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

docs: correct routing configuration guidance (#786)

ryan-lempka62494ea70d75
diff --git a/crates/switchyard-server/README.md b/crates/switchyard-server/README.md
index bdf2f22fd..45176afb6 100644
--- a/crates/switchyard-server/README.md
+++ b/crates/switchyard-server/README.md
@@ -79,7 +79,10 @@ upstream, and a route's `id` is the model clients send to select that algorithm.
Each target references an entry under `llm_clients`. All configured clients use
`TranslatingLlmClient`; supported formats are `openai_chat`, `openai_responses`, and
`anthropic_messages`. Supported algorithms are `noop`, `random`, `passthrough`,
-`llm_classifier`, and `stage_router`. The optional `prefill-router` feature also enables
+`llm_classifier`, `stage_router`, [`auto`](../../docs/reference/toml_schema.md#auto),
+[`composite`](../../docs/routing_algorithms/composite_routing.md), and
+[`advisor`](../../docs/routing_algorithms/advisor_gate_routing.md).
+The optional `prefill-router` feature also enables
the experimental `prefill_router`. See its
[artifact requirements](../../docs/reference/toml_schema.md#prefill_router).
An `api_key_env` value names an environment variable. The TOML never contains the
diff --git a/docs/reference/toml_schema.md b/docs/reference/toml_schema.md
index d168a64df..4be77ed86 100644
--- a/docs/reference/toml_schema.md
+++ b/docs/reference/toml_schema.md
@@ -342,9 +342,10 @@ configuration. Today a classifier sets the tier a stage router falls open to whe
The tier is retained per session. A deployment that sends no session ID needs
`classifier.message_hash_fallback = true`, which keys on the first user message
-instead. The stage table takes no `picker`: the classifier supplies that tier per turn. A turn the
-classifier cannot reach falls open to the efficient tier. Leaving out
-`classifier` is recommended: that judge runs ahead of the fall-open tier.
+instead. The outer `[routes.<name>.classifier]` table is required. The nested
+`[routes.<name>.stage]` table accepts neither `picker` nor a second `classifier`.
+The outer classifier supplies the default tier. A turn the classifier cannot
+reach falls open to the efficient tier.
### `advisor`
diff --git a/docs/routing_algorithms/composite_routing.md b/docs/routing_algorithms/composite_routing.md
index aaa6c3e9b..e70e96c14 100644
--- a/docs/routing_algorithms/composite_routing.md
+++ b/docs/routing_algorithms/composite_routing.md
@@ -41,12 +41,11 @@ plan = ["create_research_plan"]
new = ["send_message"]
```
-`[routes.switchyard.classifier]` takes the `stage_router` classifier fields.
+`[routes.switchyard.classifier]` is required and takes the `stage_router` classifier fields.
`classify_trigger` sets how often the judge runs: `user_turn` re-picks the tier
whenever the user speaks, `new_session` picks once and holds it.
-`[routes.switchyard.stage]` takes the `stage_router` fields except `picker`, whose
-job the classifier does per turn. Leave `classifier` out as well: that judge runs
-ahead of the fall-open tier, so it answers most of the turns this one decided.
+`[routes.switchyard.stage]` takes the stage settings, but accepts neither `picker`
+nor a second `classifier`. The outer classifier supplies the default tier.
Custom `tool_semantics` remain stage-local and have the same additive behavior as
on a standalone stage route.
diff --git a/docs/routing_algorithms/overview.md b/docs/routing_algorithms/overview.md
index fa26620c5..18941521e 100644
--- a/docs/routing_algorithms/overview.md
+++ b/docs/routing_algorithms/overview.md
@@ -46,7 +46,7 @@ These options remain available when you need a different routing policy.
| [Escalation](escalation_router_routing.md) | Start on the efficient model and escalate when an LLM judge detects trouble. | `llm_classifier` with `mode = "escalation"` |
| [Custom](llm_classifier_routing.md#custom-multi-target-routing) | Route among two or more models using your own classification schema and rules. | `llm_classifier` with `mode = "custom"` |
| [Advisor Gate](advisor_gate_routing.md) | Keep one executor model and have a stronger advisor review its plans and completion claims. | `advisor` |
-| [Sub-Agent-Aware Routing](subagent_routing.md) | Delegated sub-agents should use a separate routing policy from the parent agent. | `passthrough` or `stage_router` with `subagents` |
+| [Sub-Agent-Aware Routing](subagent_routing.md) | Delegated sub-agents should use a separate routing policy from the parent agent. | `passthrough`, `stage_router`, or `composite` with `subagents` |
| [Random Routing](random_routing.md) | You need a fixed traffic split for A/B tests, baselines, or cost experiments. | `random` |
| [Fixed Model](#direct-model-routes) | Send every request to one target without a routing decision. | `passthrough` |
diff --git a/docs/routing_algorithms/subagent_routing.md b/docs/routing_algorithms/subagent_routing.md
index c02af0de2..be6ce7ab8 100644
--- a/docs/routing_algorithms/subagent_routing.md
+++ b/docs/routing_algorithms/subagent_routing.md
@@ -2,7 +2,8 @@
Sub-agent-aware routing leaves parent-agent traffic with its configured routing
algorithm while routing delegated sub-agent work separately. It is available on
-`passthrough` and `stage_router` routes through the optional `subagents` table.
+`passthrough`, `stage_router`, and `composite` routes through the optional
+`subagents` table.
> Requires unreleased features. [Build from source](../getting_started.md#build-from-source) to run this example.