[Core][Model] Fit PersonaPlex into the Unified Full-duplex Framework (RFC #7181 PR 2) (#7695)

Signed-off-by: chickeyton <ngton2014@gmail.com>
Signed-off-by: Yueqian Lin <linyueqian@outlook.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: Yueqian Lin <linyueqian@outlook.com>

tests/entrypoints/duplex/test_duplex_warmup.py

@@ -0,0 +1,53 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+
+"""The startup warmup sends the model's own silence unit, not a fixed 16 kHz one."""
+
+from __future__ import annotations
+
+import base64
+
+import pytest
+
+from vllm_omni.entrypoints.duplex.warmup import warmup_silence_unit
+
+pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
+
+
+class _Plugin:
+    silence_continuation_samples = 1920
+
+    def silence_unit_payload(self) -> dict[str, object]:
+        return {
+            "type": "audio",
+            "audio": base64.b64encode(bytes(1920 * 4)).decode("ascii"),
+            "format": "pcm_f32le",
+            "sample_rate_hz": 24000,
+        }
+
+
+class _LegacyPlugin:
+    silence_continuation_samples = 8000
+
+
+def test_the_plugin_unit_is_used_verbatim() -> None:
+    unit = warmup_silence_unit(_Plugin())
+
+    assert unit["format"] == "pcm_f32le"
+    assert unit["sample_rate_hz"] == 24000
+    assert len(base64.b64decode(str(unit["audio"]))) == 1920 * 4
+
+
+def test_a_plugin_without_the_hook_falls_back_to_16k_zeros_of_its_sample_count() -> None:
+    unit = warmup_silence_unit(_LegacyPlugin())
+
+    assert unit["sample_rate_hz"] == 16000
+    assert unit["format"] == "pcm_f32le"
+    assert len(base64.b64decode(str(unit["audio"]))) == 8000 * 4
+
+
+def test_no_plugin_means_the_minicpm_sized_default() -> None:
+    unit = warmup_silence_unit(None)
+
+    assert unit["sample_rate_hz"] == 16000
+    assert len(base64.b64decode(str(unit["audio"]))) == 16000 * 4