Return an async context manager that connects to the WS endpoint.

Disable a tool for a specific reason - uses enum, no magic strings!

Stream response from LLM client. Args: client: LLM client with streaming support messages: Conversation messages

Unwrap middleware ``ToolExecutionResult`` wrappers and MCP result dicts. When middleware is enabled, ``ToolManager`` returns the result wrapped i

Extract structuredContent from text blocks per MCP spec. The MCP spec says servers SHOULD include structuredContent serialised as JSO

Return a mock ToolManager.

Create a ServerInfo instance for testing.

Execute a tool via mcp-cli's ToolManager with guard checks. Args: request: Planner's execution request (tool_name, args, step_id)

Export messages to JSON format. Args: messages: Conversation messages metadata: Optional session metadata

Get the number of cached clients.

Attach the active WebSocket connection, closing any previous one.

Normalise a tool result into MCP CallToolResult structure. Returns ``{"content": [...], "structuredContent": {...}}`` matching the MC

Return a ToolStateManager scoped to the given *agent_id*. * ``"default"`` → delegates to the upstream global singleton * anything else → per-

Get the default model for a provider. NO HARDCODED MODELS - All come from configuration. Args: provider: Provid

Get a list of all available chat commands. Returns: List of command names with descriptions.

Store a back-reference to ChatContext for history/config queries.

Convert an object to a JSON-serializable form. Handles MCP SDK ``ToolResult``, Pydantic models, and other non-serializable types. MCP ``Tool

Build a mock RuntimeConfig.

Map a URL path to a file in the static directory. Returns None if not found.

Format tool arguments for inline preview. Shows first N arguments, truncated to reasonable length. Args: arguments: Tool arguments d

Get active model.

Convert size in bytes to human-readable string (KB/MB/GB). Args: size: Size in bytes (None if unavailable) Returns: Hum

Render a plan as an ASCII DAG for terminal display. Shows steps with their tools, dependencies, and execution status. Parallel steps (same ba

Extract value from a list of MCP content blocks.

Format tool response for conversation history.

Get a client for a custom OpenAI-compatible provider using chuk_llm. Args: provider: Provider name model: Mo

Build a mock ModelManager.

Return a ToolProcessor with sensible defaults.

Build a mock stream_manager with configurable behaviour.

Serialize conversation history for API, with cleanup. Replaces inline ``[msg.to_dict() for msg in ...]`` + validate pattern. Strips o

Set up logging and theme for individual commands.

Unwrap nested ToolResult/dict structures from MCP responses. MCP tools can return deeply nested result structures. This method extrac

Execute multiple tool calls in parallel with optional callbacks. Uses chuk-tool-processor's ToolCall/ToolResult models for consistency.

Finalize and validate all accumulated tool calls. Returns only valid, complete tool calls as dicts for API compatibility.

Get the currently active provider.

Get a specific custom provider configuration.

Convenience function to initialize the application context. Args: **kwargs: Arguments to pass to ApplicationContext.create() Re

Called after every tool execution completes.

Reset all counters.

Reset the configuration (useful for testing).

Set the global tool confirmation mode. Args: mode: One of 'always', 'never', or 'smart'

Compute parallel execution batches via topological sort. Groups steps into batches where all steps in a batch have their dependencies satisfi

Build a mock PreferenceManager with the given current mode.

Ensure every assistant tool_call_id has a matching tool result. Defense-in-depth: repairs orphaned tool_calls before sending to the API.

Execute the confirm command.

Get list of available models for a provider.

Execute a plan tool and return the result as a JSON string. Args: tool_name: One of plan_create, plan_execute, plan_create_and_execute.

Launch the interactive mode CLI with slash-menu autocompletion.

List all entries in a scope.

Push a tool result notification to the app.

Save session data to disk. Args: data: Session data to save Returns: Path to the saved file

Stop tool execution display and show result. Args: result: Tool execution result success: Whether execution succeeded

Validate a tool definition against OpenAI's function calling schema. SIMPLIFIED: Always attempt auto-fix instead of strict validation

Extract HTML content from a resources/read response.

Check if a ToolManager result is actually an error. ToolManager.execute_tool marks success=True when stream_manager.call_tool returns without

Merge two JSON strings intelligently. Tries multiple strategies: 1. Parse both and merge dicts 2. Concatenate and validate

Send agent list and replay focused agent state to a new client.

Check whether a tool result carries a ui_patch structuredContent.

Sanitize text for display by escaping control characters. Args: text: Raw text that may contain control characters Returns:

Simulate an LLM that streams responses.

Display the result of a tool call using chuk-term.

Execute an agent orchestration tool. Parameters ---------- tool_name: One of the ``_AGENT_TOOL_NAMES``. arguments: To

Launch the interactive chat loop with streaming support. Runtime uses adaptive policy: strict core with smooth wrapper. - Always enforce

Launch an MCP App in the browser. 1. Fetch the UI resource HTML from the MCP server 2. Start a local HTTP + WebSocket server

Load MCP server config file with fallback to bundled package config.

Get connection info string for display.

Get configured token manager instance with mcp-cli namespace.

Handle OAuth authentication flow for a server. Uses _oauth_lock to prevent concurrent flows from racing (e.g. multiple parallel tool

Load and adapt tools for the current provider. FIXED: Updated to use universal tool compatibility system.

Register a single command with Typer.

Build a mock LLM response containing a tool call. Uses the chuk_llm native format (top-level tool_calls key).

Truncate tool result if it exceeds max_chars. Keeps head + tail with a truncation notice in the middle. max_chars <= 0 disables trunc

Update cached reasoning preview with debouncing. Updates preview only if: - 1 second has passed since last update, OR - 10 ch

Create a prompt_toolkit Style for autocomplete menu that matches terminal background. Args: theme_colors: Theme ColorScheme object w

Disable a server in preferences.

Create a failed validation result.

Filter tools, separating valid from invalid ones. AGGRESSIVE: Always try to auto-fix first, then validate. Returns:

Format reasoning content for inline preview. Shows a clean excerpt of the reasoning with proper word boundaries. By default shows last N char

Get all command names. Args: mode: Optional mode filter. include_aliases: Whether to include command aliases

Get all custom providers.

Get the standard environment variable name for a provider. Args: provider_name: Provider name (e.g., 'openai', 'anthropic') Ret

Load saved user layouts from disk. Returns empty list on any error.