Switchyard lets LLM applications route traffic across models and providers while preserving native OpenAI and Anthropic API compatibility - enabling flexible model selection, benchmarking, and cost/performance optimization.
Rust
// SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
// SPDX-License-Identifier: Apache-2.0
//! Sub-agent override as a single SDK component.
//!
//! [`SubagentOverride`] scores one fixed worker target for requests carrying delegated
//! sub-agent work ([`Metadata::is_subagent_work`]) and abstains for everything else, so a
//! cascade falls through to its later classifiers on ordinary traffic.
//!
//! It is stateless and holds only the worker's *name*: the fall-through cascade resolves it
//! against its
//! target set. Keeping the policy independent of any memory of past decisions is what lets
//! it compose with a stateful classifier such as
//! [`AffinityRouter`](crate::algorithms::AffinityRouter) — the override decides *which*
//! target delegated work belongs on, affinity decides *how long* a decision lives, and
//! neither needs to know about the other.
use std::sync::Arc;
use async_trait::async_trait;
use crate::Result;
use crate::core::algorithm::Driver;
use crate::core::classifier::{Classification, Classifier, Score};
use crate::core::state::State;
use switchyard_protocol::{
ContentBlock, LlmRequest, Message, Metadata, ModelId, Request, Response, Role,
};
/// Classifies delegated work from its parent-supplied prompt and abstains otherwise.
///
/// The inner classifier receives a prompt-only request clone. The original request remains
/// unchanged for the selected child model.
pub struct SubagentGate {
inner: Arc<dyn Classifier>,
}
impl SubagentGate {
/// Wraps `inner` with delegated-work detection.
pub fn new(inner: Arc<dyn Classifier>) -> Self {
Self { inner }
}
}
/// Builds a classifier request containing only the delegated task text.
fn delegated_prompt_request(request: &Request) -> Option<Request> {
// Agent clients put the delegated task after added context and reminders.
// A user message containing only tool-result blocks has no top-level text.
// Find the latest nonblank user text without using text inside tool results.
let prompt = request
.llm_request
.messages
.iter()
.rev()
.filter(|message| message.role == Role::User)
.find_map(|message| {
message.content.iter().rev().find_map(|block| match block {
ContentBlock::Text { text } if !text.trim().is_empty() => Some(text.clone()),
_ => None,
})
})?;
Some(Request {
llm_request: LlmRequest {
model: request.llm_request.model.clone(),
messages: vec![Message::text(Role::User, prompt)],
..LlmRequest::default()
},
raw_request: None,
metadata: request.metadata.clone(),
})
}
#[async_trait]
impl Classifier for SubagentGate {
async fn score(
&self,
state: &mut State,
request: &mut Request,
driver: &Driver,
) -> Result<(Classification, Option<Response>)> {
if !request
.metadata
.as_ref()
.is_some_and(Metadata::is_subagent_work)
{
return Ok((Classification::Scores(Vec::new()), None));
}
let Some(mut classifier_request) = delegated_prompt_request(request) else {
return Ok((Classification::Scores(Vec::new()), None));
};
self.inner
.score(state, &mut classifier_request, driver)
.await
}
}
/// Scores a fixed worker target for delegated sub-agent work; abstains otherwise.
pub struct SubagentOverride {
/// Name of the worker target, resolved by the cascade against its target set.
worker: ModelId,
}
impl SubagentOverride {
/// Creates an override scoring `worker` for delegated sub-agent work.
///
/// `worker` must name a target in the cascade's set, or routing a sub-agent request
/// fails with [`LibsyError::TargetNotFound`](crate::LibsyError::TargetNotFound).
pub fn new(worker: impl Into<ModelId>) -> Self {
Self {
worker: worker.into(),
}
}
}
#[async_trait]
impl Classifier for SubagentOverride {
async fn score(
&self,
_state: &mut State,
request: &mut Request,
driver: &Driver,
) -> Result<(Classification, Option<Response>)> {
// Delegated *work* only. A harness maintenance turn (e.g. Codex `compact`) carries
// sub-agent lineage but is not delegated work, so it abstains and routes normally.
let is_delegated_work = request
.metadata
.as_ref()
.is_some_and(Metadata::is_subagent_work);
if is_delegated_work {
driver.set_evidence(serde_json::json!({"source": "subagent"}));
}
Ok((
Classification::Scores(if is_delegated_work {
vec![Score {
confidence: 1.0,
target: self.worker.clone(),
category: None,
}]
} else {
Vec::new()
}),
None,
))
}
}
#[cfg(test)]
mod tests {
use super::*;
use crate::core::testing::empty_driver;
use switchyard_protocol::{slice_to_header_map, text_request};
fn request(headers: &[(&str, &str)]) -> Request {
let metadata =
(!headers.is_empty()).then(|| Metadata::from_headers(&slice_to_header_map(headers)));
Request {
llm_request: text_request(Some(ModelId::from("auto").to_string()), "hi"),
raw_request: None,
metadata,
}
}
/// Scores `headers` through the override, returning the winning target if it scored.
async fn selected(headers: &[(&str, &str)]) -> Result<Option<ModelId>> {
let mut state = State::default();
let classification = SubagentOverride::new("worker")
.score(&mut state, &mut request(headers), &empty_driver())
.await?;
Ok(classification.0.argmax(false)?.map(|score| score.target))
}
#[tokio::test]
async fn requests_without_metadata_abstain() -> Result<()> {
assert_eq!(selected(&[]).await?, None);
Ok(())
}
#[tokio::test]
async fn subagent_work_scores_the_worker() -> Result<()> {
// Claude Code child-agent lineage.
let claude = &[
("x-claude-code-session-id", "root"),
("x-claude-code-agent-id", "child-1"),
];
assert_eq!(selected(claude).await?, Some(ModelId::from("worker")));
// Codex delegated-work kinds.
assert_eq!(
selected(&[("x-openai-subagent", "review")]).await?,
Some(ModelId::from("worker"))
);
assert_eq!(
selected(&[("x-openai-subagent", "collab_spawn")]).await?,
Some(ModelId::from("worker"))
);
Ok(())
}
#[tokio::test]
async fn harness_maintenance_turns_abstain() -> Result<()> {
assert_eq!(selected(&[("x-openai-subagent", "compact")]).await?, None);
assert_eq!(
selected(&[("x-switchyard-is-subagent", "false")]).await?,
None
);
Ok(())
}
#[tokio::test]
async fn delegated_work_is_scored_definitively() -> Result<()> {
// Confidence 1.0 under `Scores` (never `Ambiguous`), so the cascade stops here
// rather than consulting later classifiers.
let mut state = State::default();
let classification = SubagentOverride::new("worker")
.score(
&mut state,
&mut request(&[("x-openai-subagent", "review")]),
&empty_driver(),
)
.await?;
match classification.0 {
Classification::Scores(scores) => {
assert_eq!(scores.len(), 1);
assert_eq!(scores[0].confidence, 1.0);
}
Classification::Ambiguous(_) => panic!("override must score definitively"),
}
Ok(())
}
}