* Log message

Bulk insert entries

Classify a single action.

* Export the current graph state as a SparseMatrix. * * `since` is an optional snapshot ISO timestamp — adapters may return * a frozen snapsh

* Supersede an existing assertion with a new one. * * Marks the old assertion as superseded and creates a new assertion that * declares it re

* Abort all active operations.

* Distill a trajectory into a reusable memory * * @param trajectory - Trajectory to distill * @returns Distilled memory or null if quality to

* Evaluate whether a write operation should be allowed. * * Steps: * 1. Check authority (namespace allowed, role sufficient) * 2. Check ra

* Execute the tool. Returns a plain string (success) or throws (will be * caught by the agent loop and wrapped in an `is_error: true` result).

* Notify the bridge that an insight has been recorded in AgentDB. * Creates a learning trajectory so the neural system can track the * insight's

Create a runner from an already-loaded buffer.

* Add a section to the appliance image. * * @param id Section identifier (e.g. 'kernel', 'runtime', 'ruflo'). * @param data Raw (unco

* Escape SQL identifier.

* Retrieve a belief by its ID. * * @param id - The belief ID * @returns The belief, or undefined if not found

* Get memory statistics

* Initialize memory namespaces

* Manually set an agent's trust score (e.g., from persistence restore). * Clamps to [0, 1].

* Set verbosity level * - quiet: Only errors and direct results * - normal: Errors, warnings, info, and results * - verbose: All of normal +

Check if request is allowed

* Check if operation is allowed under rate limit

* Export the chain as a serializable object.