Return the source line of a key inside the *idx*-th rule. Falls back to the rule's own line if the key is not in the AST (e.g., when

Evaluate trust for *agent_id*. Returns ``trusted`` (bool), ``score`` (float), and ``action`` ("pass", "review", or "block").

Execute the agent with trust verification. Args: query: The input query. handler: Callable that processes the query.

Execute the evidence collection pipeline. Returns: An ``EvidenceReport`` containing the Annex IV document, evidence s

Execute the benchmark and return (overall_results, category_results).

Inject a fault, run traffic, then reset.

Run uvicorn with env-configurable host/port.

Run every golden trace in *suite* and aggregate results. Args: suite: The golden-trace suite to execute. agent_fn: Ca

Run a supervision cycle to check for violations. Returns violations detected by all supervisors.

Save policy to a YAML file.

Save a trace to storage.

* Save agent to disk.

Scan existing memory entries for poisoning indicators. Checks both content patterns and hash integrity for every entry.

Compute loop score for a conversation.

Return a z-score for the current window's call frequency. Returns 0.0 when insufficient data is available.

Select the best agent based on strategy. Args: bids: List of bids from agents selection_strategy: St

Send a message over the gRPC channel. Serialises the payload as JSON and transmits it as a unary call. Args: topic: RPC

Register an async callback for DEFER policy decisions. Args: callback: Async callable receiving (action, params) and returning

Mark the span as failed.

Set resource quota for an agent

Set the span status.

Store a session variable (e.g. AGT trust score).

Return SHA-256 hex digest.

Check if score should allow actions.

Check if an agent should be demoted based on trust drop.

Determine if requests to this agent should be quarantined.

Check if score should trigger revocation.

Sign this AI Card with an AgentIdentity. Args: identity: An ``AgentIdentity`` instance with a private key.

Simulate agent execution with different performance characteristics. Args: version: "old" or "new" scenario: "normal", "

Store a message. Returns False if duplicate (by message_id).

Store content in cache. Args: context: The context to cache response: Optional response to cache wit

Store agent identity data. Args: agent_did: The agent's DID identifier. identity_data: Identity information dict.

Store a semantic rule to the rules file. Args: rule: The semantic rule to store Returns:

Subscribe a handler to an event type.

Check if terminal supports colors.

Increment the clock for an agent.

Get time remaining until expiration. Returns: A non-negative timedelta; zero if already expired.

Seconds until the requested number of tokens are available.

Export as a JWK Set.

Convert undefended vectors into ComplianceViolation-compatible dicts. Only produces violations for vectors that are not defended. Ar

Export as CycloneDX 1.5 JSON.

Serialize to dictionary for API responses or storage.

Convert to dictionary.

Serialize the finding to a JSON-compatible dict.

Serialize this verdict to a dictionary for audit logging. Returns: Dict with all verdict fields suitable for JSON serialization.

Convert to dictionary.

Convert to dictionary.

Export as HTTP headers for injection.

Return the v4 ``(allowed, reason)`` tuple.

Convert to OpenAI function calling format. Returns: Dictionary in OpenAI function calling format

Serialise as W3C traceparent header value.

Build a password-free Redis URL from host/port/db. The password is intentionally NOT embedded in the URL. Anything that reads back th

Convert an AMB topic to a CloudEvents type. CloudEvents recommends reverse-DNS naming for types. Args: topic: AMB topic

Context manager for tracing approval workflow spans.

Start an execution trace.

Start a policy check trace.

Context manager for tracing trust verification spans.

Transition to a new state. Raises ValueError if the transition is not allowed.

Remove a backend from the registry. Args: name: Backend identifier to remove.

Remove the governance wrapper and return the original agent. Args: governed_agent: A previously wrapped agent or any object.

Get original crew from wrapped version.

Update instructions with new version and log the improvement. Args: new_instructions: New system instructions