Struct Conversation 

pub struct Conversation { /* private fields */ }

Stateful conversation that accumulates chat history across turns

Implementations

impl Conversation

pub fn new(client: HttpClient) -> Self

Create a new conversation using the given HTTP client

Example

use wauldo::{HttpClient, Conversation};
let client = HttpClient::localhost().unwrap();
let conv = Conversation::new(client);

pub fn with_system(self, system: &str) -> Self

Add a system message to the conversation history

The system message is always inserted at position 0 so it precedes all user and assistant turns.

Example

let conv = Conversation::new(client)
    .with_system("You are a helpful Rust expert");

pub fn with_model(self, model: &str) -> Self

Set the model to use for chat completions

Defaults to "default" (server-selected) if not called.

Example

let conv = Conversation::new(client).with_model("llama3");

pub async fn say(&mut self, message: &str) -> Result<String>

Send a user message, receive the assistant reply, and store both in history

The full accumulated history (system + prior turns) is sent with each request so the LLM has complete conversational context.

Example

let mut conv = Conversation::new(client).with_system("Be concise");
let reply = conv.say("What is Rust?").await?;
println!("{}", reply);

pub fn history(&self) -> &[ChatMessage]

Read-only access to the accumulated history

Example

let conv = Conversation::new(client).with_system("sys");
assert_eq!(conv.history().len(), 1);

pub fn clear(&mut self)

Clear conversation history (keeps model setting and system prompt)

Removes all user and assistant messages but preserves the system prompt (if any) so subsequent say() calls retain the same persona.

Example

let mut conv = Conversation::new(client).with_system("sys");
conv.clear();
assert_eq!(conv.history().len(), 1); // system prompt preserved