Source Module Reference

Every production module has an English module-level docstring and an SPDX MIT header. The table below explains the responsibility boundary.

Module	Responsibility
`acceptance.py`	Real-provider acceptance harness, matrix orchestration, safe retry, consolidated certification
`acceptance_logging.py`	Trace normalization and portable acceptance evidence
`acceptance_policy.py`	Acceptance-only environment/provider safety policy
`agent_config.py`	Provider-neutral agent configuration generation and validation
`agents.py`	Managed agent rules, skills, instructions, provider integration helpers
`application.py`	Provider-neutral application API and graphical-client read/write contracts
`app_server.py`	Loopback-only HTTP/JSON adapter and Local Studio static server
`artifacts.py`	Constitution, baseline, specification, plan and Markdown artifact management
`baselines.py`	Accepted reference efficiency baselines
`benchmark.py`	Offline acceptance comparison helpers
`cli.py`	`dynosai` command-line interface
`context_control.py`	Context pressure and context-strategy selection
`cost_telemetry.py`	Usage/cost telemetry and cost-per-successful-governed-change aggregation
`context_handles.py`	Persistent pass-by-reference handles for large artifacts
`eval_registry.py`	Versioned eval scenarios and offline harness comparisons
`eval_intelligence.py`	Failure attribution, local-trace mining, acceptance ZIP import and inbox-only improvement loop
`prompt_prefix.py`	Deterministic MCP authority prefix hash; cacheable structure, not a cache-hit claim
`execution_profiles.py`	Strict/Balanced/Autonomous host policy, local runtime bind and unsupported-runtime refusal
`capability_manifests.py`	Cursor ACP / Codex app-server contracts and refusal of uncertified clients/packs
`certification_matrix.py`	MATRIX_1.0 live Codex/Cursor × greenfield/brownfield cells; historical 0.13 evidence stays separate
`execution_runtime.py`	Provider-neutral ExecutionRuntime and local hands implementation
`harness_contracts.py`	Typed execution/recovery/handle contracts and optional harness feature flags
`secrets.py`	Credential redaction and secret broker (model refuse-closed; optional local runtime vault)
`validation_integrity.py`	Requirement → evidence → validation integrity reports
`db.py`	SQLite schema, migrations, persistence and audit storage
`debug.py`	Diagnostic and end-to-end debug orchestration
`debug_fixtures.py`	Deterministic greenfield/brownfield test fixtures
`engine.py`	Main Core facade and workflow coordination
`estimator_calibration.py`	Token/context estimator calibration
`events.py`	Event primitives
`git_guard.py`	Managed-session Git restrictions
`git_manager.py`	Governed Git read/write operations and evidence
`indexer.py`	Files/symbols/tests/code relationship indexing
`interactions.py`	Human gate and elicitation lifecycle
`managed_runtime.py`	Strict provider runtime isolation
`mcp.py`	MCP server, tool contracts and provider transport adapters
`model_benchmark.py`	Offline provider/model benchmark ranking
`model_control.py`	Phase budgets, adaptive control decisions and transition evidence
`model_routing.py`	Provider/activity routing profiles
`policy.py`	Path/scope, validation-profile, network and dependency policies
`predictive_routing.py`	Historical predictive routing memory (shadow use)
`predictive_validation.py`	Offline replay and authority-gate analysis
`provider_debug.py`	Provider-native debug/certification adapters
`provider_process.py`	Provider subprocess spawn/shutdown and Windows process-tree cleanup
`providers.py`	Provider capability detection and command construction
`retrieval.py`	Hybrid bounded retrieval of text, symbols, tasks and evidence
`runtime.py`	Machine-local project/session broker
`runtime_audit.py`	Runtime/provider isolation audit
`semantic.py`	Local embeddings and semantic index
`state.py`	Portable snapshots and atomic restore
`team_scheduler.py`	Plan-DAG waves, file-disjoint leases, claim rules and wave-scoped fan-in
`studio_projects.py`	Local Studio project hub registry and directory browsing
`token_usage.py`	Token normalization, phase attribution and usage reporting
`util.py`	Shared low-level helpers
`validation.py`	Workflow quality scoring and validation evaluation
`validation_discovery.py`	Stack-aware, read-only discovery and explicit approval of validation profiles
`risk.py`	Deterministic advisory project/work risk assessment
`version.py`	Package version