Assess urgency of log entry.

Assess urgency of audio input.

Convert one ATR detection condition into an AGT rule entry.

Attach OTLP span exporter to the tracer provider (best-effort).

Log data analysis to audit chain.

Auto-resolve a timed-out dispute.

Background loop for automatic scaling

Get from backend through circuit breaker with tracing.

Set in backend through circuit breaker with tracing.

Return the platform-agnostic basename without extension.

Bootstrap OTEL if configured.

Broadcast current data to all clients.

Broadcast an event to all connected agents.

Broadcast loop for slash events.

Build a standardized claim object from extracted data.

Build a concise comment body with risk levels only.

Build action_id → compensator mapping from STEPS.

Build policy evaluation context from function arguments.

Build detailed drift explanation.

Best-effort PolicyEvaluator construction — returns ``None`` when ``agent-os-kernel`` isn't installed (the docker provider does the sam

Generate synthetic history (12 hours at 15-min intervals).

Bind the bridge policy at the intervention points that match the v4 GovernancePolicy fields. pre_tool_call covers allowed_tools / blocked_pat

Build a standardized observation object.

Build the standalone argparse parser for this example.

Load the research-agent policy from its YAML file.

Build the canonical policy input per SPECIFICATION.md §7.

Build a resource ID from the request

Construct the upstream Sandbox. We try the full kwarg form first; if upstream rejects an unknown keyword, fall back to the minimal po

Section 1: General description of the AI system.

Section 2: Detailed description of development process.

Section 3: Monitoring, functioning, and control.

Section 4: Risk management system.

Section 5: Accuracy, robustness, and cybersecurity.

Build a detailed inspection dict for a session.

Build a summary dict for a single session.

Build hierarchical span tree

Build a chronological timeline from protocol spans.

Return an AGT manifest ``tools`` section from ``allowed_tools``. v4 semantics: an empty allowed_tools list means "no allowlist". Returning ``

Simulate trust score fluctuations over the last *hours*.

Return ``(allowed, reason)``. The CaaS app refuses to start unless either (a) AGENT_OS_ENV is local/dev/development, or (b) the operator has e

Cache a verification result (key includes DID + public key for security).

Cache a verification result.

Cache a verification result with timestamp.

Calculate how anomalous an action is compared to baseline

Calculate average weights across all documents.

Calculate behavior component score.

Calculate compliance component score.

Calculate confidence score for the observation. Factors: 1. Signal clarity: Do we have all required fields?

Calculate content density (information per character).

Calculate detailed breakdown for audit trail.

Calculate confidence score for embedding verification.

Calculate average action frequency

Calculate identity component score.

Calculate an importance score for a section (0-1).

Calculate density of important keywords.

Calculate network component score.

Calculate overall contract risk level.

Calculate drift between all pairs of responses.

Calculate total penalty from violations.

Calculate objective score based on job requirements.

Calculate weight for a specific section using tier-based approach.

Calculate cosine similarity between two vectors. Simplified implementation.

Calculate suspicion score based on heuristics

Calculate the time span covered by episodes in days.

Calculate trust score for a peer.