
|
+
and not bool(contract.get("requires_external_service")),
|
|
659
|
+
"memory probe target must be local and no-external-service",
|
|
660
|
+
{"contract": dict(contract)},
|
|
661
|
+
)
|
|
662
|
+
_append_finding(
|
|
663
|
+
findings,
|
|
664
|
+
"memory_probe_current_retrieval",
|
|
665
|
+
bool(current_doc_ids)
|
|
666
|
+
and bool(cited_doc_ids)
|
|
667
|
+
and cited_doc_ids.issubset(current_doc_ids)
|
|
668
|
+
and all(citation.get("freshness_checked") is True for citation in citations),
|
|
669
|
+
"retrieval citations must cite current documents with freshness checks",
|
|
670
|
+
{
|
|
671
|
+
"current_doc_ids": sorted(current_doc_ids),
|
|
672
|
+
"cited_doc_ids": sorted(cited_doc_ids),
|
|
673
|
+
"citation_count": len(citations),
|
|
674
|
+
},
|
|
675
|
+
)
|
|
676
|
+
operation_types = set(_plain_list(lineage_summary.get("operation_types")))
|
|
677
|
+
required_operations = set(_plain_list(contract.get("operations")) or ["read", "write", "recall"])
|
|
678
|
+
_append_finding(
|
|
679
|
+
findings,
|
|
680
|
+
"memory_probe_read_write_recall",
|
|
681
|
+
required_operations.issubset(operation_types)
|
|
682
|
+
and _as_int(lineage_summary.get("audited_operation_count"))
|
|
683
|
+
>= _as_int(lineage_summary.get("operation_count"))
|
|
684
|
+
and _as_int(lineage_summary.get("operation_count")) >= len(required_operations),
|
|
685
|
+
"memory lineage must include audited read/write/recall operations",
|
|
686
|
+
{
|
|
687
|
+
"required_operations": sorted(required_operations),
|
|
688
|
+
"operation_types": sorted(operation_types),
|
|
689
|
+
"operation_count": lineage_summary.get("operation_count"),
|
|
690
|
+
"audited_operation_count": lineage_summary.get("audited_operation_count"),
|
|
691
|
+
},
|
|
692
|
+
)
|
|
693
|
+
_append_finding(
|
|
694
|
+
findings,
|
|
695
|
+
"memory_probe_lineage_attribution",
|
|
696
|
+
_as_int(lineage_summary.get("memory_count")) > 0
|
|
697
|
+
and bool(lineage_summary.get("has_lineage"))
|
|
698
|
+
and bool(lineage_summary.get("has_source_attribution"))
|
|
699
|
+
and not _plain_list(lineage_summary.get("missing_required_evidence"))
|
|
700
|
+
and not _plain_list(lineage_summary.get("missing_required_signals")),
|
|
701
|
+
"memory records must have source attribution and closed lineage",
|
|
702
|
+
{
|
|
703
|
+
"memory_count": lineage_summary.get("memory_count"),
|
|
704
|
+
"has_lineage": lineage_summary.get("has_lineage"),
|
|
705
|
+
"missing_required_evidence": lineage_summary.get("missing_required_evidence"),
|
|
706
|
+
"missing_required_signals": lineage_summary.get("missing_required_signals"),
|
|
707
|
+
},
|
|
708
|
+
)
|
|
709
|
+
governance_keys = (
|
|
710
|
+
"has_tenant_isolation",
|
|
711
|
+
"has_audit",
|
|
712
|
+
"has_retention_policy",
|
|
713
|
+
"has_deletion_policy",
|
|
714
|
+
"has_redaction",
|
|
715
|
+
"has_canaries",
|
|
716
|
+
"has_observability",
|
|
717
|
+
"has_artifacts",
|
|
718
|
+
)
|
|
719
|
+
_append_finding(
|
|
720
|
+
findings,
|
|
721
|
+
"memory_probe_governance",
|
|
722
|
+
all(bool(lineage_summary.get(key)) for key in governance_keys)
|
|
723
|
+
and _as_int(lineage_summary.get("policy_violation_count")) == 0
|
|
724
|
+
and _as_int(lineage_summary.get("open_poisoning_count")) == 0
|
|
725
|
+
and _as_int(lineage_summary.get("isolation_violation_count")) == 0
|
|
726
|
+
and _as_int(lineage_summary.get("retention_violation_count")) == 0
|
|
727
|
+
and _as_int(lineage_summary.get("blocking_gap_count")) == 0,
|
|
728
|
+
"memory governance, poisoning, isolation, retention, and artifacts must close",
|
|
729
|
+
{
|
|
730
|
+
key: lineage_summary.get(key)
|
|
731
|
+
for key in (
|
|
732
|
+
*governance_keys,
|
|
733
|
+
"policy_violation_count",
|
|
734
|
+
"open_poisoning_count",
|
|
735
|
+
"isolation_violation_count",
|
|
736
|
+
"retention_violation_count",
|
|
737
|
+
"blocking_gap_count",
|
|
738
|
+
)
|
|
739
|
+
},
|
|
740
|
+
)
|
|
741
|
+
return findings
|
|
742
|
+
|
|
743
|
+
|
|
744
|
+
def _append_finding(
|
|
745
|
+
findings: list[dict[str, Any]],
|
|
746
|
+
check: str,
|
|
747
|
+
passed: bool,
|
|
748
|
+
message: str,
|
|
749
|
+
evidence: Mapping[str, Any],
|
|
750
|
+
) -> None:
|
|
751
|
+
if passed:
|
|
752
|
+
return
|
|
753
|
+
findings.append(
|
|
754
|
+
{
|
|
755
|
+
"check": check,
|
|
756
|
+
"level": "error",
|
|
757
|
+
"message": message,
|
|
758
|
+
"evidence": dict(evidence),
|
|
759
|
+
}
|
|
760
|
+
)
|
|
761
|
+
|
|
762
|
+
|
|
763
|
+
def _memory_probe_cases(cases: Sequence[Mapping[str, Any]] | None) -> list[dict[str, Any]]:
|
|
764
|
+
if cases:
|
|
765
|
+
return [dict(item) for item in cases]
|
|
766
|
+
return [
|
|
767
|
+
{
|
|
768
|
+
"id": "memory-probe",
|
|
769
|
+
"input": "Store and recall a grounded refund policy memory.",
|
|
770
|
+
"memory_key": "refund_policy_memory",
|
|
771
|
+
"memory_value": "Refund policy memory with source attribution.",
|
|
772
|
+
}
|
|
773
|
+
]
|
|
774
|
+
|
|
775
|
+
|
|
776
|
+
def _required_operations(cases: Sequence[Mapping[str, Any]]) -> list[str]:
|
|
777
|
+
operations: list[str] = []
|
|
778
|
+
for case in cases:
|
|
779
|
+
operations.extend(str(item) for item in _plain_list(case.get("required_operations")))
|
|
780
|
+
return _unique_strings(operations or ["read", "write", "recall"])
|
|
781
|
+
|
|
782
|
+
|
|
783
|
+
def _plain_mapping(value: Any) -> dict[str, Any]:
|
|
784
|
+
return dict(value) if isinstance(value, Mapping) else {}
|
|
785
|
+
|
|
786
|
+
|
|
787
|
+
def _plain_list(value: Any) -> list[Any]:
|
|
788
|
+
if value is None:
|
|
789
|
+
return []
|
|
790
|
+
if isinstance(value, list):
|
|
791
|
+
return value
|
|
792
|
+
if isinstance(value, tuple):
|
|
793
|
+
return list(value)
|
|
794
|
+
return [value]
|
|
795
|
+
|
|
796
|
+
|
|
797
|
+
def _scope_key(value: Any) -> str:
|
|
798
|
+
return str(value or "").strip().lower().replace("-", "_").replace(" ", "_")
|
|
799
|
+
|
|
800
|
+
|
|
801
|
+
def _operation_key(value: Any) -> str:
|
|
802
|
+
aliases = {
|
|
803
|
+
"memory_write": "write",
|
|
804
|
+
"write_memory": "write",
|
|
805
|
+
"memory_read": "read",
|
|
806
|
+
"retrieve_memory": "read",
|
|
807
|
+
"memory_retrieval": "read",
|
|
808
|
+
"memory_recall": "recall",
|
|
809
|
+
"recall_memory": "recall",
|
|
810
|
+
"memory_update": "update",
|
|
811
|
+
"memory_delete": "delete",
|
|
812
|
+
"delete_memory": "delete",
|
|
813
|
+
}
|
|
814
|
+
normalized = _scope_key(value)
|
|
815
|
+
return aliases.get(normalized, normalized)
|
|
816
|
+
|
|
817
|
+
|
|
818
|
+
def _has_policy(policy_keys: set[str], *names: str) -> bool:
|
|
819
|
+
return bool(policy_keys & {_scope_key(item) for item in names})
|
|
820
|
+
|
|
821
|
+
|
|
822
|
+
def _operation_audited(operation: Mapping[str, Any]) -> bool:
|
|
823
|
+
return bool(
|
|
824
|
+
operation.get("trace_id")
|
|
825
|
+
or operation.get("audit_id")
|
|
826
|
+
or operation.get("evidence")
|
|
827
|
+
)
|
|
828
|
+
|
|
829
|
+
|
|
830
|
+
def _all_operations_audited(operations: Sequence[Mapping[str, Any]]) -> bool:
|
|
831
|
+
return bool(operations) and all(_operation_audited(item) for item in operations)
|
|
832
|
+
|
|
833
|
+
|
|
834
|
+
def _observability_hook_count(observability: Mapping[str, Any]) -> int:
|
|
835
|
+
if not observability:
|
|
836
|
+
return 0
|
|
837
|
+
return sum(
|
|
838
|
+
len(_plain_list(observability.get(key)))
|
|
839
|
+
for key in ("traces", "logs", "hooks", "spans", "events")
|
|
840
|
+
) or 1
|
|
841
|
+
|
|
842
|
+
|
|
843
|
+
def _as_int(value: Any) -> int:
|
|
844
|
+
if isinstance(value, bool):
|
|
845
|
+
return int(value)
|
|
846
|
+
if isinstance(value, int):
|
|
847
|
+
return value
|
|
848
|
+
if isinstance(value, float):
|
|
849
|
+
return int(value)
|
|
850
|
+
if isinstance(value, str):
|
|
851
|
+
try:
|
|
852
|
+
return int(float(value.strip()))
|
|
853
|
+
except ValueError:
|
|
854
|
+
return 0
|
|
855
|
+
return 0
|
|
856
|
+
|
|
857
|
+
|
|
858
|
+
def _unique_strings(values: Sequence[Any]) -> list[str]:
|
|
859
|
+
seen: set[str] = set()
|
|
860
|
+
result: list[str] = []
|
|
861
|
+
for value in values:
|
|
862
|
+
text = str(value)
|
|
863
|
+
if text and text not in seen:
|
|
864
|
+
seen.add(text)
|
|
865
|
+
result.append(text)
|
|
866
|
+
return result
|
|
867
|
+
|
|
868
|
+
|
|
869
|
+
def _dedupe_documents(documents: Sequence[Mapping[str, Any]]) -> list[dict[str, Any]]:
|
|
870
|
+
seen: set[str] = set()
|
|
871
|
+
result: list[dict[str, Any]] = []
|
|
872
|
+
for document in documents:
|
|
873
|
+
item = dict(document)
|
|
874
|
+
doc_id = str(item.get("id") or "")
|
|
875
|
+
if doc_id and doc_id in seen:
|
|
876
|
+
continue
|
|
877
|
+
if doc_id:
|
|
878
|
+
seen.add(doc_id)
|
|
879
|
+
result.append(item)
|
|
880
|
+
return result
|
|
881
|
+
|
|
882
|
+
|
|
883
|
+
def _lineage_source_ids(lineage: Mapping[str, Any]) -> list[str]:
|
|
884
|
+
source_ids: list[str] = []
|
|
885
|
+
for memory in _plain_list(lineage.get("memories")):
|
|
886
|
+
item = _plain_mapping(memory)
|
|
887
|
+
source_ids.extend(
|
|
888
|
+
str(source_id)
|
|
889
|
+
for source_id in _plain_list(
|
|
890
|
+
item.get("source_ids") or item.get("sources") or item.get("doc_ids")
|
|
891
|
+
)
|
|
892
|
+
if str(source_id)
|
|
893
|
+
)
|
|
894
|
+
return _unique_strings(source_ids)
|
|
895
|
+
|
|
896
|
+
|
|
897
|
+
def _is_external_target(target: str) -> bool:
|
|
898
|
+
return urlparse(str(target)).scheme.lower() in {"http", "https"}
|
|
899
|
+
|
|
900
|
+
|
|
901
|
+
__all__ = [
|
|
902
|
+
"memory_layer_contract",
|
|
903
|
+
"probe_memory_layer",
|
|
904
|
+
"run_memory_layer_probe",
|
|
905
|
+
]
|