Struct CommandLlm 

pub struct CommandLlm { /* private fields */ }

A subprocess LLM backend. One process per call; argv is whitespace-split with no shell (identical rules to CommandEmbed).

Implementations

impl CommandLlm

pub fn new(cmd: &str, model: Option<&str>) -> Result<Self>

Construct and probe. The probe ({"loop":1,"op":"probe"}) must return JSON with a model (or one is supplied), so a misconfigured command fails at construction, not mid-run.

Trait Implementations

impl LlmBackend for CommandLlm

fn model(&self) -> &str

Model identifier, stamped as provenance on origin = llm grains.

fn complete(&self, request: &str) -> Result<String>

Run one request. request is a JSON string; the returned text is expected to be JSON and is validated by the caller.