Enum OutputStrategy 

#[non_exhaustive]
pub enum OutputStrategy {
    Auto,
    Native,
    Tool,
    Prompted,
}

How the codec should ask the model to honour the schema.

Industry consensus (LangChain 1.0 ProviderStrategy/ ToolStrategy, pydantic-ai 1.90 NativeOutput/ToolOutput/ PromptedOutput/TextOutput, BAML SAP, Vercel AI SDK 5 generateObject, Instructor’s mode flag) converges on three dispatch shapes plus an automatic picker:

• Native — vendor-native structured output channel (OpenAI text.format = json_schema, Gemini responseJsonSchema, Anthropic output_config.format = json_schema). Strictest; the vendor itself rejects malformed responses.
• Tool — single forced tool call whose input schema is the target schema. Mature on every vendor (the tool-call surface predates native structured output by a year+); slightly less efficient because the model emits a tool_use block instead of plain assistant text.
• Prompted — schema injected into the system prompt; the reply is parsed best-effort. Last resort for vendors with neither native nor tool support, and for “I want a typed answer but the model is non-reasoning” flows. Deferred to 1.1 — complete_typed rejects this strategy at runtime today.
• Auto — codec picks per-vendor at codec-construction time (NOT per request — per-request resolution would let the same logical request resolve differently across replays, breaking the SessionGraph event log’s deterministic-replay guarantee). The picked strategy is what Codec::auto_output_strategy(model) returns.

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

Auto

Codec picks per-vendor at codec-construction time. Default.

Native

Vendor-native structured output channel.

Tool

Forced single tool call carrying the target schema.

Prompted

Schema injected into the system prompt (1.1 — currently rejected at encode time).

Trait Implementations

impl Clone for OutputStrategy

fn clone(&self) -> OutputStrategy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for OutputStrategy

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

Formats the value using the given formatter.

impl Default for OutputStrategy

fn default() -> OutputStrategy

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for OutputStrategy

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for OutputStrategy

fn eq(&self, other: &OutputStrategy) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for OutputStrategy

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Copy for OutputStrategy

impl Eq for OutputStrategy

impl StructuralPartialEq for OutputStrategy