Skip to content

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

feat(server): return outcome metadata from /v1/decision (#1122)

Chinmay Gabelfb89e12ba70b
diff --git a/crates/libsy/src/core/outcome_metadata.rs b/crates/libsy/src/core/outcome_metadata.rs
index ee8277727..2ae10f106 100644
--- a/crates/libsy/src/core/outcome_metadata.rs
+++ b/crates/libsy/src/core/outcome_metadata.rs
@@ -6,7 +6,7 @@
use serde_json::Value;
/// Identity and optional algorithm evidence attached to a successful routing outcome.
-#[derive(Clone, Debug, PartialEq, Eq)]
+#[derive(Clone, Debug, PartialEq, Eq, serde::Serialize)]
pub struct OutcomeMetadata {
outcome_id: String,
/// Stable name of the algorithm that produced the outcome.
diff --git a/crates/switchyard-runner/src/runner.rs b/crates/switchyard-runner/src/runner.rs
index f69712c29..0ddf249b8 100644
--- a/crates/switchyard-runner/src/runner.rs
+++ b/crates/switchyard-runner/src/runner.rs
@@ -35,6 +35,8 @@ pub struct ModelInfo<'a> {
pub struct DecisionDescription {
pub selected: DecisionTarget,
pub fallbacks: Vec<DecisionTarget>,
+ /// Metadata the algorithm attached to the routing outcome, if any.
+ pub metadata: Option<libsy::OutcomeMetadata>,
}
/// Non-secret configured target details returned by the decision endpoint.
@@ -163,6 +165,7 @@ impl Runner {
Some(DecisionDescription {
selected: resolve(model_ids.next()?)?,
fallbacks: model_ids.map(resolve).collect::<Option<Vec<_>>>()?,
+ metadata: outcome.metadata.clone(),
})
}
}
diff --git a/crates/switchyard-server/README.md b/crates/switchyard-server/README.md
index fd71a5c8f..c3edb7454 100644
--- a/crates/switchyard-server/README.md
+++ b/crates/switchyard-server/README.md
@@ -345,7 +345,9 @@ returns the selected target and ordered fallbacks with their model, format, base
`extra_body`. It does not make a post-routing answer call. Routing-time calls still execute, and
response-dependent algorithms such as escalation and advisor routing may produce an answer while
deciding. When they do, the endpoint includes the buffered answer as `response`, encoded in
-`input_format`; otherwise the field is omitted.
+`input_format`; otherwise the field is omitted. When the algorithm attaches outcome metadata, the
+endpoint returns it as `metadata`, with the outcome ID, the algorithm name and the algorithm's
+evidence.
For `stage_router`, `algorithm_stats.stage_router` groups routing decisions by source and semantic
target and summarizes its score, confidence, and input-dimension histograms. These values reset
diff --git a/crates/switchyard-server/src/lib.rs b/crates/switchyard-server/src/lib.rs
index b27dc20b0..d34b5222d 100644
--- a/crates/switchyard-server/src/lib.rs
+++ b/crates/switchyard-server/src/lib.rs
@@ -140,6 +140,8 @@ struct DecisionResponse {
selected: DecisionTargetResponse,
fallbacks: Vec<DecisionTargetResponse>,
#[serde(skip_serializing_if = "Option::is_none")]
+ metadata: Option<libsy::OutcomeMetadata>,
+ #[serde(skip_serializing_if = "Option::is_none")]
response: Option<Value>,
}
@@ -284,6 +286,7 @@ impl ServerState {
Some(DecisionResponse {
selected: convert(description.selected),
fallbacks: description.fallbacks.into_iter().map(convert).collect(),
+ metadata: description.metadata,
response,
})
}
diff --git a/crates/switchyard-server/tests/server.rs b/crates/switchyard-server/tests/server.rs
index 51299552f..7261261d8 100644
--- a/crates/switchyard-server/tests/server.rs
+++ b/crates/switchyard-server/tests/server.rs
@@ -2432,8 +2432,13 @@ escalation = {{ confirmations = 1 }}
.await?;
assert_eq!(response.status, StatusCode::OK);
+ let mut body = response.json()?;
+ let outcome_id = body["metadata"]
+ .as_object_mut()
+ .and_then(|metadata| metadata.remove("outcome_id"));
+ assert!(outcome_id.is_some_and(|id| id.as_str().is_some_and(|id| !id.is_empty())));
assert_eq!(
- response.json()?,
+ body,
json!({
"selected": {
"target": "economy",
@@ -2453,6 +2458,10 @@ escalation = {{ confirmations = 1 }}
},
"extra_body": {},
}],
+ "metadata": {
+ "algorithm": "llm_task_classifier",
+ "evidence": {"source": "llm-classifier", "score": 0.9, "threshold": 0.5},
+ },
})
);
assert_eq!(