Struct Generation 

pub struct Generation { /* private fields */ }

One call being assembled.

Every setting is optional; ai.prompt("…").generate().await is a complete program.

Implementations

impl Generation

pub fn model(self, model: impl Into<String>) -> Generation

pub fn system(self, system: impl Into<String>) -> Generation

pub fn prompt(self, text: impl Into<String>) -> Generation

What to ask. Sugar for a single user message.

pub fn message(self, message: Message) -> Generation

pub fn conversation(self, conversation: Conversation) -> Generation

pub fn temperature(self, temperature: f64) -> Generation

pub fn max_tokens(self, max_tokens: u32) -> Generation

pub fn stop(self, sequence: impl Into<String>) -> Generation

pub fn tool(self, tool: Tool) -> Generation

Offer the model a tool it may ask for but that this crate will not run — for a caller who wants to handle the tool call themselves.

pub fn tools(self, toolbox: Toolbox) -> Generation

Offer tools and run them: Generation::generate then loops until the model answers in words.

pub fn max_rounds(self, max_rounds: usize) -> Generation

How many model turns the tool loop may take.

pub fn request(&self) -> &Request

The request as it will be sent, for tests and for inspection.

pub async fn generate(self) -> Result<Completion>

Ask, and wait for the whole answer.

pub async fn text(self) -> Result<String>

Ask, and keep only the words — the shortest useful call there is.

pub async fn stream(self) -> Result<TextStream>

Ask, and read the answer as it is written.

Tools are not run while streaming: a tool call is not text, and a caller who wants both should use Generation::generate.

pub async fn generate_as(self, schema: &Schema) -> Result<Json>

Ask for JSON matching a schema, and get it parsed.

Retries once if the model’s answer will not parse; see structured::generate_as.