Struct ModelPolicy 

pub struct ModelPolicy {

    pub provider: Option<String>,
    pub model: Option<String>,
    pub model_tier: Option<String>,
    pub temperature: Option<f64>,
    pub max_tokens: Option<i64>,
    pub max_iterations: Option<usize>,
    pub max_nudges: Option<usize>,
    pub nudge: Option<String>,
    pub tool_examples: Option<String>,
    pub post_turn_callback: Option<EqIgnored<VmValue>>,
    pub stop_after_successful_tools: Option<Vec<String>>,
    pub require_successful_tools: Option<Vec<String>>,
    pub turn_policy: Option<TurnPolicy>,
    pub tool_format: Option<String>,
    pub native_tool_fallback: NativeToolFallbackPolicy,
}

Fields

provider: Option<String>

model: Option<String>

model_tier: Option<String>

temperature: Option<f64>

max_tokens: Option<i64>

max_iterations: Option<usize>

Maximum agent_loop iterations for this stage. Overrides the default 16.

max_nudges: Option<usize>

Maximum consecutive text-only (no tool call) responses before declaring stuck.

nudge: Option<String>

Custom nudge message injected when the model produces text without tool calls. If omitted, the VM uses a generic “Continue — use a tool call” message.

tool_examples: Option<String>

Few-shot tool-call examples injected into the tool contract prompt, shown before the tool schema listing. Pipelines provide these — the VM has no hardcoded tool names.

post_turn_callback: Option<EqIgnored<VmValue>>

Optional Harn closure called after each tool-calling turn. Receives turn metadata; returns either a string user message to inject, a bool stop flag, or a dict like {message, stop}. Wrapped in EqIgnored so it doesn’t affect PartialEq derivation.

stop_after_successful_tools: Option<Vec<String>>

When set, the stage stops after any tool-calling turn whose successful results include one of these tool names. This is useful for workflow-owned verify loops where a productive write turn should hand control back to verification immediately.

require_successful_tools: Option<Vec<String>>

When set, the stage is reported as failed unless at least one of these tool names succeeds during the interaction. Pipelines use this to assert a stage cannot quietly finish without running a specific tool.

turn_policy: Option<TurnPolicy>

Turn-shape constraints for action stages.

tool_format: Option<String>

Tool-calling contract format for the per-stage agent loop. "native" routes user tools through the provider’s native function-call channel; "text" embeds the <tool_call> contract prompt. Mirrors the top-level agent_loop(..., {tool_format: ...}) option. When None, the workflow falls back to HARN_AGENT_TOOL_FORMAT env / the provider-model default.

native_tool_fallback: NativeToolFallbackPolicy

Policy for native-tool stages when a provider emits text-mode <tool_call> output instead of native tool calls.

Trait Implementations

impl Clone for ModelPolicy

fn clone(&self) -> ModelPolicy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelPolicy

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

Formats the value using the given formatter.

impl Default for ModelPolicy

fn default() -> ModelPolicy

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for ModelPolicy

where ModelPolicy: Default,

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for ModelPolicy

fn eq(&self, other: &ModelPolicy) -> 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 ModelPolicy

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

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for ModelPolicy