pig/agent/msg

Agent message type for the sans-IO core.

Messages drive the state machine. They represent both user inputs and runtime responses fed back into the core:

UserPrompt — a new prompt to process
ProviderResponded — the LLM’s response (or error)
ToolResults — tool execution outcomes

Types

AgentMsg

</>

Messages that drive the agent state machine.

pub type AgentMsg {
  UserPrompt(String)
  ProviderResponded(Result(message.Message, error.AiError))
  ToolResults(
    List(#(message.ToolCall, Result(json.Json, tool.ToolError))),
  )
}

Constructors

```
UserPrompt(String)
```
A new user prompt. Starts or continues the conversation.
```
ProviderResponded(Result(message.Message, error.AiError))
```
The LLM provider’s response, delivered by the runtime after executing a CallProvider effect.
```
ToolResults(
  List(#(message.ToolCall, Result(json.Json, tool.ToolError))),
)
```
Tool execution outcomes, delivered by the runtime after executing an ExecuteTools effect.

✨