refactor(python): remove deprecated server stack (#343)

Signed-off-by: nachiketb <nachiketb@nvidia.com>

Author: nachiketb-nvidia

.agents/skills/switchyard-stage-router-scorer/SKILL.md

@@ -16,7 +16,7 @@ Minimal steps to reproduce. Include the command line, the inbound request shape,
 
 ```bash
 # example
-switchyard serve --routes examples/route.yaml --port 4000
+switchyard-server --config routes.toml --port 4000
 curl -s http://localhost:4000/v1/chat/completions -d '{"model":"...","messages":[...]}'
 ```
 

.github/ISSUE_TEMPLATE/bug_report.md

@@ -168,7 +168,7 @@ jobs:
     runs-on: ubuntu-latest
     env:
       SWITCHYARD_DEFAULT_PACKAGE: "nemo-switchyard @ file://${{ github.workspace }}"
-      SWITCHYARD_EXTRAS_PACKAGE: "nemo-switchyard[cli,server] @ file://${{ github.workspace }}"
+      SWITCHYARD_EXTRAS_PACKAGE: "nemo-switchyard[cli] @ file://${{ github.workspace }}"
     steps:
       - uses: actions/checkout@v4
       - uses: astral-sh/setup-uv@v6
@@ -213,7 +213,7 @@ jobs:
               sys.exit(f"FAIL: heavy packages pulled into slim install: {extras}")
           print(f"OK: all {len(forbidden)} heavy packages absent from slim install.")
           PY
-      - name: Verify CLI and server imports work with extras
+      - name: Verify CLI works with its extra
         working-directory: /tmp
         run: |
           uv run --isolated --no-project --python 3.12 \
@@ -225,9 +225,6 @@ jobs:
           uv run --isolated --no-project --python 3.12 \
             --with "${SWITCHYARD_EXTRAS_PACKAGE}" \
             switchyard launch claude --help 2>&1 | grep -q -- '--config' || { echo 'FAIL: --config flag missing from help'; exit 1; }
-          uv run --isolated --no-project --python 3.12 \
-            --with "${SWITCHYARD_EXTRAS_PACKAGE}" \
-            switchyard serve --help > /dev/null
 
   # Single required check for branch protection. Gates on every job above so
   # the ruleset never needs editing when jobs are added or renamed.

.github/workflows/ci.yml

@@ -55,8 +55,8 @@ jobs:
         run: uv pip install aiperf
 
       # ------------------------------------------------------------------
-      # Start a local zero-latency mock OpenAI upstream. This replaces the
-      # The proxy serves a real `type: passthrough` chain that
+      # Start a local zero-latency mock OpenAI upstream. The proxy serves a
+      # real `type: passthrough` route that
       # forwards to this loopback stub, which returns a fixed completion
       # instantly. The extra loopback hop adds a small constant overhead.
       # ------------------------------------------------------------------
@@ -127,21 +127,30 @@ jobs:
           echo "STUB_PID=$!" >> "$GITHUB_ENV"
 
       # ------------------------------------------------------------------
-      # Start the proxy: a `type: passthrough` route pointed at the local stub.
+      # Build and start the native proxy against the local stub.
       # ------------------------------------------------------------------
+      - name: Build switchyard server
+        run: cargo build --locked --release -p switchyard-server
+
       - name: Start switchyard proxy
         run: |
-          cat > bench.yaml <<YAML
-          defaults:
-            api_key: dummy
-            base_url: http://localhost:$STUB_PORT/v1
-            format: openai
-          routes:
-            mock-model:
-              type: passthrough
-              target: mock-model
-          YAML
-          uv run switchyard serve --routes bench.yaml --port $PROXY_PORT &
+          cat > bench.toml <<TOML
+          schema_version = 1
+
+          [llm_clients.stub]
+          format = "openai_chat"
+          base_url = "http://localhost:$STUB_PORT/v1"
+
+          [targets.mock]
+          id = "mock-model"
+          llm_client = "stub"
+
+          [routes.mock]
+          id = "mock-model"
+          type = "passthrough"
+          target = "mock"
+          TOML
+          target/release/switchyard-server --config bench.toml --port $PROXY_PORT &
           echo "PROXY_PID=$!" >> "$GITHUB_ENV"
 
       - name: Wait for proxy to be ready