From c4eaf5088b2d96dc8500bc839748a17e8ea83b9aae2389ea2af36a314d5ae43b Mon Sep 17 00:00:00 2001
From: tlg <thomas.langer@destengs.com>
Date: Sat, 4 Apr 2026 07:31:10 +0200
Subject: [PATCH] feat: model registry with virtual-to-physical resolution

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 kischdle/llmux/llmux/model_registry.py      | 36 +++++++++++
 kischdle/llmux/tests/test_model_registry.py | 66 +++++++++++++++++++++
 2 files changed, 102 insertions(+)
 create mode 100644 kischdle/llmux/llmux/model_registry.py
 create mode 100644 kischdle/llmux/tests/test_model_registry.py

diff --git a/kischdle/llmux/llmux/model_registry.py b/kischdle/llmux/llmux/model_registry.py
new file mode 100644
index 0000000..ca52846
--- /dev/null
+++ b/kischdle/llmux/llmux/model_registry.py
@@ -0,0 +1,36 @@
+from llmux.config import PhysicalModel, VirtualModel, load_models_config
+
+
+class ModelRegistry:
+    def __init__(
+        self,
+        physical: dict[str, PhysicalModel],
+        virtual: dict[str, VirtualModel],
+    ):
+        self._physical = physical
+        self._virtual = virtual
+
+    @classmethod
+    def from_config(cls) -> "ModelRegistry":
+        physical, virtual = load_models_config()
+        return cls(physical, virtual)
+
+    def list_virtual_models(self) -> list[dict]:
+        return [
+            {
+                "id": name,
+                "object": "model",
+                "created": 0,
+                "owned_by": "llmux",
+            }
+            for name in self._virtual
+        ]
+
+    def resolve(self, virtual_name: str) -> tuple[str, PhysicalModel, dict]:
+        """Resolve a virtual model name to (physical_id, PhysicalModel, params)."""
+        vm = self._virtual[virtual_name]  # raises KeyError if unknown
+        pm = self._physical[vm.physical]
+        return vm.physical, pm, dict(vm.params)
+
+    def get_physical(self, physical_id: str) -> PhysicalModel:
+        return self._physical[physical_id]  # raises KeyError if unknown
diff --git a/kischdle/llmux/tests/test_model_registry.py b/kischdle/llmux/tests/test_model_registry.py
new file mode 100644
index 0000000..f9279bd
--- /dev/null
+++ b/kischdle/llmux/tests/test_model_registry.py
@@ -0,0 +1,66 @@
+import pytest
+
+from llmux.model_registry import ModelRegistry
+
+
+@pytest.fixture
+def registry():
+    return ModelRegistry.from_config()
+
+
+def test_list_virtual_models(registry):
+    models = registry.list_virtual_models()
+    assert len(models) == 16
+    names = [m["id"] for m in models]
+    assert "Qwen3.5-9B-FP8-Thinking" in names
+    assert "GPT-OSS-20B-High" in names
+    assert "cohere-transcribe" in names
+    assert "Chatterbox-Multilingual" in names
+
+
+def test_virtual_model_openai_format(registry):
+    models = registry.list_virtual_models()
+    m = next(m for m in models if m["id"] == "Qwen3.5-9B-FP8-Thinking")
+    assert m["object"] == "model"
+    assert m["owned_by"] == "llmux"
+
+
+def test_resolve_virtual_to_physical(registry):
+    physical_id, physical, params = registry.resolve("Qwen3.5-9B-FP8-Thinking")
+    assert physical_id == "qwen3.5-9b-fp8"
+    assert physical.backend == "transformers"
+    assert params == {"enable_thinking": True}
+
+
+def test_resolve_instruct_variant(registry):
+    physical_id, physical, params = registry.resolve("Qwen3.5-9B-FP8-Instruct")
+    assert physical_id == "qwen3.5-9b-fp8"
+    assert params == {"enable_thinking": False}
+
+
+def test_resolve_gpt_oss_reasoning(registry):
+    physical_id, physical, params = registry.resolve("GPT-OSS-20B-Medium")
+    assert physical_id == "gpt-oss-20b"
+    assert params == {"system_prompt_prefix": "Reasoning: medium"}
+
+
+def test_resolve_same_physical_for_variants(registry):
+    pid1, _, _ = registry.resolve("Qwen3.5-9B-FP8-Thinking")
+    pid2, _, _ = registry.resolve("Qwen3.5-9B-FP8-Instruct")
+    assert pid1 == pid2
+
+
+def test_resolve_unknown_model_raises(registry):
+    with pytest.raises(KeyError):
+        registry.resolve("nonexistent-model")
+
+
+def test_get_physical(registry):
+    physical = registry.get_physical("qwen3.5-9b-fp8")
+    assert physical.type == "llm"
+    assert physical.estimated_vram_gb == 9
+
+
+def test_get_physical_unknown_raises(registry):
+    with pytest.raises(KeyError):
+        registry.get_physical("nonexistent")