agent-os-kernel 1.1.0__py3-none-any.whl → 1.3.0__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_os/__init__.py +66 -4
- agent_os/agents_compat.py +286 -0
- agent_os/base_agent.py +308 -0
- agent_os/cli.py +1079 -19
- agent_os/integrations/__init__.py +37 -2
- agent_os/integrations/openai_adapter.py +502 -0
- agent_os/integrations/semantic_kernel_adapter.py +569 -0
- agent_os/stateless.py +349 -0
- agent_os_kernel-1.3.0.dist-info/METADATA +676 -0
- agent_os_kernel-1.3.0.dist-info/RECORD +1053 -0
- {agent_os_kernel-1.1.0.dist-info → agent_os_kernel-1.3.0.dist-info}/entry_points.txt +0 -1
- modules/amb/.github/workflows/ci.yml +102 -0
- modules/amb/.github/workflows/publish.yml +146 -0
- modules/amb/.gitignore +134 -0
- modules/amb/CHANGELOG.md +118 -0
- modules/amb/CONTRIBUTING.md +141 -0
- modules/amb/LICENSE +21 -0
- modules/amb/README.md +188 -0
- modules/amb/amb_core/__init__.py +175 -0
- modules/amb/amb_core/adapters/__init__.py +55 -0
- modules/amb/amb_core/adapters/aws_sqs_broker.py +374 -0
- modules/amb/amb_core/adapters/azure_servicebus_broker.py +338 -0
- modules/amb/amb_core/adapters/kafka_broker.py +258 -0
- modules/amb/amb_core/adapters/nats_broker.py +283 -0
- modules/amb/amb_core/adapters/rabbitmq_broker.py +233 -0
- modules/amb/amb_core/adapters/redis_broker.py +260 -0
- modules/amb/amb_core/broker.py +143 -0
- modules/amb/amb_core/bus.py +479 -0
- modules/amb/amb_core/cloudevents.py +507 -0
- modules/amb/amb_core/dlq.py +343 -0
- modules/amb/amb_core/hf_utils.py +534 -0
- modules/amb/amb_core/memory_broker.py +408 -0
- modules/amb/amb_core/models.py +139 -0
- modules/amb/amb_core/persistence.py +527 -0
- modules/amb/amb_core/schema.py +292 -0
- modules/amb/amb_core/tracing.py +356 -0
- modules/amb/examples/advanced_features.py +223 -0
- modules/amb/examples/backpressure_demo.py +225 -0
- modules/amb/examples/basic_usage.py +117 -0
- modules/amb/examples/tracing_demo.py +104 -0
- modules/amb/experiments/README.md +52 -0
- modules/amb/experiments/reproduce_results.py +467 -0
- modules/amb/experiments/results.json +324 -0
- modules/amb/paper/README.md +40 -0
- modules/amb/paper/paper.tex +365 -0
- modules/amb/paper/whitepaper.md +377 -0
- modules/amb/pyproject.toml +117 -0
- modules/amb/tests/__init__.py +1 -0
- modules/amb/tests/test_backpressure_priority.py +280 -0
- modules/amb/tests/test_bus.py +198 -0
- modules/amb/tests/test_cloudevents.py +443 -0
- modules/amb/tests/test_features.py +531 -0
- modules/amb/tests/test_models.py +74 -0
- modules/amb/tests/test_tracing.py +254 -0
- modules/atr/.github/workflows/ci.yml +101 -0
- modules/atr/.github/workflows/publish.yml +140 -0
- modules/atr/.gitignore +134 -0
- modules/atr/.pre-commit-config.yaml +37 -0
- modules/atr/CHANGELOG.md +39 -0
- modules/atr/CONTRIBUTING.md +96 -0
- modules/atr/IMPLEMENTATION_SUMMARY.md +143 -0
- modules/atr/README.md +180 -0
- modules/atr/atr/__init__.py +638 -0
- modules/atr/atr/access.py +346 -0
- modules/atr/atr/composition.py +643 -0
- modules/atr/atr/decorator.py +355 -0
- modules/atr/atr/executor.py +382 -0
- modules/atr/atr/health.py +555 -0
- modules/atr/atr/hf_utils.py +447 -0
- modules/atr/atr/injection.py +420 -0
- modules/atr/atr/metrics.py +438 -0
- modules/atr/atr/policies.py +401 -0
- modules/atr/atr/py.typed +2 -0
- modules/atr/atr/registry.py +450 -0
- modules/atr/atr/schema.py +478 -0
- modules/atr/atr/tools/safe/__init__.py +73 -0
- modules/atr/atr/tools/safe/calculator.py +380 -0
- modules/atr/atr/tools/safe/datetime_tool.py +441 -0
- modules/atr/atr/tools/safe/file_reader.py +400 -0
- modules/atr/atr/tools/safe/http_client.py +314 -0
- modules/atr/atr/tools/safe/json_parser.py +372 -0
- modules/atr/atr/tools/safe/text_tool.py +526 -0
- modules/atr/atr/tools/safe/toolkit.py +173 -0
- modules/atr/docs/PYPI_SETUP.md +113 -0
- modules/atr/examples/README.md +27 -0
- modules/atr/examples/demo.py +144 -0
- modules/atr/examples/sandbox_demo.py +218 -0
- modules/atr/experiments/README.md +69 -0
- modules/atr/experiments/reproduce_results.py +509 -0
- modules/atr/experiments/results/.gitkeep +0 -0
- modules/atr/experiments/results/results_20260123_140334.json +71 -0
- modules/atr/paper/README.md +36 -0
- modules/atr/paper/figures/.gitkeep +0 -0
- modules/atr/paper/references.bib +84 -0
- modules/atr/paper/structure.tex +293 -0
- modules/atr/paper/whitepaper.md +234 -0
- modules/atr/pyproject.toml +148 -0
- modules/atr/requirements.txt +1 -0
- modules/atr/setup.py +30 -0
- modules/atr/tests/__init__.py +1 -0
- modules/atr/tests/test_decorator.py +317 -0
- modules/atr/tests/test_executor.py +245 -0
- modules/atr/tests/test_integration_executor.py +184 -0
- modules/atr/tests/test_registry.py +312 -0
- modules/atr/tests/test_schema.py +182 -0
- modules/atr/tests/test_v2_features.py +708 -0
- modules/caas/.dockerignore +63 -0
- modules/caas/.github/ISSUE_TEMPLATE/bug_report.md +38 -0
- modules/caas/.github/ISSUE_TEMPLATE/custom.md +10 -0
- modules/caas/.github/ISSUE_TEMPLATE/feature_request.md +20 -0
- modules/caas/.github/workflows/ci.yml +100 -0
- modules/caas/.github/workflows/lint.yml +39 -0
- modules/caas/.github/workflows/publish-pypi.yml +124 -0
- modules/caas/.gitignore +73 -0
- modules/caas/.pre-commit-config.yaml +33 -0
- modules/caas/CHANGELOG.md +58 -0
- modules/caas/CONTRIBUTING.md +346 -0
- modules/caas/Dockerfile +41 -0
- modules/caas/LICENSE +21 -0
- modules/caas/MANIFEST.in +11 -0
- modules/caas/README.md +158 -0
- modules/caas/benchmarks/README.md +255 -0
- modules/caas/benchmarks/create_hf_dataset.py +502 -0
- modules/caas/benchmarks/data/sample_corpus/README.md +86 -0
- modules/caas/benchmarks/data/sample_corpus/auth_module.py +211 -0
- modules/caas/benchmarks/data/sample_corpus/contribution_guide.md +185 -0
- modules/caas/benchmarks/data/sample_corpus/remote_work_policy.html +57 -0
- modules/caas/benchmarks/hf_dataset/README.md +214 -0
- modules/caas/benchmarks/hf_dataset/caas_benchmark_corpus.py +73 -0
- modules/caas/benchmarks/hf_dataset/corpus_preview.json +193 -0
- modules/caas/benchmarks/results/README.md +66 -0
- modules/caas/benchmarks/results/evaluation_2026-01-20.json +121 -0
- modules/caas/benchmarks/run_evaluation.py +561 -0
- modules/caas/benchmarks/statistical_tests.py +289 -0
- modules/caas/benchmarks/verify_sample_corpus.py +83 -0
- modules/caas/docker-compose.yml +38 -0
- modules/caas/docs/CONTEXT_TRIAD.md +462 -0
- modules/caas/docs/CONTRIBUTING.md +346 -0
- modules/caas/docs/ETHICS_AND_LIMITATIONS.md +336 -0
- modules/caas/docs/HEURISTIC_ROUTER.md +442 -0
- modules/caas/docs/IMPLEMENTATION_SUMMARY.md +363 -0
- modules/caas/docs/IMPLEMENTATION_SUMMARY_CONTEXT_TRIAD.md +277 -0
- modules/caas/docs/IMPLEMENTATION_SUMMARY_HEURISTIC_ROUTER.md +231 -0
- modules/caas/docs/IMPLEMENTATION_SUMMARY_METADATA_INJECTION.md +258 -0
- modules/caas/docs/IMPLEMENTATION_SUMMARY_PRAGMATIC_TRUTH.md +212 -0
- modules/caas/docs/IMPLEMENTATION_SUMMARY_TRUST_GATEWAY.md +319 -0
- modules/caas/docs/LAYER_1_PRIMITIVE.md +202 -0
- modules/caas/docs/METADATA_INJECTION.md +404 -0
- modules/caas/docs/PRAGMATIC_TRUTH.md +431 -0
- modules/caas/docs/RELATED_WORK.md +312 -0
- modules/caas/docs/RELEASE_CHECKLIST.md +219 -0
- modules/caas/docs/RELEASE_GUIDE.md +285 -0
- modules/caas/docs/REPRODUCIBILITY.md +386 -0
- modules/caas/docs/SLIDING_WINDOW.md +387 -0
- modules/caas/docs/STRUCTURE_AWARE_INDEXING.md +158 -0
- modules/caas/docs/TESTING.md +259 -0
- modules/caas/docs/THREAT_MODEL.md +247 -0
- modules/caas/docs/TRUST_GATEWAY.md +575 -0
- modules/caas/docs/VFS.md +298 -0
- modules/caas/examples/agents/enterprise_security_agent.py +414 -0
- modules/caas/examples/agents/intelligent_document_analyzer.py +380 -0
- modules/caas/examples/demos/demo.py +309 -0
- modules/caas/examples/demos/demo_context_triad.py +225 -0
- modules/caas/examples/demos/demo_conversation_manager.py +285 -0
- modules/caas/examples/demos/demo_heuristic_router.py +133 -0
- modules/caas/examples/demos/demo_metadata_injection.py +198 -0
- modules/caas/examples/demos/demo_pragmatic_truth.py +303 -0
- modules/caas/examples/demos/demo_structure_aware.py +140 -0
- modules/caas/examples/demos/demo_time_decay.py +247 -0
- modules/caas/examples/demos/demo_trust_gateway.py +383 -0
- modules/caas/examples/multi_agent/README.md +159 -0
- modules/caas/examples/multi_agent/research_team.py +369 -0
- modules/caas/examples/multi_agent/vfs_collaboration.py +393 -0
- modules/caas/examples/usage/auth_module.py +142 -0
- modules/caas/examples/usage/usage_example.py +173 -0
- modules/caas/experiments/README.md +42 -0
- modules/caas/experiments/reproduce_results.py +462 -0
- modules/caas/paper/ARXIV_METADATA.md +145 -0
- modules/caas/paper/ARXIV_README.md +47 -0
- modules/caas/paper/CHECKLIST.md +103 -0
- modules/caas/paper/GITHUB_RELEASE_NOTES.md +105 -0
- modules/caas/paper/README.md +71 -0
- modules/caas/paper/abstract.md +24 -0
- modules/caas/paper/arxiv_submission.tar +0 -0
- modules/caas/paper/arxiv_submission.zip +0 -0
- modules/caas/paper/build_pdf.py +355 -0
- modules/caas/paper/experiments.md +149 -0
- modules/caas/paper/figures/.gitkeep +0 -0
- modules/caas/paper/figures/README.md +237 -0
- modules/caas/paper/figures/fig1_system_architecture.png +0 -0
- modules/caas/paper/figures/fig1_system_architecture.svg +198 -0
- modules/caas/paper/figures/fig2_context_triad.png +0 -0
- modules/caas/paper/figures/fig2_context_triad.svg +105 -0
- modules/caas/paper/figures/fig3_ablation_results.png +0 -0
- modules/caas/paper/figures/fig3_ablation_results.svg +113 -0
- modules/caas/paper/figures/fig4_routing_latency.png +0 -0
- modules/caas/paper/figures/fig4_routing_latency.svg +97 -0
- modules/caas/paper/intro.md +103 -0
- modules/caas/paper/latex/figures/fig1_system_architecture.png +0 -0
- modules/caas/paper/latex/figures/fig2_context_triad.png +0 -0
- modules/caas/paper/latex/figures/fig3_ablation_results.png +0 -0
- modules/caas/paper/latex/figures/fig4_routing_latency.png +0 -0
- modules/caas/paper/latex/main.tex +468 -0
- modules/caas/paper/latex/references.bib +140 -0
- modules/caas/paper/method.md +350 -0
- modules/caas/paper/outline.md +123 -0
- modules/caas/paper/related_work.md +101 -0
- modules/caas/paper/tables/.gitkeep +0 -0
- modules/caas/paper/tables/results_tables.md +50 -0
- modules/caas/pyproject.toml +172 -0
- modules/caas/requirements.txt +11 -0
- modules/caas/src/caas/__init__.py +232 -0
- modules/caas/src/caas/api/__init__.py +7 -0
- modules/caas/src/caas/api/server.py +1326 -0
- modules/caas/src/caas/caching.py +832 -0
- modules/caas/src/caas/cli.py +208 -0
- modules/caas/src/caas/conversation.py +221 -0
- modules/caas/src/caas/decay.py +118 -0
- modules/caas/src/caas/detection/__init__.py +7 -0
- modules/caas/src/caas/detection/detector.py +236 -0
- modules/caas/src/caas/enrichment.py +127 -0
- modules/caas/src/caas/gateway/__init__.py +24 -0
- modules/caas/src/caas/gateway/trust_gateway.py +471 -0
- modules/caas/src/caas/hf_utils.py +477 -0
- modules/caas/src/caas/ingestion/__init__.py +21 -0
- modules/caas/src/caas/ingestion/processors.py +251 -0
- modules/caas/src/caas/ingestion/structure_parser.py +185 -0
- modules/caas/src/caas/models.py +354 -0
- modules/caas/src/caas/pragmatic_truth.py +441 -0
- modules/caas/src/caas/routing/__init__.py +8 -0
- modules/caas/src/caas/routing/heuristic_router.py +242 -0
- modules/caas/src/caas/storage/__init__.py +7 -0
- modules/caas/src/caas/storage/store.py +450 -0
- modules/caas/src/caas/triad.py +472 -0
- modules/caas/src/caas/tuning/__init__.py +7 -0
- modules/caas/src/caas/tuning/tuner.py +322 -0
- modules/caas/src/caas/vfs/__init__.py +12 -0
- modules/caas/src/caas/vfs/filesystem.py +450 -0
- modules/caas/tests/__init__.py +3 -0
- modules/caas/tests/conftest.py +8 -0
- modules/caas/tests/test_caching.py +628 -0
- modules/caas/tests/test_context_triad.py +385 -0
- modules/caas/tests/test_conversation_manager.py +289 -0
- modules/caas/tests/test_functionality.py +215 -0
- modules/caas/tests/test_heuristic_router.py +370 -0
- modules/caas/tests/test_metadata_injection.py +328 -0
- modules/caas/tests/test_pragmatic_truth.py +322 -0
- modules/caas/tests/test_structure_aware_indexing.py +283 -0
- modules/caas/tests/test_time_decay.py +268 -0
- modules/caas/tests/test_trust_gateway.py +445 -0
- modules/caas/tests/test_vfs.py +298 -0
- modules/cmvk/.github/FUNDING.yml +9 -0
- modules/cmvk/.github/dependabot.yml +54 -0
- modules/cmvk/.github/workflows/ci.yml +205 -0
- modules/cmvk/.github/workflows/publish.yml +143 -0
- modules/cmvk/.gitignore +147 -0
- modules/cmvk/.pre-commit-config.yaml +58 -0
- modules/cmvk/CHANGELOG.md +146 -0
- modules/cmvk/CITATION.cff +48 -0
- modules/cmvk/CONTRIBUTING.md +229 -0
- modules/cmvk/Dockerfile +87 -0
- modules/cmvk/HF_MODEL_CARD.md +185 -0
- modules/cmvk/LICENSE +21 -0
- modules/cmvk/README.md +149 -0
- modules/cmvk/SECURITY.md +114 -0
- modules/cmvk/config/prompts/generator_v1.txt +23 -0
- modules/cmvk/config/prompts/verifier_hostile.txt +32 -0
- modules/cmvk/config/settings.yaml +40 -0
- modules/cmvk/coverage_html/.gitignore +2 -0
- modules/cmvk/coverage_html/class_index.html +658 -0
- modules/cmvk/coverage_html/coverage_html_cb_188fc9a4.js +735 -0
- modules/cmvk/coverage_html/favicon_32_cb_c827f16f.png +0 -0
- modules/cmvk/coverage_html/function_index.html +1978 -0
- modules/cmvk/coverage_html/index.html +255 -0
- modules/cmvk/coverage_html/keybd_closed_cb_900cfef5.png +0 -0
- modules/cmvk/coverage_html/status.json +1 -0
- modules/cmvk/coverage_html/style_cb_5c747636.css +389 -0
- modules/cmvk/coverage_html/z_2c49bd2ed3e01e38___init___py.html +315 -0
- modules/cmvk/coverage_html/z_2c49bd2ed3e01e38_audit_py.html +499 -0
- modules/cmvk/coverage_html/z_2c49bd2ed3e01e38_benchmarks_py.html +575 -0
- modules/cmvk/coverage_html/z_2c49bd2ed3e01e38_constitutional_py.html +1001 -0
- modules/cmvk/coverage_html/z_2c49bd2ed3e01e38_hf_utils_py.html +398 -0
- modules/cmvk/coverage_html/z_2c49bd2ed3e01e38_metrics_py.html +570 -0
- modules/cmvk/coverage_html/z_2c49bd2ed3e01e38_profiles_py.html +397 -0
- modules/cmvk/coverage_html/z_2c49bd2ed3e01e38_types_py.html +109 -0
- modules/cmvk/coverage_html/z_2c49bd2ed3e01e38_verification_py.html +1053 -0
- modules/cmvk/docs/DIAGRAMS.md +325 -0
- modules/cmvk/docs/architecture.md +345 -0
- modules/cmvk/docs/features.md +308 -0
- modules/cmvk/docs/getting_started.md +279 -0
- modules/cmvk/docs/innovation_layer.md +377 -0
- modules/cmvk/docs/safety.md +281 -0
- modules/cmvk/docs/traceability.md +150 -0
- modules/cmvk/examples/basic_example.py +62 -0
- modules/cmvk/examples/demo_complete_pipeline.py +209 -0
- modules/cmvk/examples/demo_innovation_layer.py +197 -0
- modules/cmvk/examples/example.py +112 -0
- modules/cmvk/examples/model_diversity_comparison.py +110 -0
- modules/cmvk/examples/real_api_integration.py +121 -0
- modules/cmvk/examples/test_full_pipeline.py +303 -0
- modules/cmvk/experiments/FEATURE_2_LATERAL_THINKING.md +187 -0
- modules/cmvk/experiments/README.md +216 -0
- modules/cmvk/experiments/ablation_runner.py +666 -0
- modules/cmvk/experiments/baseline_runner.py +158 -0
- modules/cmvk/experiments/blind_spot_benchmark.py +364 -0
- modules/cmvk/experiments/datasets/README.md +85 -0
- modules/cmvk/experiments/datasets/humaneval_50.json +352 -0
- modules/cmvk/experiments/datasets/humaneval_full.json +1150 -0
- modules/cmvk/experiments/datasets/humaneval_sample.json +32 -0
- modules/cmvk/experiments/datasets/sabotage.json +262 -0
- modules/cmvk/experiments/datasets/sample.json +40 -0
- modules/cmvk/experiments/demo_with_traces.py +110 -0
- modules/cmvk/experiments/efficiency_curve.py +259 -0
- modules/cmvk/experiments/experiment_runner.py +243 -0
- modules/cmvk/experiments/paper_data_generator.py +183 -0
- modules/cmvk/experiments/reproduce_results.py +407 -0
- modules/cmvk/experiments/reproducible_runner.py +352 -0
- modules/cmvk/experiments/sabotage_stress_test.py +311 -0
- modules/cmvk/experiments/test_lateral_thinking.py +116 -0
- modules/cmvk/experiments/test_prosecutor.py +41 -0
- modules/cmvk/experiments/visualize_results.py +735 -0
- modules/cmvk/logs/traces/demo_HumanEval_0_20260121-204900.json +36 -0
- modules/cmvk/notebooks/analysis.ipynb +124 -0
- modules/cmvk/paper/PAPER.md +561 -0
- modules/cmvk/paper/arxiv_checklist.md +230 -0
- modules/cmvk/paper/cmvk_neurips.aux +77 -0
- modules/cmvk/paper/cmvk_neurips.bbl +81 -0
- modules/cmvk/paper/cmvk_neurips.blg +48 -0
- modules/cmvk/paper/cmvk_neurips.out +16 -0
- modules/cmvk/paper/cmvk_neurips.pdf +0 -0
- modules/cmvk/paper/cmvk_neurips.tex +309 -0
- modules/cmvk/paper/figures/ablation.png +0 -0
- modules/cmvk/paper/figures/ablation.svg +39 -0
- modules/cmvk/paper/figures/architecture.png +0 -0
- modules/cmvk/paper/figures/architecture.svg +115 -0
- modules/cmvk/paper/figures/results_bar.png +0 -0
- modules/cmvk/paper/figures/results_bar.svg +70 -0
- modules/cmvk/paper/generate_figures.py +383 -0
- modules/cmvk/paper/neurips_2024.sty +101 -0
- modules/cmvk/paper/references.bib +98 -0
- modules/cmvk/paper/structure.tex +200 -0
- modules/cmvk/pyproject.toml +189 -0
- modules/cmvk/requirements-dev.txt +19 -0
- modules/cmvk/requirements.txt +14 -0
- modules/cmvk/src/cmvk/__init__.py +216 -0
- modules/cmvk/src/cmvk/audit.py +400 -0
- modules/cmvk/src/cmvk/benchmarks.py +476 -0
- modules/cmvk/src/cmvk/constitutional.py +902 -0
- modules/cmvk/src/cmvk/hf_utils.py +299 -0
- modules/cmvk/src/cmvk/metrics.py +471 -0
- modules/cmvk/src/cmvk/profiles.py +298 -0
- modules/cmvk/src/cmvk/py.typed +0 -0
- modules/cmvk/src/cmvk/types.py +10 -0
- modules/cmvk/src/cmvk/verification.py +954 -0
- modules/cmvk/src/cross_model_verification_kernel/__init__.py +91 -0
- modules/cmvk/src/cross_model_verification_kernel/__main__.py +10 -0
- modules/cmvk/src/cross_model_verification_kernel/agents/__init__.py +16 -0
- modules/cmvk/src/cross_model_verification_kernel/agents/base_agent.py +142 -0
- modules/cmvk/src/cross_model_verification_kernel/agents/generator_openai.py +223 -0
- modules/cmvk/src/cross_model_verification_kernel/agents/verifier_anthropic.py +448 -0
- modules/cmvk/src/cross_model_verification_kernel/agents/verifier_gemini.py +481 -0
- modules/cmvk/src/cross_model_verification_kernel/cli.py +570 -0
- modules/cmvk/src/cross_model_verification_kernel/core/__init__.py +26 -0
- modules/cmvk/src/cross_model_verification_kernel/core/graph_memory.py +308 -0
- modules/cmvk/src/cross_model_verification_kernel/core/kernel.py +413 -0
- modules/cmvk/src/cross_model_verification_kernel/core/trace_logger.py +75 -0
- modules/cmvk/src/cross_model_verification_kernel/core/types.py +121 -0
- modules/cmvk/src/cross_model_verification_kernel/datasets/__init__.py +20 -0
- modules/cmvk/src/cross_model_verification_kernel/datasets/humaneval_loader.py +271 -0
- modules/cmvk/src/cross_model_verification_kernel/generator.py +118 -0
- modules/cmvk/src/cross_model_verification_kernel/kernel.py +292 -0
- modules/cmvk/src/cross_model_verification_kernel/models.py +111 -0
- modules/cmvk/src/cross_model_verification_kernel/py.typed +1 -0
- modules/cmvk/src/cross_model_verification_kernel/simple_kernel.py +185 -0
- modules/cmvk/src/cross_model_verification_kernel/tools/__init__.py +94 -0
- modules/cmvk/src/cross_model_verification_kernel/tools/huggingface_upload.py +394 -0
- modules/cmvk/src/cross_model_verification_kernel/tools/sandbox.py +159 -0
- modules/cmvk/src/cross_model_verification_kernel/tools/statistics.py +468 -0
- modules/cmvk/src/cross_model_verification_kernel/tools/visualizer.py +312 -0
- modules/cmvk/src/cross_model_verification_kernel/tools/web_search.py +86 -0
- modules/cmvk/src/cross_model_verification_kernel/verifier.py +257 -0
- modules/cmvk/tests/__init__.py +3 -0
- modules/cmvk/tests/conftest.py +61 -0
- modules/cmvk/tests/integration/__init__.py +1 -0
- modules/cmvk/tests/integration/test_anthropic_verifier.py +269 -0
- modules/cmvk/tests/integration/test_integration.py +53 -0
- modules/cmvk/tests/integration/test_lateral_thinking_integration.py +199 -0
- modules/cmvk/tests/integration/test_lateral_thinking_witness.py +208 -0
- modules/cmvk/tests/integration/test_prosecutor_mode.py +131 -0
- modules/cmvk/tests/test_constitutional.py +611 -0
- modules/cmvk/tests/test_enhanced_features.py +603 -0
- modules/cmvk/tests/test_verification.py +255 -0
- modules/cmvk/tests/unit/__init__.py +1 -0
- modules/cmvk/tests/unit/test_agents.py +64 -0
- modules/cmvk/tests/unit/test_cli.py +224 -0
- modules/cmvk/tests/unit/test_core.py +126 -0
- modules/cmvk/tests/unit/test_humaneval_loader.py +197 -0
- modules/cmvk/tests/unit/test_kernel.py +255 -0
- modules/cmvk/tests/unit/test_reproducibility.py +160 -0
- modules/cmvk/tests/unit/test_trace_logger.py +115 -0
- modules/cmvk/tests/unit/test_visualizer.py +218 -0
- modules/control-plane/.github/ISSUE_TEMPLATE/bug_report.yml +82 -0
- modules/control-plane/.github/ISSUE_TEMPLATE/config.yml +11 -0
- modules/control-plane/.github/ISSUE_TEMPLATE/feature_request.yml +104 -0
- modules/control-plane/.github/ISSUE_TEMPLATE/question.yml +70 -0
- modules/control-plane/.github/ISSUE_TEMPLATE/security_vulnerability.yml +84 -0
- modules/control-plane/.github/discussions.yml +73 -0
- modules/control-plane/.github/pull_request_template.md +82 -0
- modules/control-plane/.github/workflows/publish.yml +146 -0
- modules/control-plane/.github/workflows/release.yml +39 -0
- modules/control-plane/.github/workflows/tests.yml +58 -0
- modules/control-plane/.gitignore +55 -0
- modules/control-plane/CHANGELOG.md +203 -0
- modules/control-plane/CONTRIBUTING.md +311 -0
- modules/control-plane/CONTRIBUTORS.md +88 -0
- modules/control-plane/Dockerfile +82 -0
- modules/control-plane/LICENSE +21 -0
- modules/control-plane/MANIFEST.in +17 -0
- modules/control-plane/README.md +1264 -0
- modules/control-plane/ROADMAP.md +228 -0
- modules/control-plane/SECURITY.md +210 -0
- modules/control-plane/SUPPORT.md +106 -0
- modules/control-plane/acp-cli.py +212 -0
- modules/control-plane/benchmark/README.md +257 -0
- modules/control-plane/benchmark/__init__.py +19 -0
- modules/control-plane/benchmark/red_team_dataset.py +517 -0
- modules/control-plane/benchmark.py +563 -0
- modules/control-plane/build_and_publish.sh +130 -0
- modules/control-plane/docker-compose.yml +74 -0
- modules/control-plane/docs/ABLATION_STUDIES.md +528 -0
- modules/control-plane/docs/ADAPTER_GUIDE.md +544 -0
- modules/control-plane/docs/ADVANCED_FEATURES.md +543 -0
- modules/control-plane/docs/AIOS_COMPARISON.md +296 -0
- modules/control-plane/docs/BIBLIOGRAPHY.md +367 -0
- modules/control-plane/docs/CASE_STUDIES.md +645 -0
- modules/control-plane/docs/DOCKER_DEPLOYMENT.md +184 -0
- modules/control-plane/docs/ECOSYSTEM_STATUS.md +98 -0
- modules/control-plane/docs/HF_MODEL_CARD.md +168 -0
- modules/control-plane/docs/KERNEL_V1_RELEASE.md +454 -0
- modules/control-plane/docs/LAYER3_FRAMEWORK.md +227 -0
- modules/control-plane/docs/LIMITATIONS.md +523 -0
- modules/control-plane/docs/PYPI_PUBLISHING.md +195 -0
- modules/control-plane/docs/README.md +58 -0
- modules/control-plane/docs/RELATED_WORK.md +319 -0
- modules/control-plane/docs/RELEASE_v1.1.0.md +252 -0
- modules/control-plane/docs/REPRODUCIBILITY.md +540 -0
- modules/control-plane/docs/RESEARCH_FOUNDATION.md +197 -0
- modules/control-plane/docs/api/CORE.md +270 -0
- modules/control-plane/docs/architecture/architecture.md +120 -0
- modules/control-plane/docs/community/ANNOUNCEMENT_TEMPLATES.md +52 -0
- modules/control-plane/docs/guides/IMPLEMENTATION.md +225 -0
- modules/control-plane/docs/guides/PHILOSOPHY.md +354 -0
- modules/control-plane/docs/guides/QUICKSTART.md +217 -0
- modules/control-plane/examples/README.md +138 -0
- modules/control-plane/examples/a2a_demo.py +410 -0
- modules/control-plane/examples/adapter_demo.py +347 -0
- modules/control-plane/examples/advanced_features.py +403 -0
- modules/control-plane/examples/basic_usage.py +261 -0
- modules/control-plane/examples/benchmark_demo.py +186 -0
- modules/control-plane/examples/compliance_demo.py +333 -0
- modules/control-plane/examples/configuration.py +265 -0
- modules/control-plane/examples/getting_started.py +178 -0
- modules/control-plane/examples/hibernation_and_time_travel_demo.py +406 -0
- modules/control-plane/examples/interactive_tutorial.ipynb +497 -0
- modules/control-plane/examples/kernel_interceptor_demo.py +202 -0
- modules/control-plane/examples/kernel_v1_demo.py +273 -0
- modules/control-plane/examples/langchain_demo.py +281 -0
- modules/control-plane/examples/lifecycle_demo.py +724 -0
- modules/control-plane/examples/mcp_demo.py +378 -0
- modules/control-plane/examples/ml_safety_demo.py +157 -0
- modules/control-plane/examples/multimodal_demo.py +347 -0
- modules/control-plane/examples/observability_demo.py +370 -0
- modules/control-plane/examples/use_cases.py +336 -0
- modules/control-plane/experiments/long_horizon_purge.py +235 -0
- modules/control-plane/experiments/multi_agent_rag.py +165 -0
- modules/control-plane/experiments/reproduce_results.py +667 -0
- modules/control-plane/paper/ARXIV_SUBMISSION_INFO.txt +122 -0
- modules/control-plane/paper/ETHICS_STATEMENT.md +248 -0
- modules/control-plane/paper/PAPER_CHECKLIST.md +72 -0
- modules/control-plane/paper/Paper.pdf +0 -0
- modules/control-plane/paper/README.md +71 -0
- modules/control-plane/paper/appendix.md +152 -0
- modules/control-plane/paper/architecture.md +15 -0
- modules/control-plane/paper/arxiv/figures/ablation_chart.png +0 -0
- modules/control-plane/paper/arxiv/figures/architecture.png +0 -0
- modules/control-plane/paper/arxiv/figures/constraint_graphs.png +0 -0
- modules/control-plane/paper/arxiv/figures/results_chart.png +0 -0
- modules/control-plane/paper/arxiv/main.aux +97 -0
- modules/control-plane/paper/arxiv/main.bbl +112 -0
- modules/control-plane/paper/arxiv/main.blg +48 -0
- modules/control-plane/paper/arxiv/main.out +33 -0
- modules/control-plane/paper/arxiv/main.pdf +0 -0
- modules/control-plane/paper/arxiv/main.tex +479 -0
- modules/control-plane/paper/arxiv/references.bib +234 -0
- modules/control-plane/paper/arxiv_submission.tar +0 -0
- modules/control-plane/paper/arxiv_submission.zip +0 -0
- modules/control-plane/paper/build.sh +68 -0
- modules/control-plane/paper/figures/README.md +47 -0
- modules/control-plane/paper/figures/ablation_chart.pdf +0 -0
- modules/control-plane/paper/figures/ablation_chart.png +0 -0
- modules/control-plane/paper/figures/architecture.pdf +0 -0
- modules/control-plane/paper/figures/architecture.png +0 -0
- modules/control-plane/paper/figures/constraint_graphs.pdf +0 -0
- modules/control-plane/paper/figures/constraint_graphs.png +0 -0
- modules/control-plane/paper/figures/generate_figures.py +252 -0
- modules/control-plane/paper/figures/results_chart.pdf +0 -0
- modules/control-plane/paper/figures/results_chart.png +0 -0
- modules/control-plane/paper/main.md +273 -0
- modules/control-plane/paper/main.tex +214 -0
- modules/control-plane/paper/main_arxiv.aux +53 -0
- modules/control-plane/paper/main_arxiv.out +17 -0
- modules/control-plane/paper/main_arxiv.pdf +0 -0
- modules/control-plane/paper/main_arxiv.tex +264 -0
- modules/control-plane/paper/references.bib +234 -0
- modules/control-plane/pyproject.toml +124 -0
- modules/control-plane/reproducibility/ABLATIONS.md +136 -0
- modules/control-plane/reproducibility/README.md +288 -0
- modules/control-plane/reproducibility/commands.md +467 -0
- modules/control-plane/reproducibility/docker_config/Dockerfile +39 -0
- modules/control-plane/reproducibility/experiment_configs/purge_config.json +46 -0
- modules/control-plane/reproducibility/experiment_configs/rag_config.json +36 -0
- modules/control-plane/reproducibility/hardware_specs.md +317 -0
- modules/control-plane/reproducibility/requirements_frozen.txt +0 -0
- modules/control-plane/reproducibility/run_all_experiments.sh +45 -0
- modules/control-plane/reproducibility/seeds.json +106 -0
- modules/control-plane/scripts/prepare_pypi.py +46 -0
- modules/control-plane/scripts/prepare_release.py +176 -0
- modules/control-plane/scripts/upload_dataset_to_hf.py +316 -0
- modules/control-plane/setup.py +69 -0
- modules/control-plane/src/agent_control_plane/__init__.py +639 -0
- modules/control-plane/src/agent_control_plane/a2a_adapter.py +541 -0
- modules/control-plane/src/agent_control_plane/adapter.py +415 -0
- modules/control-plane/src/agent_control_plane/agent_hibernation.py +364 -0
- modules/control-plane/src/agent_control_plane/agent_kernel.py +464 -0
- modules/control-plane/src/agent_control_plane/compliance.py +718 -0
- modules/control-plane/src/agent_control_plane/constraint_graphs.py +475 -0
- modules/control-plane/src/agent_control_plane/control_plane.py +848 -0
- modules/control-plane/src/agent_control_plane/example_executors.py +193 -0
- modules/control-plane/src/agent_control_plane/execution_engine.py +229 -0
- modules/control-plane/src/agent_control_plane/flight_recorder.py +600 -0
- modules/control-plane/src/agent_control_plane/governance_layer.py +432 -0
- modules/control-plane/src/agent_control_plane/hf_utils.py +561 -0
- modules/control-plane/src/agent_control_plane/interfaces/__init__.py +53 -0
- modules/control-plane/src/agent_control_plane/interfaces/kernel_interface.py +359 -0
- modules/control-plane/src/agent_control_plane/interfaces/plugin_interface.py +495 -0
- modules/control-plane/src/agent_control_plane/interfaces/protocol_interfaces.py +385 -0
- modules/control-plane/src/agent_control_plane/kernel_space.py +707 -0
- modules/control-plane/src/agent_control_plane/langchain_adapter.py +422 -0
- modules/control-plane/src/agent_control_plane/lifecycle.py +3111 -0
- modules/control-plane/src/agent_control_plane/mcp_adapter.py +517 -0
- modules/control-plane/src/agent_control_plane/ml_safety.py +560 -0
- modules/control-plane/src/agent_control_plane/multimodal.py +724 -0
- modules/control-plane/src/agent_control_plane/mute_agent.py +419 -0
- modules/control-plane/src/agent_control_plane/observability.py +785 -0
- modules/control-plane/src/agent_control_plane/orchestrator.py +480 -0
- modules/control-plane/src/agent_control_plane/plugin_registry.py +748 -0
- modules/control-plane/src/agent_control_plane/policy_engine.py +525 -0
- modules/control-plane/src/agent_control_plane/shadow_mode.py +307 -0
- modules/control-plane/src/agent_control_plane/signals.py +491 -0
- modules/control-plane/src/agent_control_plane/supervisor_agents.py +427 -0
- modules/control-plane/src/agent_control_plane/time_travel_debugger.py +554 -0
- modules/control-plane/src/agent_control_plane/tool_registry.py +350 -0
- modules/control-plane/src/agent_control_plane/vfs.py +695 -0
- modules/control-plane/tests/README.md +33 -0
- modules/control-plane/tests/test_a2a_adapter.py +336 -0
- modules/control-plane/tests/test_adapter.py +422 -0
- modules/control-plane/tests/test_advanced_features.py +389 -0
- modules/control-plane/tests/test_benchmark.py +223 -0
- modules/control-plane/tests/test_compliance.py +214 -0
- modules/control-plane/tests/test_control_plane.py +295 -0
- modules/control-plane/tests/test_hibernation.py +274 -0
- modules/control-plane/tests/test_kernel_interception.py +284 -0
- modules/control-plane/tests/test_langchain_adapter.py +258 -0
- modules/control-plane/tests/test_lifecycle.py +1174 -0
- modules/control-plane/tests/test_mcp_adapter.py +293 -0
- modules/control-plane/tests/test_ml_safety.py +142 -0
- modules/control-plane/tests/test_multimodal.py +317 -0
- modules/control-plane/tests/test_new_features.py +435 -0
- modules/control-plane/tests/test_observability.py +338 -0
- modules/control-plane/tests/test_time_travel.py +387 -0
- modules/emk/.github/workflows/ci.yml +105 -0
- modules/emk/.github/workflows/publish.yml +144 -0
- modules/emk/.gitignore +74 -0
- modules/emk/CHANGELOG.md +41 -0
- modules/emk/CONTRIBUTING.md +295 -0
- modules/emk/IMPLEMENTATION.md +174 -0
- modules/emk/LICENSE +21 -0
- modules/emk/MANIFEST.in +8 -0
- modules/emk/README.md +135 -0
- modules/emk/RELEASE_NOTES.md +82 -0
- modules/emk/SECURITY.md +52 -0
- modules/emk/codecov.yml +39 -0
- modules/emk/docs/MEMORY_MANAGEMENT.md +285 -0
- modules/emk/emk/__init__.py +106 -0
- modules/emk/emk/hf_utils.py +419 -0
- modules/emk/emk/indexer.py +144 -0
- modules/emk/emk/py.typed +0 -0
- modules/emk/emk/schema.py +204 -0
- modules/emk/emk/sleep_cycle.py +345 -0
- modules/emk/emk/store.py +479 -0
- modules/emk/examples/basic_usage.py +123 -0
- modules/emk/examples/memory_features_demo.py +154 -0
- modules/emk/experiments/README.md +59 -0
- modules/emk/experiments/reproduce_results.py +461 -0
- modules/emk/experiments/results.json +61 -0
- modules/emk/paper/structure.tex +192 -0
- modules/emk/paper/whitepaper.md +273 -0
- modules/emk/pyproject.toml +91 -0
- modules/emk/setup.py +5 -0
- modules/emk/tests/test_file_adapter.py +195 -0
- modules/emk/tests/test_indexer.py +174 -0
- modules/emk/tests/test_init.py +55 -0
- modules/emk/tests/test_negative_memory.py +83 -0
- modules/emk/tests/test_schema.py +150 -0
- modules/emk/tests/test_semantic_rules.py +175 -0
- modules/emk/tests/test_sleep_cycle.py +335 -0
- modules/emk/tests/test_store_anti_patterns.py +239 -0
- modules/iatp/.github/workflows/docker-build.yml +124 -0
- modules/iatp/.github/workflows/publish.yml +174 -0
- modules/iatp/.github/workflows/python-package.yml +121 -0
- modules/iatp/.gitignore +67 -0
- modules/iatp/.pre-commit-config.yaml +64 -0
- modules/iatp/CHANGELOG.md +120 -0
- modules/iatp/Dockerfile +91 -0
- modules/iatp/IMPLEMENTATION_SUMMARY.md +218 -0
- modules/iatp/MANIFEST.in +9 -0
- modules/iatp/README.md +180 -0
- modules/iatp/docker/Dockerfile.agent +27 -0
- modules/iatp/docker/Dockerfile.sidecar-python +86 -0
- modules/iatp/docker/README.md +258 -0
- modules/iatp/docker-compose.yml +194 -0
- modules/iatp/docs/ARCHITECTURE.md +243 -0
- modules/iatp/docs/CLI_GUIDE.md +220 -0
- modules/iatp/docs/DEPLOYMENT.md +304 -0
- modules/iatp/examples/README.md +132 -0
- modules/iatp/examples/backend_agent.py +39 -0
- modules/iatp/examples/client.py +168 -0
- modules/iatp/examples/demo_attestation_reputation.py +274 -0
- modules/iatp/examples/demo_client.py +240 -0
- modules/iatp/examples/demo_rbac.py +143 -0
- modules/iatp/examples/integration_demo.py +245 -0
- modules/iatp/examples/manifests/coder_agent.json +20 -0
- modules/iatp/examples/manifests/reviewer_agent.json +19 -0
- modules/iatp/examples/manifests/secure_bank.json +14 -0
- modules/iatp/examples/manifests/standard_agent.json +14 -0
- modules/iatp/examples/manifests/untrusted_honeypot.json +14 -0
- modules/iatp/examples/run_secure_bank_sidecar.py +85 -0
- modules/iatp/examples/run_sidecar.py +105 -0
- modules/iatp/examples/run_untrusted_sidecar.py +77 -0
- modules/iatp/examples/secure_bank_agent.py +138 -0
- modules/iatp/examples/test_untrusted.py +82 -0
- modules/iatp/examples/untrusted_agent.py +119 -0
- modules/iatp/experiments/README.md +58 -0
- modules/iatp/experiments/cascading_hallucination/README.md +149 -0
- modules/iatp/experiments/cascading_hallucination/agent_a_user.py +41 -0
- modules/iatp/experiments/cascading_hallucination/agent_b_summarizer.py +54 -0
- modules/iatp/experiments/cascading_hallucination/agent_c_database.py +47 -0
- modules/iatp/experiments/cascading_hallucination/proof_of_concept.py +290 -0
- modules/iatp/experiments/cascading_hallucination/run_experiment.py +226 -0
- modules/iatp/experiments/cascading_hallucination/sidecar_c.py +61 -0
- modules/iatp/experiments/reproduce_results.py +574 -0
- modules/iatp/experiments/results.json +2336 -0
- modules/iatp/iatp/__init__.py +164 -0
- modules/iatp/iatp/attestation.py +401 -0
- modules/iatp/iatp/cli.py +253 -0
- modules/iatp/iatp/hf_utils.py +469 -0
- modules/iatp/iatp/ipc_pipes.py +578 -0
- modules/iatp/iatp/main.py +410 -0
- modules/iatp/iatp/models/__init__.py +445 -0
- modules/iatp/iatp/policy_engine.py +335 -0
- modules/iatp/iatp/py.typed +2 -0
- modules/iatp/iatp/recovery.py +319 -0
- modules/iatp/iatp/security/__init__.py +268 -0
- modules/iatp/iatp/sidecar/__init__.py +517 -0
- modules/iatp/iatp/telemetry/__init__.py +162 -0
- modules/iatp/iatp/tests/__init__.py +1 -0
- modules/iatp/iatp/tests/test_attestation.py +368 -0
- modules/iatp/iatp/tests/test_cli.py +129 -0
- modules/iatp/iatp/tests/test_models.py +128 -0
- modules/iatp/iatp/tests/test_policy_engine.py +345 -0
- modules/iatp/iatp/tests/test_recovery.py +279 -0
- modules/iatp/iatp/tests/test_security.py +220 -0
- modules/iatp/iatp/tests/test_sidecar.py +165 -0
- modules/iatp/iatp/tests/test_telemetry.py +173 -0
- modules/iatp/paper/BLOG.md +307 -0
- modules/iatp/paper/PAPER.md +236 -0
- modules/iatp/paper/RFC_SUBMISSION.md +299 -0
- modules/iatp/paper/whitepaper.md +369 -0
- modules/iatp/proto/README.md +200 -0
- modules/iatp/proto/generate_stubs.py +81 -0
- modules/iatp/proto/iatp.proto +552 -0
- modules/iatp/pyproject.toml +180 -0
- modules/iatp/requirements-dev.txt +2 -0
- modules/iatp/requirements.txt +6 -0
- modules/iatp/setup.py +60 -0
- modules/iatp/sidecar/README.md +487 -0
- modules/iatp/sidecar/go/Dockerfile +32 -0
- modules/iatp/sidecar/go/README.md +237 -0
- modules/iatp/sidecar/go/go.mod +8 -0
- modules/iatp/sidecar/go/main.go +488 -0
- modules/iatp/spec/001-handshake.md +436 -0
- modules/iatp/spec/002-reversibility.md +394 -0
- modules/iatp/spec/schema/capability_manifest.json +266 -0
- modules/iatp/test_integration.py +310 -0
- modules/mcp-kernel-server/README.md +261 -0
- modules/mcp-kernel-server/pyproject.toml +60 -0
- modules/mcp-kernel-server/src/mcp_kernel_server/__init__.py +26 -0
- modules/mcp-kernel-server/src/mcp_kernel_server/cli.py +229 -0
- modules/mcp-kernel-server/src/mcp_kernel_server/resources.py +215 -0
- modules/mcp-kernel-server/src/mcp_kernel_server/server.py +562 -0
- modules/mcp-kernel-server/src/mcp_kernel_server/tools.py +1172 -0
- modules/mute-agent/.github/workflows/safety_check.yml +45 -0
- modules/mute-agent/.gitignore +53 -0
- modules/mute-agent/ARCHITECTURE.md +531 -0
- modules/mute-agent/BENCHMARK_GUIDE.md +384 -0
- modules/mute-agent/COMPLETION_SUMMARY.md +293 -0
- modules/mute-agent/EXPERIMENT_SUMMARY.md +318 -0
- modules/mute-agent/IMPLEMENTATION_SUMMARY.md +212 -0
- modules/mute-agent/LICENSE +21 -0
- modules/mute-agent/PHASE3_SUMMARY.md +297 -0
- modules/mute-agent/README.md +360 -0
- modules/mute-agent/STEEL_MAN_RESULTS.md +353 -0
- modules/mute-agent/USAGE.md +505 -0
- modules/mute-agent/V2_IMPLEMENTATION_SUMMARY.md +253 -0
- modules/mute-agent/V2_STEEL_MAN_IMPLEMENTATION.md +274 -0
- modules/mute-agent/VERIFICATION_REPORT.md +435 -0
- modules/mute-agent/charts/cost_comparison.png +0 -0
- modules/mute-agent/charts/cost_vs_ambiguity.png +0 -0
- modules/mute-agent/charts/metrics_comparison.png +0 -0
- modules/mute-agent/charts/scenario_breakdown.png +0 -0
- modules/mute-agent/charts/trace_attack_blocked.html +140 -0
- modules/mute-agent/charts/trace_attack_blocked.png +0 -0
- modules/mute-agent/charts/trace_failure.html +140 -0
- modules/mute-agent/charts/trace_failure.png +0 -0
- modules/mute-agent/charts/trace_success.html +140 -0
- modules/mute-agent/charts/trace_success.png +0 -0
- modules/mute-agent/examples/__init__.py +1 -0
- modules/mute-agent/examples/advanced_example.py +384 -0
- modules/mute-agent/examples/graph_debugger_demo.py +241 -0
- modules/mute-agent/examples/listener_example.py +297 -0
- modules/mute-agent/examples/simple_example.py +242 -0
- modules/mute-agent/examples/steel_man_demo.py +297 -0
- modules/mute-agent/experiments/README.md +135 -0
- modules/mute-agent/experiments/__init__.py +3 -0
- modules/mute-agent/experiments/agent_comparison.csv +6 -0
- modules/mute-agent/experiments/agent_comparison_50runs.csv +6 -0
- modules/mute-agent/experiments/ambiguity_test.py +335 -0
- modules/mute-agent/experiments/ambiguity_test_results.csv +31 -0
- modules/mute-agent/experiments/ambiguity_test_results_50runs.csv +51 -0
- modules/mute-agent/experiments/baseline_agent.py +189 -0
- modules/mute-agent/experiments/benchmark.py +402 -0
- modules/mute-agent/experiments/demo.py +172 -0
- modules/mute-agent/experiments/generate_cost_curve.py +474 -0
- modules/mute-agent/experiments/jailbreak_test.py +137 -0
- modules/mute-agent/experiments/latent_state_scenario.py +361 -0
- modules/mute-agent/experiments/mute_agent_experiment.py +349 -0
- modules/mute-agent/experiments/run_extended_experiment.py +40 -0
- modules/mute-agent/experiments/run_v2_experiments.py +266 -0
- modules/mute-agent/experiments/run_v2_experiments_auto.py +247 -0
- modules/mute-agent/experiments/v2_scenarios/README.md +214 -0
- modules/mute-agent/experiments/v2_scenarios/__init__.py +4 -0
- modules/mute-agent/experiments/v2_scenarios/scenario_1_deep_dependency.py +325 -0
- modules/mute-agent/experiments/v2_scenarios/scenario_2_adversarial.py +328 -0
- modules/mute-agent/experiments/v2_scenarios/scenario_3_false_positive.py +303 -0
- modules/mute-agent/experiments/v2_scenarios/scenario_4_performance.py +319 -0
- modules/mute-agent/experiments/visualize.py +400 -0
- modules/mute-agent/mute_agent/__init__.py +66 -0
- modules/mute-agent/mute_agent/core/__init__.py +1 -0
- modules/mute-agent/mute_agent/core/execution_agent.py +164 -0
- modules/mute-agent/mute_agent/core/handshake_protocol.py +199 -0
- modules/mute-agent/mute_agent/core/reasoning_agent.py +236 -0
- modules/mute-agent/mute_agent/knowledge_graph/__init__.py +1 -0
- modules/mute-agent/mute_agent/knowledge_graph/graph_elements.py +63 -0
- modules/mute-agent/mute_agent/knowledge_graph/multidimensional_graph.py +168 -0
- modules/mute-agent/mute_agent/knowledge_graph/subgraph.py +222 -0
- modules/mute-agent/mute_agent/listener/__init__.py +41 -0
- modules/mute-agent/mute_agent/listener/adapters/__init__.py +29 -0
- modules/mute-agent/mute_agent/listener/adapters/base_adapter.py +187 -0
- modules/mute-agent/mute_agent/listener/adapters/caas_adapter.py +342 -0
- modules/mute-agent/mute_agent/listener/adapters/control_plane_adapter.py +434 -0
- modules/mute-agent/mute_agent/listener/adapters/iatp_adapter.py +330 -0
- modules/mute-agent/mute_agent/listener/adapters/scak_adapter.py +249 -0
- modules/mute-agent/mute_agent/listener/listener.py +608 -0
- modules/mute-agent/mute_agent/listener/state_observer.py +434 -0
- modules/mute-agent/mute_agent/listener/threshold_config.py +311 -0
- modules/mute-agent/mute_agent/super_system/__init__.py +1 -0
- modules/mute-agent/mute_agent/super_system/router.py +202 -0
- modules/mute-agent/mute_agent/visualization/__init__.py +8 -0
- modules/mute-agent/mute_agent/visualization/graph_debugger.py +495 -0
- modules/mute-agent/requirements-dev.txt +6 -0
- modules/mute-agent/requirements.txt +9 -0
- modules/mute-agent/setup.py +64 -0
- modules/mute-agent/src/__init__.py +0 -0
- modules/mute-agent/src/agents/__init__.py +0 -0
- modules/mute-agent/src/agents/baseline_agent.py +524 -0
- modules/mute-agent/src/agents/interactive_agent.py +113 -0
- modules/mute-agent/src/agents/mute_agent.py +622 -0
- modules/mute-agent/src/benchmarks/__init__.py +0 -0
- modules/mute-agent/src/benchmarks/evaluator.py +481 -0
- modules/mute-agent/src/benchmarks/scenarios.json +985 -0
- modules/mute-agent/src/core/__init__.py +0 -0
- modules/mute-agent/src/core/mock_state.py +320 -0
- modules/mute-agent/src/core/tools.py +441 -0
- modules/nexus/__init__.py +49 -0
- modules/nexus/arbiter.py +357 -0
- modules/nexus/client.py +464 -0
- modules/nexus/dmz.py +417 -0
- modules/nexus/escrow.py +428 -0
- modules/nexus/exceptions.py +284 -0
- modules/nexus/registry.py +391 -0
- modules/nexus/reputation.py +423 -0
- modules/nexus/schemas/__init__.py +49 -0
- modules/nexus/schemas/compliance.py +274 -0
- modules/nexus/schemas/escrow.py +249 -0
- modules/nexus/schemas/manifest.py +223 -0
- modules/nexus/schemas/receipt.py +206 -0
- modules/observability/README.md +192 -0
- modules/observability/alertmanager/alertmanager.yml +116 -0
- modules/observability/alerts/agent-os-alerts.yaml +197 -0
- modules/observability/docker-compose.yml +128 -0
- modules/observability/grafana/dashboards/agent-os-amb.json +448 -0
- modules/observability/grafana/dashboards/agent-os-cmvk.json +441 -0
- modules/observability/grafana/dashboards/agent-os-overview.json +268 -0
- modules/observability/grafana/dashboards/agent-os-performance.json +15 -0
- modules/observability/grafana/dashboards/agent-os-safety.json +50 -0
- modules/observability/grafana/provisioning/dashboards/dashboards.yml +15 -0
- modules/observability/grafana/provisioning/datasources/datasources.yml +33 -0
- modules/observability/otel/otel-collector-config.yml +61 -0
- modules/observability/prometheus/prometheus.yml +63 -0
- modules/observability/pyproject.toml +53 -0
- modules/observability/scripts/export_dashboards.py +55 -0
- modules/observability/src/agent_os_observability/__init__.py +25 -0
- modules/observability/src/agent_os_observability/dashboards.py +896 -0
- modules/observability/src/agent_os_observability/metrics.py +396 -0
- modules/observability/src/agent_os_observability/server.py +221 -0
- modules/observability/src/agent_os_observability/tracer.py +226 -0
- modules/primitives/.gitignore +8 -0
- modules/primitives/README.md +62 -0
- modules/primitives/agent_primitives/__init__.py +22 -0
- modules/primitives/agent_primitives/failures.py +82 -0
- modules/primitives/agent_primitives/py.typed +0 -0
- modules/primitives/pyproject.toml +68 -0
- modules/scak/.github/copilot-instructions.md +396 -0
- modules/scak/.github/workflows/release.yml +117 -0
- modules/scak/.gitignore +32 -0
- modules/scak/CHANGELOG.md +173 -0
- modules/scak/CITATION.cff +62 -0
- modules/scak/CONTRIBUTING.md +429 -0
- modules/scak/Dockerfile +58 -0
- modules/scak/ENTERPRISE_FEATURES.md +518 -0
- modules/scak/IMPLEMENTATION_SUMMARY.md +206 -0
- modules/scak/LIMITATIONS.md +565 -0
- modules/scak/MANIFEST.in +16 -0
- modules/scak/NOVELTY.md +535 -0
- modules/scak/README.md +928 -0
- modules/scak/RESEARCH.md +670 -0
- modules/scak/agent_kernel/__init__.py +66 -0
- modules/scak/agent_kernel/analyzer.py +432 -0
- modules/scak/agent_kernel/auditor.py +31 -0
- modules/scak/agent_kernel/completeness_auditor.py +234 -0
- modules/scak/agent_kernel/detector.py +200 -0
- modules/scak/agent_kernel/kernel.py +741 -0
- modules/scak/agent_kernel/memory_manager.py +82 -0
- modules/scak/agent_kernel/models.py +372 -0
- modules/scak/agent_kernel/nudge_mechanism.py +260 -0
- modules/scak/agent_kernel/outcome_analyzer.py +335 -0
- modules/scak/agent_kernel/patcher.py +579 -0
- modules/scak/agent_kernel/semantic_analyzer.py +313 -0
- modules/scak/agent_kernel/semantic_purge.py +346 -0
- modules/scak/agent_kernel/simulator.py +447 -0
- modules/scak/agent_kernel/teacher.py +82 -0
- modules/scak/agent_kernel/triage.py +149 -0
- modules/scak/build_and_publish.ps1 +74 -0
- modules/scak/build_and_publish.sh +74 -0
- modules/scak/cli.py +471 -0
- modules/scak/dashboard.py +462 -0
- modules/scak/datasets/DATASET_CARD.md +219 -0
- modules/scak/datasets/README.md +143 -0
- modules/scak/datasets/gaia_vague_queries/vague_queries.json +262 -0
- modules/scak/datasets/hf_upload/README.md +219 -0
- modules/scak/datasets/hf_upload/scak_gaia_laziness.jsonl +50 -0
- modules/scak/datasets/prepare_hf_datasets.py +145 -0
- modules/scak/datasets/red_team/jailbreak_patterns.json +202 -0
- modules/scak/docker-compose.yml +99 -0
- modules/scak/docs/Adaptive-Memory-Hierarchy.md +319 -0
- modules/scak/docs/Data-Contracts-and-Schemas.md +285 -0
- modules/scak/docs/Dual-Loop-Architecture.md +344 -0
- modules/scak/docs/Enhanced-Features.md +612 -0
- modules/scak/docs/LANGCHAIN_INTEGRATION.md +572 -0
- modules/scak/docs/README.md +128 -0
- modules/scak/docs/Reference-Implementations.md +163 -0
- modules/scak/docs/SCAK_V2.md +374 -0
- modules/scak/docs/Three-Failure-Types.md +178 -0
- modules/scak/examples/basic_example.py +155 -0
- modules/scak/examples/circuit_breaker_lazy_eval_demo.py +243 -0
- modules/scak/examples/langchain_integration_example.py +339 -0
- modules/scak/examples/layer4_demo.py +243 -0
- modules/scak/examples/production_features_demo.py +353 -0
- modules/scak/examples/quick_demo.py +79 -0
- modules/scak/examples/scak_v2_demo.py +252 -0
- modules/scak/experiments/README.md +438 -0
- modules/scak/experiments/ablation_studies/README.md +192 -0
- modules/scak/experiments/ablation_studies/ablation_no_audit.py +116 -0
- modules/scak/experiments/ablation_studies/ablation_no_purge.py +133 -0
- modules/scak/experiments/chaos_engineering/README.md +332 -0
- modules/scak/experiments/context_efficiency_test.py +328 -0
- modules/scak/experiments/gaia_benchmark/README.md +208 -0
- modules/scak/experiments/laziness_benchmark.py +179 -0
- modules/scak/experiments/long_horizon_task_experiment.py +252 -0
- modules/scak/experiments/multi_agent_rag_experiment.py +284 -0
- modules/scak/experiments/results/ablation_table.md +12 -0
- modules/scak/experiments/results/long_horizon.json +36 -0
- modules/scak/experiments/results/multi_agent_rag.json +66 -0
- modules/scak/experiments/run_comprehensive_ablations.py +332 -0
- modules/scak/experiments/test_auditor_patcher_integration.py +251 -0
- modules/scak/notebooks/getting_started.ipynb +33 -0
- modules/scak/paper/ARXIV_SUBMISSION_METADATA.txt +109 -0
- modules/scak/paper/PAPER_CHECKLIST.md +304 -0
- modules/scak/paper/Paper.pdf +0 -0
- modules/scak/paper/README.md +113 -0
- modules/scak/paper/appendix.md +351 -0
- modules/scak/paper/arxiv/bibliography.bib +284 -0
- modules/scak/paper/arxiv/fig1_ooda_architecture.pdf +0 -0
- modules/scak/paper/arxiv/fig2_memory_hierarchy.pdf +0 -0
- modules/scak/paper/arxiv/fig3_gaia_results.pdf +0 -0
- modules/scak/paper/arxiv/fig4_ablation_heatmap.pdf +0 -0
- modules/scak/paper/arxiv/fig5_context_reduction.pdf +0 -0
- modules/scak/paper/arxiv/fig6_mttr_boxplot.pdf +0 -0
- modules/scak/paper/arxiv/main.aux +103 -0
- modules/scak/paper/arxiv/main.bbl +113 -0
- modules/scak/paper/arxiv/main.blg +55 -0
- modules/scak/paper/arxiv/main.out +31 -0
- modules/scak/paper/arxiv/main.pdf +0 -0
- modules/scak/paper/arxiv/main.tex +482 -0
- modules/scak/paper/arxiv_submission/bibliography.bib +284 -0
- modules/scak/paper/arxiv_submission/fig1_ooda_architecture.pdf +0 -0
- modules/scak/paper/arxiv_submission/fig2_memory_hierarchy.pdf +0 -0
- modules/scak/paper/arxiv_submission/fig3_gaia_results.pdf +0 -0
- modules/scak/paper/arxiv_submission/fig4_ablation_heatmap.pdf +0 -0
- modules/scak/paper/arxiv_submission/fig5_context_reduction.pdf +0 -0
- modules/scak/paper/arxiv_submission/fig6_mttr_boxplot.pdf +0 -0
- modules/scak/paper/arxiv_submission/main.aux +103 -0
- modules/scak/paper/arxiv_submission/main.bbl +113 -0
- modules/scak/paper/arxiv_submission/main.blg +55 -0
- modules/scak/paper/arxiv_submission/main.out +31 -0
- modules/scak/paper/arxiv_submission/main.pdf +0 -0
- modules/scak/paper/arxiv_submission/main.tex +482 -0
- modules/scak/paper/arxiv_submission.tar.gz +0 -0
- modules/scak/paper/bibliography.bib +284 -0
- modules/scak/paper/build.sh +55 -0
- modules/scak/paper/figures/README.md +32 -0
- modules/scak/paper/figures/fig1_ooda_architecture.md +75 -0
- modules/scak/paper/figures/fig1_ooda_architecture.pdf +0 -0
- modules/scak/paper/figures/fig1_ooda_architecture.png +0 -0
- modules/scak/paper/figures/fig2_memory_hierarchy.md +83 -0
- modules/scak/paper/figures/fig2_memory_hierarchy.pdf +0 -0
- modules/scak/paper/figures/fig2_memory_hierarchy.png +0 -0
- modules/scak/paper/figures/fig3_gaia_results.md +64 -0
- modules/scak/paper/figures/fig3_gaia_results.pdf +0 -0
- modules/scak/paper/figures/fig3_gaia_results.png +0 -0
- modules/scak/paper/figures/fig4_ablation_heatmap.md +64 -0
- modules/scak/paper/figures/fig4_ablation_heatmap.pdf +0 -0
- modules/scak/paper/figures/fig4_ablation_heatmap.png +0 -0
- modules/scak/paper/figures/fig5_context_reduction.md +71 -0
- modules/scak/paper/figures/fig5_context_reduction.pdf +0 -0
- modules/scak/paper/figures/fig5_context_reduction.png +0 -0
- modules/scak/paper/figures/fig6_mttr_boxplot.md +80 -0
- modules/scak/paper/figures/fig6_mttr_boxplot.pdf +0 -0
- modules/scak/paper/figures/fig6_mttr_boxplot.png +0 -0
- modules/scak/paper/figures/generate_figures.py +463 -0
- modules/scak/paper/main.aux +103 -0
- modules/scak/paper/main.bbl +113 -0
- modules/scak/paper/main.blg +55 -0
- modules/scak/paper/main.md +192 -0
- modules/scak/paper/main.out +31 -0
- modules/scak/paper/main.pdf +0 -0
- modules/scak/paper/main.tex +482 -0
- modules/scak/reproducibility/ABLATIONS.md +225 -0
- modules/scak/reproducibility/Dockerfile.reproducibility +34 -0
- modules/scak/reproducibility/README.md +421 -0
- modules/scak/reproducibility/requirements-pinned.txt +32 -0
- modules/scak/reproducibility/run_all_experiments.py +395 -0
- modules/scak/reproducibility/seed_control.py +53 -0
- modules/scak/reproducibility/statistical_analysis.py +302 -0
- modules/scak/requirements.txt +50 -0
- modules/scak/setup.py +93 -0
- modules/scak/src/__init__.py +124 -0
- modules/scak/src/agents/__init__.py +13 -0
- modules/scak/src/agents/conflict_resolution.py +732 -0
- modules/scak/src/agents/orchestrator.py +761 -0
- modules/scak/src/agents/pubsub.py +484 -0
- modules/scak/src/agents/shadow_teacher.py +344 -0
- modules/scak/src/agents/swarm.py +661 -0
- modules/scak/src/agents/worker.py +357 -0
- modules/scak/src/integrations/__init__.py +81 -0
- modules/scak/src/integrations/cmvk_adapter.py +430 -0
- modules/scak/src/integrations/control_plane_adapter.py +601 -0
- modules/scak/src/integrations/langchain_integration.py +902 -0
- modules/scak/src/interfaces/__init__.py +59 -0
- modules/scak/src/interfaces/llm_clients.py +505 -0
- modules/scak/src/interfaces/openapi_tools.py +611 -0
- modules/scak/src/interfaces/plugin_system.py +605 -0
- modules/scak/src/interfaces/protocols.py +365 -0
- modules/scak/src/interfaces/telemetry.py +464 -0
- modules/scak/src/interfaces/tool_registry.py +547 -0
- modules/scak/src/kernel/__init__.py +100 -0
- modules/scak/src/kernel/auditor.py +305 -0
- modules/scak/src/kernel/circuit_breaker.py +398 -0
- modules/scak/src/kernel/core.py +724 -0
- modules/scak/src/kernel/distributed.py +667 -0
- modules/scak/src/kernel/evolution.py +455 -0
- modules/scak/src/kernel/failover.py +621 -0
- modules/scak/src/kernel/governance.py +710 -0
- modules/scak/src/kernel/governance_v2.py +603 -0
- modules/scak/src/kernel/lazy_evaluator.py +514 -0
- modules/scak/src/kernel/load_testing.py +633 -0
- modules/scak/src/kernel/memory.py +945 -0
- modules/scak/src/kernel/patcher.py +581 -0
- modules/scak/src/kernel/rubric.py +419 -0
- modules/scak/src/kernel/schemas.py +390 -0
- modules/scak/src/kernel/skill_mapper.py +309 -0
- modules/scak/src/kernel/triage.py +149 -0
- modules/scak/src/mocks/__init__.py +99 -0
- modules/scak/tests/__init__.py +1 -0
- modules/scak/tests/test_circuit_breaker.py +403 -0
- modules/scak/tests/test_conflict_resolution.py +287 -0
- modules/scak/tests/test_dual_loop.py +463 -0
- modules/scak/tests/test_enhanced_features.py +421 -0
- modules/scak/tests/test_failover_and_load.py +438 -0
- modules/scak/tests/test_governance.py +185 -0
- modules/scak/tests/test_kernel.py +359 -0
- modules/scak/tests/test_langchain_integration.py +451 -0
- modules/scak/tests/test_lazy_evaluator.py +465 -0
- modules/scak/tests/test_llm_clients.py +122 -0
- modules/scak/tests/test_memory_controller.py +528 -0
- modules/scak/tests/test_orchestrator.py +181 -0
- modules/scak/tests/test_phase3_integration.py +265 -0
- modules/scak/tests/test_pubsub_swarm.py +203 -0
- modules/scak/tests/test_reference_implementations.py +240 -0
- modules/scak/tests/test_rubric.py +363 -0
- modules/scak/tests/test_scak_v2.py +651 -0
- modules/scak/tests/test_skill_mapper.py +217 -0
- modules/scak/tests/test_specific_failures.py +393 -0
- modules/scak/tests/test_tool_registry.py +264 -0
- modules/scak/tests/test_tools_and_plugins.py +303 -0
- modules/scak/tests/test_triage.py +596 -0
- modules/scak/tests/test_write_through.py +319 -0
- agent_os_kernel-1.1.0.dist-info/METADATA +0 -400
- agent_os_kernel-1.1.0.dist-info/RECORD +0 -12
- {agent_os_kernel-1.1.0.dist-info → agent_os_kernel-1.3.0.dist-info}/WHEEL +0 -0
- {agent_os_kernel-1.1.0.dist-info → agent_os_kernel-1.3.0.dist-info}/licenses/LICENSE +0 -0
|
@@ -0,0 +1,105 @@
|
|
|
1
|
+
<?xml version="1.0" encoding="UTF-8"?>
|
|
2
|
+
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 700 500" font-family="Arial, sans-serif">
|
|
3
|
+
<defs>
|
|
4
|
+
<linearGradient id="hotGrad2" x1="0%" y1="0%" x2="100%" y2="100%">
|
|
5
|
+
<stop offset="0%" style="stop-color:#FF6B6B;stop-opacity:1" />
|
|
6
|
+
<stop offset="100%" style="stop-color:#EE5A5A;stop-opacity:1" />
|
|
7
|
+
</linearGradient>
|
|
8
|
+
<linearGradient id="warmGrad2" x1="0%" y1="0%" x2="100%" y2="100%">
|
|
9
|
+
<stop offset="0%" style="stop-color:#FFB347;stop-opacity:1" />
|
|
10
|
+
<stop offset="100%" style="stop-color:#FFA500;stop-opacity:1" />
|
|
11
|
+
</linearGradient>
|
|
12
|
+
<linearGradient id="coldGrad2" x1="0%" y1="0%" x2="100%" y2="100%">
|
|
13
|
+
<stop offset="0%" style="stop-color:#5DADE2;stop-opacity:1" />
|
|
14
|
+
<stop offset="100%" style="stop-color:#3498DB;stop-opacity:1" />
|
|
15
|
+
</linearGradient>
|
|
16
|
+
<filter id="shadow2" x="-20%" y="-20%" width="140%" height="140%">
|
|
17
|
+
<feDropShadow dx="3" dy="3" stdDeviation="4" flood-opacity="0.2"/>
|
|
18
|
+
</filter>
|
|
19
|
+
</defs>
|
|
20
|
+
|
|
21
|
+
<!-- Title -->
|
|
22
|
+
<text x="350" y="35" text-anchor="middle" font-size="24" font-weight="bold" fill="#2C3E50">Context Triad: Intimacy-Based Prioritization</text>
|
|
23
|
+
<text x="350" y="58" text-anchor="middle" font-size="14" fill="#6C757D">Token budget allocation based on contextual intimacy</text>
|
|
24
|
+
|
|
25
|
+
<!-- Triangle visualization -->
|
|
26
|
+
<polygon points="350,90 150,380 550,380" fill="none" stroke="#DEE2E6" stroke-width="3" stroke-dasharray="8,4"/>
|
|
27
|
+
|
|
28
|
+
<!-- HOT (Top) -->
|
|
29
|
+
<circle cx="350" cy="120" r="70" fill="url(#hotGrad2)" filter="url(#shadow2)"/>
|
|
30
|
+
<text x="350" y="105" text-anchor="middle" font-size="28" fill="white">🔥</text>
|
|
31
|
+
<text x="350" y="130" text-anchor="middle" font-size="16" font-weight="bold" fill="white">HOT</text>
|
|
32
|
+
<text x="350" y="148" text-anchor="middle" font-size="11" fill="white">2,000 tokens</text>
|
|
33
|
+
|
|
34
|
+
<!-- Hot details box -->
|
|
35
|
+
<rect x="450" y="75" width="220" height="100" rx="8" fill="white" stroke="#EE5A5A" stroke-width="2" filter="url(#shadow2)"/>
|
|
36
|
+
<text x="560" y="100" text-anchor="middle" font-size="13" font-weight="bold" fill="#EE5A5A">🔥 Hot Context</text>
|
|
37
|
+
<line x1="470" y1="110" x2="650" y2="110" stroke="#EEE" stroke-width="1"/>
|
|
38
|
+
<text x="480" y="130" font-size="11" fill="#495057">• Current conversation turns</text>
|
|
39
|
+
<text x="480" y="148" font-size="11" fill="#495057">• Last 10 messages</text>
|
|
40
|
+
<text x="480" y="166" font-size="11" fill="#495057">• Highest retrieval priority</text>
|
|
41
|
+
|
|
42
|
+
<!-- Arrow to hot box -->
|
|
43
|
+
<path d="M420 120 L448 100" stroke="#EE5A5A" stroke-width="2" fill="none"/>
|
|
44
|
+
|
|
45
|
+
<!-- WARM (Bottom Left) -->
|
|
46
|
+
<circle cx="200" cy="340" r="70" fill="url(#warmGrad2)" filter="url(#shadow2)"/>
|
|
47
|
+
<text x="200" y="325" text-anchor="middle" font-size="28" fill="white">🌡️</text>
|
|
48
|
+
<text x="200" y="350" text-anchor="middle" font-size="16" font-weight="bold" fill="white">WARM</text>
|
|
49
|
+
<text x="200" y="368" text-anchor="middle" font-size="11" fill="white">1,000 tokens</text>
|
|
50
|
+
|
|
51
|
+
<!-- Warm details box -->
|
|
52
|
+
<rect x="30" y="200" width="200" height="100" rx="8" fill="white" stroke="#FFA500" stroke-width="2" filter="url(#shadow2)"/>
|
|
53
|
+
<text x="130" y="225" text-anchor="middle" font-size="13" font-weight="bold" fill="#FFA500">🌡️ Warm Context</text>
|
|
54
|
+
<line x1="50" y1="235" x2="210" y2="235" stroke="#EEE" stroke-width="1"/>
|
|
55
|
+
<text x="60" y="255" font-size="11" fill="#495057">• User preferences</text>
|
|
56
|
+
<text x="60" y="273" font-size="11" fill="#495057">• Recent documents</text>
|
|
57
|
+
<text x="60" y="291" font-size="11" fill="#495057">• Session memory</text>
|
|
58
|
+
|
|
59
|
+
<!-- Arrow to warm box -->
|
|
60
|
+
<path d="M160 290 L145 300" stroke="#FFA500" stroke-width="2" fill="none"/>
|
|
61
|
+
|
|
62
|
+
<!-- COLD (Bottom Right) -->
|
|
63
|
+
<circle cx="500" cy="340" r="70" fill="url(#coldGrad2)" filter="url(#shadow2)"/>
|
|
64
|
+
<text x="500" y="325" text-anchor="middle" font-size="28" fill="white">❄️</text>
|
|
65
|
+
<text x="500" y="350" text-anchor="middle" font-size="16" font-weight="bold" fill="white">COLD</text>
|
|
66
|
+
<text x="500" y="368" text-anchor="middle" font-size="11" fill="white">5,000 tokens</text>
|
|
67
|
+
|
|
68
|
+
<!-- Cold details box -->
|
|
69
|
+
<rect x="470" y="200" width="200" height="100" rx="8" fill="white" stroke="#3498DB" stroke-width="2" filter="url(#shadow2)"/>
|
|
70
|
+
<text x="570" y="225" text-anchor="middle" font-size="13" font-weight="bold" fill="#3498DB">❄️ Cold Context</text>
|
|
71
|
+
<line x1="490" y1="235" x2="650" y2="235" stroke="#EEE" stroke-width="1"/>
|
|
72
|
+
<text x="500" y="255" font-size="11" fill="#495057">• Historical archives</text>
|
|
73
|
+
<text x="500" y="273" font-size="11" fill="#495057">• Reference documentation</text>
|
|
74
|
+
<text x="500" y="291" font-size="11" fill="#495057">• Background knowledge</text>
|
|
75
|
+
|
|
76
|
+
<!-- Arrow to cold box -->
|
|
77
|
+
<path d="M540 290 L555 300" stroke="#3498DB" stroke-width="2" fill="none"/>
|
|
78
|
+
|
|
79
|
+
<!-- Central LLM -->
|
|
80
|
+
<rect x="290" y="230" width="120" height="50" rx="25" fill="#2C3E50" filter="url(#shadow2)"/>
|
|
81
|
+
<text x="350" y="260" text-anchor="middle" font-size="14" font-weight="bold" fill="white">🤖 LLM</text>
|
|
82
|
+
|
|
83
|
+
<!-- Arrows to LLM -->
|
|
84
|
+
<path d="M350 190 L350 230" stroke="#EE5A5A" stroke-width="3" fill="none" marker-end="url(#redarrow)"/>
|
|
85
|
+
<path d="M250 300 L295 260" stroke="#FFA500" stroke-width="3" fill="none" marker-end="url(#orangearrow)"/>
|
|
86
|
+
<path d="M450 300 L405 260" stroke="#3498DB" stroke-width="3" fill="none" marker-end="url(#bluearrow2)"/>
|
|
87
|
+
|
|
88
|
+
<defs>
|
|
89
|
+
<marker id="redarrow" markerWidth="10" markerHeight="7" refX="9" refY="3.5" orient="auto">
|
|
90
|
+
<polygon points="0 0, 10 3.5, 0 7" fill="#EE5A5A"/>
|
|
91
|
+
</marker>
|
|
92
|
+
<marker id="orangearrow" markerWidth="10" markerHeight="7" refX="9" refY="3.5" orient="auto">
|
|
93
|
+
<polygon points="0 0, 10 3.5, 0 7" fill="#FFA500"/>
|
|
94
|
+
</marker>
|
|
95
|
+
<marker id="bluearrow2" markerWidth="10" markerHeight="7" refX="9" refY="3.5" orient="auto">
|
|
96
|
+
<polygon points="0 0, 10 3.5, 0 7" fill="#3498DB"/>
|
|
97
|
+
</marker>
|
|
98
|
+
</defs>
|
|
99
|
+
|
|
100
|
+
<!-- Token Budget Summary -->
|
|
101
|
+
<rect x="150" y="420" width="400" height="70" rx="10" fill="#F8F9FA" stroke="#DEE2E6" stroke-width="2"/>
|
|
102
|
+
<text x="350" y="445" text-anchor="middle" font-size="14" font-weight="bold" fill="#2C3E50">Total Token Budget: 8,000</text>
|
|
103
|
+
<text x="350" y="468" text-anchor="middle" font-size="12" fill="#6C757D">71% average utilization → 5,820 useful tokens per request</text>
|
|
104
|
+
<text x="350" y="485" text-anchor="middle" font-size="11" fill="#27AE60" font-weight="bold">Precision truncation, not lossy summarization</text>
|
|
105
|
+
</svg>
|
|
Binary file
|
|
@@ -0,0 +1,113 @@
|
|
|
1
|
+
<?xml version="1.0" encoding="UTF-8"?>
|
|
2
|
+
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 800 500" font-family="Arial, sans-serif">
|
|
3
|
+
<defs>
|
|
4
|
+
<linearGradient id="barGrad" x1="0%" y1="0%" x2="100%" y2="0%">
|
|
5
|
+
<stop offset="0%" style="stop-color:#3498DB;stop-opacity:1" />
|
|
6
|
+
<stop offset="100%" style="stop-color:#2980B9;stop-opacity:1" />
|
|
7
|
+
</linearGradient>
|
|
8
|
+
<linearGradient id="fullGrad" x1="0%" y1="0%" x2="100%" y2="0%">
|
|
9
|
+
<stop offset="0%" style="stop-color:#27AE60;stop-opacity:1" />
|
|
10
|
+
<stop offset="100%" style="stop-color:#1E8449;stop-opacity:1" />
|
|
11
|
+
</linearGradient>
|
|
12
|
+
<filter id="barShadow" x="-5%" y="-5%" width="110%" height="120%">
|
|
13
|
+
<feDropShadow dx="1" dy="2" stdDeviation="2" flood-opacity="0.15"/>
|
|
14
|
+
</filter>
|
|
15
|
+
</defs>
|
|
16
|
+
|
|
17
|
+
<!-- Title -->
|
|
18
|
+
<text x="400" y="35" text-anchor="middle" font-size="22" font-weight="bold" fill="#2C3E50">Ablation Study: Component Contributions to Precision@5</text>
|
|
19
|
+
<text x="400" y="58" text-anchor="middle" font-size="13" fill="#6C757D">Each component's individual contribution measured by disabling others</text>
|
|
20
|
+
|
|
21
|
+
<!-- Chart area -->
|
|
22
|
+
<rect x="180" y="80" width="580" height="340" fill="#FAFAFA" stroke="#E0E0E0" stroke-width="1"/>
|
|
23
|
+
|
|
24
|
+
<!-- Y-axis -->
|
|
25
|
+
<line x1="180" y1="80" x2="180" y2="420" stroke="#495057" stroke-width="2"/>
|
|
26
|
+
<text x="30" y="260" text-anchor="middle" font-size="14" font-weight="bold" fill="#2C3E50" transform="rotate(-90, 30, 260)">Precision@5</text>
|
|
27
|
+
|
|
28
|
+
<!-- Y-axis labels -->
|
|
29
|
+
<text x="170" y="420" text-anchor="end" font-size="11" fill="#6C757D">0.50</text>
|
|
30
|
+
<line x1="175" y1="416" x2="760" y2="416" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
31
|
+
|
|
32
|
+
<text x="170" y="360" text-anchor="end" font-size="11" fill="#6C757D">0.60</text>
|
|
33
|
+
<line x1="175" y1="356" x2="760" y2="356" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
34
|
+
|
|
35
|
+
<text x="170" y="300" text-anchor="end" font-size="11" fill="#6C757D">0.70</text>
|
|
36
|
+
<line x1="175" y1="296" x2="760" y2="296" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
37
|
+
|
|
38
|
+
<text x="170" y="240" text-anchor="end" font-size="11" fill="#6C757D">0.80</text>
|
|
39
|
+
<line x1="175" y1="236" x2="760" y2="236" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
40
|
+
|
|
41
|
+
<text x="170" y="180" text-anchor="end" font-size="11" fill="#6C757D">0.90</text>
|
|
42
|
+
<line x1="175" y1="176" x2="760" y2="176" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
43
|
+
|
|
44
|
+
<!-- X-axis -->
|
|
45
|
+
<line x1="180" y1="420" x2="760" y2="420" stroke="#495057" stroke-width="2"/>
|
|
46
|
+
|
|
47
|
+
<!-- Baseline reference line at 0.64 -->
|
|
48
|
+
<line x1="180" y1="332" x2="760" y2="332" stroke="#E74C3C" stroke-width="2" stroke-dasharray="8,4"/>
|
|
49
|
+
<text x="765" y="336" font-size="10" fill="#E74C3C" font-weight="bold">Baseline (0.64)</text>
|
|
50
|
+
|
|
51
|
+
<!-- Bars - scaled: 0.50 = 420, 0.90 = 176, so each 0.01 = 6.1px -->
|
|
52
|
+
<!-- Height formula: 420 - ((value - 0.50) * 600) -->
|
|
53
|
+
|
|
54
|
+
<!-- Bar 1: Baseline 0.640 -->
|
|
55
|
+
<rect x="205" y="332" width="70" height="88" fill="#95A5A6" filter="url(#barShadow)" rx="3"/>
|
|
56
|
+
<text x="240" y="322" text-anchor="middle" font-size="12" font-weight="bold" fill="#2C3E50">0.640</text>
|
|
57
|
+
<text x="240" y="445" text-anchor="middle" font-size="10" fill="#495057">Baseline</text>
|
|
58
|
+
<text x="240" y="458" text-anchor="middle" font-size="9" fill="#6C757D">(Naive)</text>
|
|
59
|
+
|
|
60
|
+
<!-- Bar 2: +Structure 0.740 (+15.6%) -->
|
|
61
|
+
<rect x="295" y="272" width="70" height="148" fill="url(#barGrad)" filter="url(#barShadow)" rx="3"/>
|
|
62
|
+
<text x="330" y="262" text-anchor="middle" font-size="12" font-weight="bold" fill="#2C3E50">0.740</text>
|
|
63
|
+
<text x="330" y="248" text-anchor="middle" font-size="10" fill="#27AE60" font-weight="bold">+15.6%</text>
|
|
64
|
+
<text x="330" y="445" text-anchor="middle" font-size="10" fill="#495057">+Structure</text>
|
|
65
|
+
<text x="330" y="458" text-anchor="middle" font-size="9" fill="#6C757D">Aware</text>
|
|
66
|
+
|
|
67
|
+
<!-- Bar 3: +Metadata 0.720 (+12.5%) -->
|
|
68
|
+
<rect x="385" y="284" width="70" height="136" fill="url(#barGrad)" filter="url(#barShadow)" rx="3"/>
|
|
69
|
+
<text x="420" y="274" text-anchor="middle" font-size="12" font-weight="bold" fill="#2C3E50">0.720</text>
|
|
70
|
+
<text x="420" y="260" text-anchor="middle" font-size="10" fill="#27AE60" font-weight="bold">+12.5%</text>
|
|
71
|
+
<text x="420" y="445" text-anchor="middle" font-size="10" fill="#495057">+Metadata</text>
|
|
72
|
+
<text x="420" y="458" text-anchor="middle" font-size="9" fill="#6C757D">Injection</text>
|
|
73
|
+
|
|
74
|
+
<!-- Bar 4: +Time Decay 0.700 (+9.4%) -->
|
|
75
|
+
<rect x="475" y="296" width="70" height="124" fill="url(#barGrad)" filter="url(#barShadow)" rx="3"/>
|
|
76
|
+
<text x="510" y="286" text-anchor="middle" font-size="12" font-weight="bold" fill="#2C3E50">0.700</text>
|
|
77
|
+
<text x="510" y="272" text-anchor="middle" font-size="10" fill="#27AE60" font-weight="bold">+9.4%</text>
|
|
78
|
+
<text x="510" y="445" text-anchor="middle" font-size="10" fill="#495057">+Time</text>
|
|
79
|
+
<text x="510" y="458" text-anchor="middle" font-size="9" fill="#6C757D">Decay</text>
|
|
80
|
+
|
|
81
|
+
<!-- Bar 5: +Pragmatic 0.680 (+6.3%) -->
|
|
82
|
+
<rect x="565" y="308" width="70" height="112" fill="url(#barGrad)" filter="url(#barShadow)" rx="3"/>
|
|
83
|
+
<text x="600" y="298" text-anchor="middle" font-size="12" font-weight="bold" fill="#2C3E50">0.680</text>
|
|
84
|
+
<text x="600" y="284" text-anchor="middle" font-size="10" fill="#27AE60" font-weight="bold">+6.3%</text>
|
|
85
|
+
<text x="600" y="445" text-anchor="middle" font-size="10" fill="#495057">+Pragmatic</text>
|
|
86
|
+
<text x="600" y="458" text-anchor="middle" font-size="9" fill="#6C757D">Truth</text>
|
|
87
|
+
|
|
88
|
+
<!-- Bar 6: Full CaaS 0.820 (+28.1%) -->
|
|
89
|
+
<rect x="655" y="224" width="80" height="196" fill="url(#fullGrad)" filter="url(#barShadow)" rx="3"/>
|
|
90
|
+
<text x="695" y="214" text-anchor="middle" font-size="14" font-weight="bold" fill="#1E8449">0.820</text>
|
|
91
|
+
<text x="695" y="198" text-anchor="middle" font-size="12" fill="#27AE60" font-weight="bold">+28.1%</text>
|
|
92
|
+
<text x="695" y="445" text-anchor="middle" font-size="11" fill="#1E8449" font-weight="bold">Full</text>
|
|
93
|
+
<text x="695" y="458" text-anchor="middle" font-size="11" fill="#1E8449" font-weight="bold">CaaS</text>
|
|
94
|
+
|
|
95
|
+
<!-- Star indicator for full -->
|
|
96
|
+
<text x="695" y="183" text-anchor="middle" font-size="16">⭐</text>
|
|
97
|
+
|
|
98
|
+
<!-- Legend / Key findings box -->
|
|
99
|
+
<rect x="50" y="80" width="120" height="120" rx="6" fill="white" stroke="#DEE2E6" stroke-width="1"/>
|
|
100
|
+
<text x="110" y="100" text-anchor="middle" font-size="11" font-weight="bold" fill="#2C3E50">Key Findings</text>
|
|
101
|
+
<line x1="60" y1="108" x2="160" y2="108" stroke="#E0E0E0" stroke-width="1"/>
|
|
102
|
+
<rect x="60" y="118" width="12" height="12" fill="#3498DB" rx="2"/>
|
|
103
|
+
<text x="78" y="128" font-size="9" fill="#495057">Individual</text>
|
|
104
|
+
<rect x="60" y="138" width="12" height="12" fill="#27AE60" rx="2"/>
|
|
105
|
+
<text x="78" y="148" font-size="9" fill="#495057">Combined</text>
|
|
106
|
+
<line x1="60" y1="162" x2="160" y2="162" stroke="#E74C3C" stroke-width="2" stroke-dasharray="4,2"/>
|
|
107
|
+
<text x="78" y="172" font-size="9" fill="#495057">Baseline</text>
|
|
108
|
+
<text x="110" y="192" text-anchor="middle" font-size="8" fill="#6C757D" font-style="italic">p < 0.001</text>
|
|
109
|
+
|
|
110
|
+
<!-- Statistical note -->
|
|
111
|
+
<rect x="200" y="470" width="400" height="25" rx="4" fill="#E8F8F5" stroke="#1ABC9C" stroke-width="1"/>
|
|
112
|
+
<text x="400" y="487" text-anchor="middle" font-size="11" fill="#1ABC9C" font-weight="bold">Statistical significance: p < 0.001 | Cohen's d = 3.36 (large effect)</text>
|
|
113
|
+
</svg>
|
|
Binary file
|
|
@@ -0,0 +1,97 @@
|
|
|
1
|
+
<?xml version="1.0" encoding="UTF-8"?>
|
|
2
|
+
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 700 400" font-family="Arial, sans-serif">
|
|
3
|
+
<defs>
|
|
4
|
+
<linearGradient id="heuristicGrad" x1="0%" y1="0%" x2="0%" y2="100%">
|
|
5
|
+
<stop offset="0%" style="stop-color:#27AE60;stop-opacity:1" />
|
|
6
|
+
<stop offset="100%" style="stop-color:#1E8449;stop-opacity:1" />
|
|
7
|
+
</linearGradient>
|
|
8
|
+
<linearGradient id="mlGrad" x1="0%" y1="0%" x2="0%" y2="100%">
|
|
9
|
+
<stop offset="0%" style="stop-color:#F39C12;stop-opacity:1" />
|
|
10
|
+
<stop offset="100%" style="stop-color:#E67E22;stop-opacity:1" />
|
|
11
|
+
</linearGradient>
|
|
12
|
+
<linearGradient id="llmGrad" x1="0%" y1="0%" x2="0%" y2="100%">
|
|
13
|
+
<stop offset="0%" style="stop-color:#E74C3C;stop-opacity:1" />
|
|
14
|
+
<stop offset="100%" style="stop-color:#C0392B;stop-opacity:1" />
|
|
15
|
+
</linearGradient>
|
|
16
|
+
<filter id="barShadow2" x="-5%" y="-5%" width="110%" height="120%">
|
|
17
|
+
<feDropShadow dx="2" dy="2" stdDeviation="3" flood-opacity="0.2"/>
|
|
18
|
+
</filter>
|
|
19
|
+
</defs>
|
|
20
|
+
|
|
21
|
+
<!-- Title -->
|
|
22
|
+
<text x="350" y="35" text-anchor="middle" font-size="20" font-weight="bold" fill="#2C3E50">Router Latency Comparison (Log Scale)</text>
|
|
23
|
+
<text x="350" y="55" text-anchor="middle" font-size="12" fill="#6C757D">CaaS Heuristic Router vs. ML-based and LLM-based alternatives</text>
|
|
24
|
+
|
|
25
|
+
<!-- Chart background -->
|
|
26
|
+
<rect x="120" y="75" width="480" height="250" fill="#FAFAFA" stroke="#E0E0E0" stroke-width="1"/>
|
|
27
|
+
|
|
28
|
+
<!-- Y-axis (log scale visual) -->
|
|
29
|
+
<line x1="120" y1="75" x2="120" y2="325" stroke="#495057" stroke-width="2"/>
|
|
30
|
+
<text x="60" y="205" text-anchor="middle" font-size="13" font-weight="bold" fill="#2C3E50" transform="rotate(-90, 60, 205)">Latency (ms)</text>
|
|
31
|
+
|
|
32
|
+
<!-- Y-axis labels (log scale) -->
|
|
33
|
+
<text x="110" y="320" text-anchor="end" font-size="10" fill="#6C757D">0.001</text>
|
|
34
|
+
<line x1="115" y1="316" x2="600" y2="316" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
35
|
+
|
|
36
|
+
<text x="110" y="280" text-anchor="end" font-size="10" fill="#6C757D">0.01</text>
|
|
37
|
+
<line x1="115" y1="276" x2="600" y2="276" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
38
|
+
|
|
39
|
+
<text x="110" y="240" text-anchor="end" font-size="10" fill="#6C757D">0.1</text>
|
|
40
|
+
<line x1="115" y1="236" x2="600" y2="236" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
41
|
+
|
|
42
|
+
<text x="110" y="200" text-anchor="end" font-size="10" fill="#6C757D">1</text>
|
|
43
|
+
<line x1="115" y1="196" x2="600" y2="196" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
44
|
+
|
|
45
|
+
<text x="110" y="160" text-anchor="end" font-size="10" fill="#6C757D">10</text>
|
|
46
|
+
<line x1="115" y1="156" x2="600" y2="156" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
47
|
+
|
|
48
|
+
<text x="110" y="120" text-anchor="end" font-size="10" fill="#6C757D">100</text>
|
|
49
|
+
<line x1="115" y1="116" x2="600" y2="116" stroke="#E0E0E0" stroke-width="1" stroke-dasharray="4,4"/>
|
|
50
|
+
|
|
51
|
+
<text x="110" y="85" text-anchor="end" font-size="10" fill="#6C757D">1000</text>
|
|
52
|
+
|
|
53
|
+
<!-- X-axis -->
|
|
54
|
+
<line x1="120" y1="325" x2="600" y2="325" stroke="#495057" stroke-width="2"/>
|
|
55
|
+
|
|
56
|
+
<!-- Heuristic Router Bar (0.003ms) - log position ~300 -->
|
|
57
|
+
<rect x="160" y="298" width="100" height="27" fill="url(#heuristicGrad)" filter="url(#barShadow2)" rx="4"/>
|
|
58
|
+
<text x="210" y="287" text-anchor="middle" font-size="14" font-weight="bold" fill="#1E8449">0.003ms</text>
|
|
59
|
+
<text x="210" y="345" text-anchor="middle" font-size="12" fill="#495057" font-weight="bold">Heuristic</text>
|
|
60
|
+
<text x="210" y="360" text-anchor="middle" font-size="11" fill="#27AE60">(CaaS)</text>
|
|
61
|
+
|
|
62
|
+
<!-- ML Router Bar (15.2ms) - log position ~148 -->
|
|
63
|
+
<rect x="310" y="148" width="100" height="177" fill="url(#mlGrad)" filter="url(#barShadow2)" rx="4"/>
|
|
64
|
+
<text x="360" y="138" text-anchor="middle" font-size="14" font-weight="bold" fill="#E67E22">15.2ms</text>
|
|
65
|
+
<text x="360" y="345" text-anchor="middle" font-size="12" fill="#495057" font-weight="bold">ML-based</text>
|
|
66
|
+
<text x="360" y="360" text-anchor="middle" font-size="11" fill="#F39C12">(Classifier)</text>
|
|
67
|
+
|
|
68
|
+
<!-- LLM Router Bar (450ms) - log position ~93 -->
|
|
69
|
+
<rect x="460" y="93" width="100" height="232" fill="url(#llmGrad)" filter="url(#barShadow2)" rx="4"/>
|
|
70
|
+
<text x="510" y="83" text-anchor="middle" font-size="14" font-weight="bold" fill="#C0392B">450ms</text>
|
|
71
|
+
<text x="510" y="345" text-anchor="middle" font-size="12" fill="#495057" font-weight="bold">LLM-based</text>
|
|
72
|
+
<text x="510" y="360" text-anchor="middle" font-size="11" fill="#E74C3C">(GPT-3.5)</text>
|
|
73
|
+
|
|
74
|
+
<!-- Speedup annotations -->
|
|
75
|
+
<path d="M260 290 Q 285 250 310 200" stroke="#2C3E50" stroke-width="1.5" fill="none" stroke-dasharray="4,2"/>
|
|
76
|
+
<text x="295" y="235" text-anchor="middle" font-size="11" font-weight="bold" fill="#2C3E50">5,000×</text>
|
|
77
|
+
<text x="295" y="248" text-anchor="middle" font-size="9" fill="#6C757D">faster</text>
|
|
78
|
+
|
|
79
|
+
<path d="M260 295 Q 360 230 460 160" stroke="#2C3E50" stroke-width="1.5" fill="none" stroke-dasharray="4,2"/>
|
|
80
|
+
<text x="380" y="200" text-anchor="middle" font-size="11" font-weight="bold" fill="#2C3E50">150,000×</text>
|
|
81
|
+
<text x="380" y="213" text-anchor="middle" font-size="9" fill="#6C757D">faster</text>
|
|
82
|
+
|
|
83
|
+
<!-- Key insight box -->
|
|
84
|
+
<rect x="610" y="100" width="80" height="140" rx="6" fill="#E8F8F5" stroke="#27AE60" stroke-width="2"/>
|
|
85
|
+
<text x="650" y="125" text-anchor="middle" font-size="10" font-weight="bold" fill="#27AE60">Why?</text>
|
|
86
|
+
<line x1="620" y1="132" x2="680" y2="132" stroke="#27AE60" stroke-width="1"/>
|
|
87
|
+
<text x="650" y="150" text-anchor="middle" font-size="8" fill="#495057">No ML</text>
|
|
88
|
+
<text x="650" y="163" text-anchor="middle" font-size="8" fill="#495057">inference</text>
|
|
89
|
+
<text x="650" y="180" text-anchor="middle" font-size="8" fill="#495057">No API</text>
|
|
90
|
+
<text x="650" y="193" text-anchor="middle" font-size="8" fill="#495057">calls</text>
|
|
91
|
+
<text x="650" y="210" text-anchor="middle" font-size="8" fill="#495057">Pure</text>
|
|
92
|
+
<text x="650" y="223" text-anchor="middle" font-size="8" fill="#495057">heuristics</text>
|
|
93
|
+
|
|
94
|
+
<!-- Bottom note -->
|
|
95
|
+
<rect x="150" y="375" width="400" height="20" rx="4" fill="#FEF9E7" stroke="#F1C40F" stroke-width="1"/>
|
|
96
|
+
<text x="350" y="389" text-anchor="middle" font-size="11" fill="#B7950B" font-weight="bold">"Speed > Smarts" — deterministic routing enables real-time RAG</text>
|
|
97
|
+
</svg>
|
|
@@ -0,0 +1,103 @@
|
|
|
1
|
+
# Introduction
|
|
2
|
+
|
|
3
|
+
## The RAG Revolution and Its Hidden Pitfalls
|
|
4
|
+
|
|
5
|
+
Retrieval-Augmented Generation (RAG) has emerged as the dominant paradigm for grounding Large Language Model (LLM) outputs in factual, domain-specific knowledge [Lewis et al., 2020]. By retrieving relevant documents at inference time, RAG systems overcome the knowledge cutoff limitations of pre-trained models while enabling deployment in enterprise settings where proprietary data must remain private.
|
|
6
|
+
|
|
7
|
+
Yet beneath the surface of this revolution lies a troubling reality: **most production RAG systems are fundamentally broken**. Not in obvious ways that cause immediate failures, but in subtle architectural choices that degrade quality, waste resources, and—most critically—erode user trust over time.
|
|
8
|
+
|
|
9
|
+
## The Seven Fallacies of Production RAG
|
|
10
|
+
|
|
11
|
+
Through extensive deployment experience and analysis of existing frameworks, we identify **seven critical fallacies** that plague production RAG systems:
|
|
12
|
+
|
|
13
|
+
### 1. The Flat Chunk Fallacy (Structure Problem)
|
|
14
|
+
|
|
15
|
+
The standard approach—split documents into fixed-size chunks (e.g., 500 tokens), embed them, and retrieve by vector similarity—treats all content as equally valuable. But a class definition in source code carries fundamentally different weight than a TODO comment. A contract's liability clause matters more than its formatting instructions. By flattening this hierarchy, we lose the structural signals that humans naturally use to prioritize information.
|
|
16
|
+
|
|
17
|
+
**The Reality**: Structure encodes importance, and ignoring it is negligent.
|
|
18
|
+
|
|
19
|
+
### 2. Context Amnesia (Metadata Problem)
|
|
20
|
+
|
|
21
|
+
When a chunk is extracted from its parent document, it loses critical context. Consider the retrieved text: *"It increased by 5%."* What increased? Revenue? Costs? Infection rates? Without metadata preserving the document path (`Q3 Earnings → Revenue → North America`), the chunk is semantically orphaned.
|
|
22
|
+
|
|
23
|
+
**The Reality**: A chunk without metadata is a fact without meaning.
|
|
24
|
+
|
|
25
|
+
### 3. Time-Blind Retrieval (Temporal Problem)
|
|
26
|
+
|
|
27
|
+
Traditional retrieval optimizes for semantic similarity: "How to restart the server" in 2021 matches "How to restart the server" in 2025 with near-perfect similarity. But the 2021 procedure may be dangerously outdated—the architecture changed, the commands differ, the old approach may even cause data loss.
|
|
28
|
+
|
|
29
|
+
**The Reality**: In software, truth has a half-life. Retrieval must decay with time.
|
|
30
|
+
|
|
31
|
+
### 4. The Flat Context Fallacy (Priority Problem)
|
|
32
|
+
|
|
33
|
+
Most systems stuff context into the LLM's window until it's full, treating the user's last message, their stated preferences, and historical archives from two years ago with equal priority. But the question asked 30 seconds ago is categorically more important than background context from last month. This leads to what we term the **Accumulation Paradox**: more context can paradoxically *degrade* performance, as demonstrated empirically by Liu et al. [2023] who showed that LLMs systematically ignore information in the middle of long contexts.
|
|
34
|
+
|
|
35
|
+
**The Reality**: Context has intimacy levels. Proximity in conversation ≠ proximity in relevance.
|
|
36
|
+
|
|
37
|
+
### 5. The Official Truth Fallacy (Source Problem)
|
|
38
|
+
|
|
39
|
+
Enterprise documentation often contains aspirational or theoretical information: "The API supports 100 concurrent requests." But the engineering Slack channel contains the practical truth: "We crash around 50; the docs are lying." Traditional RAG surfaces only the official answer, leading users astray.
|
|
40
|
+
|
|
41
|
+
**The Reality**: Official ≠ Accurate. Systems must surface the *pragmatic truth*.
|
|
42
|
+
|
|
43
|
+
### 6. The Brutal Squeeze (Context Management Problem)
|
|
44
|
+
|
|
45
|
+
When conversation history exceeds context limits, the common solution is AI-powered summarization: "Let's ask an LLM to compress the history." But this is a trap:
|
|
46
|
+
- Summarization costs tokens (money) to generate
|
|
47
|
+
- Summaries lose nuance: "I tried X and it failed with error E-1234" becomes "User attempted troubleshooting"
|
|
48
|
+
- The specific error code—exactly what the user needs—is lost
|
|
49
|
+
|
|
50
|
+
**The Reality**: Lossless truncation (FIFO sliding window) beats lossy compression (summarization).
|
|
51
|
+
|
|
52
|
+
### 7. The Middleware Gap (Trust Problem)
|
|
53
|
+
|
|
54
|
+
Startups offer "intelligent routing" services: send your queries through their API, and they'll route to the cheapest model. But no enterprise CISO will send proprietary data through a random middleware startup to save 30% on tokens.
|
|
55
|
+
|
|
56
|
+
**The Reality**: The winner in enterprise AI won't have the smartest router—they'll have the one enterprises trust with the keys.
|
|
57
|
+
|
|
58
|
+
## Our Contribution: Context-as-a-Service
|
|
59
|
+
|
|
60
|
+
We present **Context-as-a-Service (CaaS)**, an open-source framework that systematically addresses all seven fallacies through a principled architecture:
|
|
61
|
+
|
|
62
|
+
1. **Structure-Aware Indexing**: A three-tier value hierarchy (High/Medium/Low) that respects document structure instead of flat chunking.
|
|
63
|
+
|
|
64
|
+
2. **Metadata Injection**: Automatic enrichment of chunks with their document path, section hierarchy, and temporal metadata.
|
|
65
|
+
|
|
66
|
+
3. **Time-Based Decay**: Exponential decay functions that prioritize recent content, with configurable half-life parameters per domain.
|
|
67
|
+
|
|
68
|
+
4. **Context Triad**: A Hot/Warm/Cold classification that treats context by intimacy:
|
|
69
|
+
- **Hot**: Current conversation (highest priority)
|
|
70
|
+
- **Warm**: User preferences, recent documents
|
|
71
|
+
- **Cold**: Historical archives, background knowledge
|
|
72
|
+
|
|
73
|
+
5. **Pragmatic Truth**: Parallel tracking of official documentation and informal sources (Slack, tickets, notes) with transparent conflict surfacing.
|
|
74
|
+
|
|
75
|
+
6. **Sliding Window Management**: FIFO truncation that preserves recent turns perfectly intact rather than summarizing everything poorly.
|
|
76
|
+
|
|
77
|
+
7. **Trust Gateway**: An on-premises/private-cloud router that enterprises can deploy behind their firewall with zero data leakage.
|
|
78
|
+
|
|
79
|
+
## Evaluation and Results
|
|
80
|
+
|
|
81
|
+
We evaluate CaaS on a new benchmark corpus of 16 enterprise documents spanning code, legal, HR, and engineering domains, released publicly on Hugging Face for reproducibility.
|
|
82
|
+
|
|
83
|
+
Key results:
|
|
84
|
+
- **+28.1% Precision@5** over flat-chunk baselines
|
|
85
|
+
- **+27.9% NDCG@10** improvement in ranking quality
|
|
86
|
+
- **0.003ms** routing latency (sub-millisecond)
|
|
87
|
+
- **3.36 Cohen's d** effect size (statistically significant, p < 0.001)
|
|
88
|
+
|
|
89
|
+
## Paper Organization
|
|
90
|
+
|
|
91
|
+
The remainder of this paper is organized as follows:
|
|
92
|
+
- **Section 2** reviews related work in RAG, context management, and enterprise AI deployment
|
|
93
|
+
- **Section 3** details the CaaS architecture and algorithms
|
|
94
|
+
- **Section 4** presents our experimental methodology and benchmark corpus
|
|
95
|
+
- **Section 5** reports results and ablation studies
|
|
96
|
+
- **Section 6** discusses limitations and future work
|
|
97
|
+
- **Section 7** concludes
|
|
98
|
+
|
|
99
|
+
---
|
|
100
|
+
|
|
101
|
+
## References (for this section)
|
|
102
|
+
|
|
103
|
+
- Lewis, P., et al. (2020). "Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks." *NeurIPS 2020*.
|
|
Binary file
|
|
Binary file
|
|
Binary file
|
|
Binary file
|