* Get the current agent state as JSON. * @returns {any}

* Get the todo list as JSON. * @returns {any}

* Get the list of available tools. * @returns {any}

* Group workers by their type for prompt generation

* Computes Hamming distance between two binary vectors. * * @param a - First binary vector * @param b - Second binary vector * @returns Ha

Public test seam — handles a single inbound message.

* Handle HTTP request

* Handle HTTP request

* Route task to optimal agent

* Route task to optimal agent

* Check if agent has a specific capability

* Check if there's a direct edge from source to target

* Check if a scope has permission for an action. * * Uses exact match and pattern matching (with wildcards).

* Check if a plugin is registered

* FNV-1a hash for cache keys

* FNV-1a hash for cache keys

* Generate deterministic hash-based embedding

* Hash function for cache keys (FNV-1a)

* Generate cache key from text

* Pure hash function for string ID (no map lookup). * Used by registerIdMapping for initial hash before collision probing.

* Compute a composite hash from an array of trace hashes.

* Health check for WASM MCP server (runs locally)

* Get hierarchy depth from hyperbolic point

* Build a hook command that resolves to the right helpers dir on every * install layout. `ruflo init` can land helpers either project-locally * (`<p

* Hyperbolic Attention for hierarchical data

* Identify confounders (common causes of treatment and outcome)

* Identify required capabilities for a task

* Create an identity morphism for a type

* Override importEvents to also persist imported events to storage.

* Import patterns (for migration or testing)

* Import state (does not load collections, just restores overrides).

Add an event to the in-memory indexes.

* Index a code file by chunking and embedding.

* Infer an agent type string from a SONA pattern result object.

* Infer agent type from task description

* Infer agent type from task description

* Infer domain from task

Ingest raw vectors into a device's vector store.

* Lazily attempt to load and initialize the NeuralLearningSystem. * The promise is cached so that repeated calls do not re-attempt * after a fai

* Initialize the WASM MCP server

* Initialize the integration bridge * * This method is idempotent - calling it multiple times is safe. * Components are lazily loaded based o

* Initialize the agent * * ADR-001: When agentic-flow is available, delegates initialization * to agentic-flow's Agent.initialize() for optim

* Initialize the queue (starts cleanup timers)

* Initialize store and load registry

* Initialize a new Codex project

* Initialize the router with WASM support

* Initialize the optimizer with WASM support

* Initialize optimizer for a team

* Initialize weights with Kaiming/He initialization

* Install a plugin from a local path