Enum BeforeModelCallResult 

pub enum BeforeModelCallResult {
    Continue(LlmRequest),
    ShortCircuit(LlmResponse),
}

Available on crate feature plugin only.

Result from a before_model_call hook invocation.

Determines whether the model call continues with a (possibly modified) request, or is short-circuited with a synthetic response.

Examples

Continuing with a modified request

use adk_plugin::BeforeModelCallResult;
use adk_core::LlmRequest;

// Modify the request (e.g., inject a system instruction)
let mut request = LlmRequest::default();
request.model = "gemini-2.5-flash".to_string();
let result = BeforeModelCallResult::Continue(request);

Short-circuiting with a cached response

use adk_plugin::BeforeModelCallResult;
use adk_core::LlmResponse;

// Return a cached response without calling the model
let cached_response = LlmResponse::default();
let result = BeforeModelCallResult::ShortCircuit(cached_response);

Variants

Continue(LlmRequest)

Continue execution with (possibly modified) LLM request.

The contained request will be passed to the next plugin in the chain, and ultimately to the LLM provider if this is the last plugin.

ShortCircuit(LlmResponse)

Short-circuit: skip the model call and use this synthetic response.

When returned, no further plugins in the chain are invoked, the LLM is not called, and this response is used as the model output.

Trait Implementations

impl Debug for BeforeModelCallResult

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

Formats the value using the given formatter.