Enum ResponseFormat 

pub enum ResponseFormat {
    JsonSchema {
        schema: Value,
        strict: bool,
    },
    Grammar {
        gbnf: String,
    },
}

Constrains the output format of a LlmCompletionPort::chat_stream call.

Pass Some(&format) when the caller requires structured output (e.g. for plan generation in the orchestrator). Adapters that target llama-server translate these variants as follows:

Variant	Wire field
JsonSchema	response_format: { type: "json_schema", json_schema: { schema, strict } }
Grammar	grammar: "<gbnf string>" (llama.cpp extension)

Normal agent-loop calls pass None, which leaves the model free-form.

Variants

JsonSchema

Constrain the output to a JSON object matching the given JSON Schema.

strict: true instructs the model to refuse outputs that do not conform to the schema. Use strict: false for best-effort guidance when strict validation would be overly rigid.

Fields

schema: Value

A valid JSON Schema object (Draft-07 or later).

strict: bool

Whether the model should refuse outputs that violate the schema.

Grammar

Constrain output using a GBNF grammar string (llama.cpp extension).

GBNF grammars are more expressive than JSON Schema for some use cases (e.g. constraining enum-only outputs without a schema round-trip).

Fields

gbnf: String

A valid GBNF grammar string understood by llama-server.

Trait Implementations

impl Clone for ResponseFormat

fn clone(&self) -> ResponseFormat

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ResponseFormat

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.