Intermediate result of a tool execution.

Standard message format.

Tool parameter definition.

Raised when a tool encounters an error.

Base class for tool errors.

Model provider configuration. For official model providers such as OpenAI and Anthropic, the base_url is optional. api_version is required fo

Model configuration.

A class to encapsulate the result of a tool execution.

Custom exception for tool-related errors.

Google Gemini client wrapper with tool schema generation.

Configuration manager for Trae Agent.

Result of a tool execution.

dataclass for function entry.

A ToolResult that can be rendered as a CLI output.

dataclass for class entry.

Represents a parsed tool call.

Standard LLM response format.

LLM usage format.

OpenRouter client wrapper that maintains compatibility while using the new architecture.

Anthropic client wrapper with tool schema generation.

Main LLM client that supports multiple providers.

Records trajectory data for agent execution and LLM interactions.

Base class for agent-related errors. Used to signal execution failures, misconfigurations, or unexpected LLM/tool behavior.

Represents a console step with its display panel and lakeview information.

Lakeview configuration.

Model parameters for a model provider.

OpenAI client wrapper with tool schema generation.

Trae Agent specialized for software engineering tasks.

Trae agent configuration.

A session of a bash shell.

Encapsulates the entire execution of an agent task. Contains the original task, all intermediate steps, final result, execution metadata

Represents a single step in an agent's execution process. Tracks the state, thought process, tool interactions, LLM response, and any as

Azure client wrapper that maintains compatibility while using the new architecture.

Azure OpenAI provider configuration.

A tool that allows the agent to run bash commands. The tool parameters are defined by Anthropic and are not editable.

Main class for running experiments and evaluations. Handles Docker image management, environment preparation, patch generation, and evaluatio

Manages Docker container lifecycle and command execution for the agent. Supports both stateless (non-interactive) and stateful (interactive)

A ToolExecutor that delegates tool calls to either a local executor or a Docker environment based on the tool's name.

Doubao client wrapper that maintains compatibility while using the new architecture.

Doubao provider configuration.

Tool for editing JSON files using JSONPath expressions.

Tool for editing JSON files using JSONPath expressions.

Supported LLM providers.

Configuration for Lakeview.

OpenRouter provider configuration.

Rich CLI console using Textual for TUI interface.

Textual app for the rich console.

Simple text-based CLI console that prints agent execution trace.

Tool to replace a string in a file.

Tool to replace a string in a file.

Widget to display real-time token usage.

Tool executor that manages tool execution.

Base class for agent configurations.

Defines possible states during an agent's execution lifecycle.

Defines possible states during an agent's execution lifecycle.

Base class for LLM-based agents.

Base class for LLM clients.

Tool to construct and query the code knowledge graph of a codebase.

Base class for CLI console implementations.

Configuration class for agents, models and model providers.

Factory class for creating CLI console instances.

Console operation modes.

Available console types.

State of MCP discovery process.

Base class for OpenAI-compatible clients with shared logic.

Abstract base class for provider-specific configurations.

A tool for sequential thinking that helps break down complex problems. This tool helps analyze problems through a flexible thinking process that

Tool to mark a task as done.