Get a cached embedding. Args: text: The text to look up Returns: The cached embedding or None

Transition to a new state. Args: new_state: The state to transition to reason: Optional reason for the trans

Emit a notification. If UI callback registered, call it; otherwise log.

Check whether `candidate` is covered by any entry in `allowed`. Allowed entries may be IPs, CIDRs, or hostnames/labels. Candidate may also be

Update status bar

Execute the tool with given arguments. Args: arguments: The arguments for the tool runtime: The runtime envi

Serialize messages to disk and update the index. If conv_id is given and already exists in the index, the file is overwritten and the

Stop the runtime environment.

Create a Tool instance from an MCP tool definition. Args: tool_def: The MCP tool definition server: The MCP server that prov

Extract candidate target strings from arguments (shallow, non-recursive). This function inspects only the top-level of the provided object (

Run the crew on a task.

Compose and update the persistent header widget.

Clear all documents and embeddings.

Generate a response from the LLM. Args: system_prompt: The system prompt messages: Conversation messages

Register a pre-built tool instance (used for MCP tools). Args: tool: The Tool instance to register

Start the local runtime.

Stop the local runtime gracefully.

Return True for models that route through an OpenAI-compatible endpoint.

Execute a command in the sandbox. Args: command: The shell command to execute timeout: Timeout in seconds

Return True when RAG should use local sentence-transformer embeddings. Decision priority: 1. Explicit ``PENTESTAGENT_EMBEDDINGS=local`` → Tru

Validate note schema based on declarative rules. Returns: Error message if validation fails, None if valid

Format an MCP tool result for display. Args: result: The raw MCP result Returns: Formatted string

Get a tool by name.

Validate arguments against the schema. Args: arguments: The arguments to validate Returns: Tuple of

Wait for specified agents (or all) to complete. Args: agent_ids: List of agent IDs to wait for. None = wait for all.

Add a node if it doesn't exist.

Emit event to callback if registered.

Build a mock LLM that returns pre-set responses. Each response is a dict with optional: - content: str - tool_calls: list of {id

Index a single file. Args: file_path: The file to index Returns: List of Document objects

Format messages as text for summarization.

Detect the current system environment.

Get all registered tools.

Get messages that fit within token limit (sync, no summarization). Falls back to truncation if over budget. Args:

Find relevant documents for a query. Args: query: The search query k: Maximum number of results to return

Add an edge.

Attach the configured OpenAI-compatible API base URL to LiteLLM kwargs. Only applied when: - ``config.openai_api_base`` is set (i.e.

Build a clean env dict with only the keys we care about.

Format conversation history for LLM.

Parse tool call arguments.

Update crew stats panel.

Get all notes (for TUI /notes command).

Get messages, summarizing older ones if needed. Args: messages: Full conversation history llm_call: Async fu

Return the configured OpenAI-compatible API base URL, or None.

Get total token count for messages.

Load and deserialize a saved conversation. Returns [] on error.

Append or set operator_notes for a workspace (plain text).

Deliver a UI event if a hook is registered, swallowing any errors.

Persist the current conversation history to the active loot store. On first call for a session, creates a new file and stores the id in

Update a worker's state.

Execute a command locally.

Generate embeddings for a list of texts using LiteLLM. Args: texts: List of texts to embed model: The embedding model to use

Generate embeddings locally using sentence-transformers. Args: texts: List of texts to embed model: The sentence-transformer

Get the global settings instance.

Resolve knowledge-related paths, preferring active workspace if present. Returns a dict with keys: base, sources, embeddings, graph, index, using

Convert to LLM function calling format.

Add a message widget to chat

Display tool execution result

Hide the sidebar.

Write an SGR mouse escape sequence to the PTY. Format: ESC [ < Cb ; Cx ; Cy M (press) ESC [ < Cb ; Cx ; Cy m (release)

Execute a tool with the given arguments. Args: tool: The tool to execute arguments: The arguments to pass to

Get memory statistics.