Trait LlmBackend

Source

pub trait LlmBackend: Send + Sync {
    // Required methods
    fn complete(&self, prompt: &str, max_tokens: u32) -> Result<String>;
    fn name(&self) -> &'static str;

    // Provided method
    fn complete_usage(
        &self,
        prompt: &str,
        max_tokens: u32,
    ) -> Result<(String, LlmUsage)> { ... }
}

Expand description

One AI call: a prompt in, the model’s text reply out.

Required Methods§

Source

fn complete(&self, prompt: &str, max_tokens: u32) -> Result<String>

Source

fn name(&self) -> &'static str

Stable label for logs / provenance.

Provided Methods§

Source

fn complete_usage( &self, prompt: &str, max_tokens: u32, ) -> Result<(String, LlmUsage)>

Like complete but also reports token usage when the backend exposes it. Default: run complete and report no usage, so mocks and minimal backends need not implement it.