Struct LlmSimConfig 

pub struct LlmSimConfig {
    pub response: ResponseConfig,
    pub tool_calls: Option<ToolCallConfig>,
    pub simulate_latency: bool,
    pub model_name: String,
    pub response_delay: Option<Duration>,
    pub response_id: Option<String>,
    pub effort_capture: Option<Arc<Mutex<Vec<Option<String>>>>>,
    pub message_capture: Option<Arc<Mutex<Vec<Vec<LlmMessage>>>>>,
}

Configuration for the LlmSim driver

Fields

response: ResponseConfig

Response generation configuration

tool_calls: Option<ToolCallConfig>

Optional tool calls to include in responses

simulate_latency: bool

Enable latency simulation (default: false for fast tests)

model_name: String

Model name to report in metadata

response_delay: Option<Duration>

Optional delay before responding (TTFT - time to first token). This is useful for testing cancellation scenarios where we need a predictable time window to cancel an active turn before completion.

response_id: Option<String>

Optional response ID to include in completion metadata. Enables testing previous_response_id chaining.

effort_capture: Option<Arc<Mutex<Vec<Option<String>>>>>

Optional capture sink for the per-call reasoning_effort (EVE-595). When set, every chat_completion_stream call appends the effort it saw in LlmCallConfig, in call order. Tests use this to assert that a mid-turn effort change is observed by subsequent LLM steps.

message_capture: Option<Arc<Mutex<Vec<Vec<LlmMessage>>>>>

Optional capture sink for the provider-visible messages of each call. When set, every chat_completion_stream call appends the exact LlmMessage slice it received, in call order. Tests use this to assert which messages actually reach the provider after context assembly and message filtering (e.g. Infinity Context history trimming).

Implementations

impl LlmSimConfig

pub fn fixed(response: impl Into<String>) -> Self

Create a new config with a fixed response

pub fn echo() -> Self

Create a new config that echoes user input

pub fn lorem(target_tokens: usize) -> Self

Create a new config with lorem ipsum text

pub fn sequence(responses: Vec<String>) -> Self

Create a new config with a sequence of responses

pub fn scripted(turns: Vec<SimTurn>) -> Self

Create a new config that replays scripted turns in order.

pub fn with_on_exhausted(self, mode: OnExhausted) -> Self

Set the behavior when a scripted response config exhausts its turns.

pub fn with_tool_calls(self, tool_calls: Vec<ToolCall>) -> Self

Add tool calls to the response

pub fn with_tool_call_sequence(self, sequences: Vec<Vec<ToolCall>>) -> Self

Add a sequence of tool calls (different per call)

pub fn with_latency(self) -> Self

Enable latency simulation

pub fn with_model(self, model: impl Into<String>) -> Self

Set model name for metadata

pub fn with_response_delay(self, delay: Duration) -> Self

Set a delay before responding (TTFT - time to first token). This creates a predictable time window for testing cancellation scenarios.

pub fn with_response_id(self, id: impl Into<String>) -> Self

Set a response ID to include in completion metadata (for testing chaining)

pub fn with_effort_capture( self, capture: Arc<Mutex<Vec<Option<String>>>>, ) -> Self

Set a shared capture sink for the per-call reasoning_effort (EVE-595). Every chat_completion_stream call appends the effort it observed in LlmCallConfig, in call order.

pub fn with_message_capture( self, capture: Arc<Mutex<Vec<Vec<LlmMessage>>>>, ) -> Self

Set a shared capture sink for the provider-visible messages of each call. Every chat_completion_stream call appends the exact LlmMessage slice it received, in call order.

pub fn error(message: impl Into<String>) -> Self

Create a new config that returns an error (for testing error handling)

pub fn model_not_available() -> Self

Create a new config that returns a model-not-available error

Trait Implementations

impl Clone for LlmSimConfig

fn clone(&self) -> LlmSimConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LlmSimConfig

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

Formats the value using the given formatter.

impl Default for LlmSimConfig

fn default() -> Self

Returns the “default value” for a type.