Module agent

Expand description

§Core components of AI Agents

This module contains core components that you can use in your AI Agents

To read more about Agents look into crate::agent::Agent

To read more about structured output look into crate::structured_output

To read more about tool look into crate::tool

Agent: The Agent struct represents an agent that interacts with a chat model.
AgentRequest: A prepared run of an Agent, created by Agent::run or Agent::run_stream.
AgentResult: The result of a successful run.
AgentStream: Stream of AgentEvents of one run, returned by awaiting Agent::run_stream.
ChatOptions: Model options sent with every request: temperature, max tokens, tool choice, reasoning effort, …
Complete: Mode of an AgentRequest created by Agent::run: awaiting it returns a AgentResult.
LlmError: Error of the LLM client: a failed request, or a client that could not be built.
RunOptions: Settings that replace the Agent configuration for one run.
Streaming: Mode of an AgentRequest created by Agent::run_stream: awaiting it returns an AgentStream.
Usage: Tokens used by a run. See AgentResult::usage.

AgentError: Errors returned by Agent.
AgentEvent: Something that happened during a streamed run, see Agent::run_stream.
Provider: LLM provider for Agent::new_with_provider.
ReasoningEffort: How much a model should reason. See ChatOptions::reasoning_effort.
StopReason: Why the model stopped writing. See AgentResult::stop_reason.
ToolChoice: Whether and which tools the model must call. See ChatOptions::tool_choice.