Result from command execution.

A message in an LLM conversation. Use Pydantic native methods: - model_dump(exclude_none=True, mode="json") to serialize to dict - Messag

Handle execution of tool calls returned by the LLM. Delegates parallel execution to ToolManager.stream_execute_tools(), handling only CL

Facade over chuk-tool-processor with mcp-cli specific features. Responsibilities: 1. Parse MCP config files and detect server types (HTT

Class to handle LLM conversation processing with streaming support. Updated to work with universal tool compatibility system. ENHANCED:

Mock context for testing.

Mock UI manager for testing.

Definition of a command parameter.

State container for plan operations. Centralizes access to the graph store, plan registry, and tool manager. Passed to the PlanRunner and Pla

Outcome of a tool execution. ENHANCED: Now wraps chuk_tool_processor.models.tool_result.ToolResult for full tracking (start_time, end_ti

Information about a tool. Lightweight adapter over chuk-tool-processor's ToolMetadata for MCP-CLI UI/display purposes.

Tool call made by the LLM.

Central registry for all commands. This replaces the separate registries in chat, CLI, and interactive modes.

Manages MCP CLI preferences with file persistence.

Configuration for a runtime provider (OpenAI-compatible API). This model represents a provider that can be added at runtime or loaded fr

Concrete mock command for chat-mode testing.

Complete state of an active streaming operation. This replaces scattered state variables with a single, type-safe model.

Loads and processes MCP configuration files with OAuth support.

Result of model discovery from an API. This model represents the outcome of attempting to discover available models from a provider's AP

Orchestrates LLM-driven plan execution with mcp-cli integration. Each step is executed by the LLM: given the step description and tool schema

A single chunk received from the streaming LLM response. This normalizes various chunk formats from different providers into a consistent str

Mock tool manager with execute_tool and stream_execute_tools methods.

A dummy context object with conversation_history and managers.

Routes dashboard messages between the server and one or more bridges.

Information about a connected server instance.

Local web server for hosting MCP Apps in the user's browser.

Local HTTP + WebSocket server for the mcp-cli dashboard shell.

State tracking for a tool currently being executed.

Stub for ToolManager — only needs read_resource for launch tests.

Complete MCP configuration - clean, immutable, type-safe. This is the source of truth loaded from config files. RuntimeConfig wraps this with

MCP-CLI specific dynamic tool provider. Extends BaseDynamicToolProvider with: - ToolManager integration for MCP tool execution - State-aw

Runtime configuration resolver with 4-tier priority. Priority (highest to lowest): 1. CLI overrides (ConfigOverride) 2. Environment varia

Concrete mock command for CLI-mode testing.

Dummy command for testing.

Mock context for testing.

Mock UI manager for testing.

Chat context with SessionManager as the native conversation backend. SessionManager is required - no fallback to local state. All conver

Factory for creating and caching LLM clients. This class manages the creation and caching of LLM clients for different providers and mod

Minimal ToolManager stub.

Minimal ToolManager stub for PlanningContext tests.

Clean, async-native streaming handler. Uses unified display system - no fallbacks, no dual paths.

Manages accumulation of tool calls across streaming chunks. Uses Pydantic models for type-safe state management. Handles fragmented JSON and

Runtime information about a running MCP App.

Manages tool filtering and disabling based on various criteria.

Result of a diagnostic test.

Manages LLM providers, models, and client creation. This is a clean, refactored implementation that: - Uses Pydantic models for type saf

Bridges WebSocket messages from browser to MCP server tool calls.

Configurable mock command for extended testing.

Mock ToolManager for testing execution functions.

Manager for application configuration. This provides a singleton-like pattern for managing configuration.

LLM completion response with streaming metadata - type-safe! This extends the basic chuk_llm CompletionResponse with MCP-CLI-specific streami

Type-safe configuration override from CLI/env. Use this instead of dict[str, Any] for CLI arguments.

Toggle or set tool call confirmation mode.

Mock command for unit tests.

Mock tool manager for testing DynamicToolProvider.

A message in the conversation history (dict-based tool_calls for SessionManager compat). This is the LOCAL message model used by ChatContext/Sess

Mock command for testing.

Manages spawning, stopping, and communicating with agents.

HTTP/SSE server configuration.

Planner → mcp-cli ToolManager adapter with guard integration. Implements the ToolExecutionBackend protocol so that chuk-ai-planner's Universa

Configuration for a single MCP server.

Configuration for a spawned agent. Controls identity, model selection, tool access, and parent relationship.

Manages persistent workspace and global memories with file-based storage.

Complete session data for persistence.

Metadata for a saved session.

Routes chat-engine events to connected browser dashboard clients.

Mock tool manager for testing DynamicToolProvider.

Minimal WebSocket stub.

Response model for server information. Example: >>> server = ServerInfoResponse( ... name="my-server", ... t

Token usage for a single conversation turn.

Unified server configuration supporting both HTTP and STDIO transports. Exactly one of (url, command) must be provided.

Manager for application contexts. This provides a singleton-like pattern for managing the application context.

Dummy command for testing.

Context that exposes a fake session.vm attribute.

Minimal stand-in that satisfies the methods ChatContext uses.

Function call within a tool call (OpenAI format). Mutable counterpart to chuk_llm.core.models.FunctionCall (frozen). Must remain mutable beca

Capabilities and features supported by a provider/model. This model describes what features a specific provider or model supports, such

Server MCP capabilities. Example: >>> caps = ServerCapabilities( ... tools=True, ... prompts=True, .

File-based session persistence. Stores sessions as JSON files in a configurable directory.

Minimal PlanningContext stub for tool tests.

Custom exception for exiting interactive mode without interfering with pytest.

Server-to-tools grouping for system prompt generation.

Mock ToolManager with configurable per-tool delays for batch timeout tests.

Successful ToolManager that tracks lifecycle.

Minimal ToolManager stub for testing McpToolBackend.

Test-specific ChatContext that works with stream_manager instead of ToolManager. This is used for: 1. The --test-mode flag in chat handl

Tracks token usage across conversation turns.

Staging area for ``/attach`` command. Lives on ChatContext.

Mock color scheme for testing.

Timeout configuration with proper defaults. All values in seconds. Immutable after creation. No more magic numbers!

Metadata tracked per in-flight tool call during execution.

A single processed attachment ready for content-block injection.

Dummy command for testing.

A single recall test case.

STDIO server configuration.