feat: API routes for models, chat, transcription, speech, and admin

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

kischdle/llmux/tests/test_routes.py

@@ -0,0 +1,62 @@
+import pytest
+from fastapi import FastAPI
+from fastapi.testclient import TestClient
+
+from llmux.config import ApiKey
+from llmux.auth import create_api_key_dependency
+from llmux.model_registry import ModelRegistry
+from llmux.vram_manager import VRAMManager
+from llmux.routes.models import create_models_router
+
+API_KEY = "sk-test-key"
+
+
+@pytest.fixture
+def registry():
+    return ModelRegistry.from_config()
+
+
+@pytest.fixture
+def vram_manager():
+    return VRAMManager(total_vram_gb=16.0)
+
+
+@pytest.fixture
+def app(registry, vram_manager):
+    keys = [ApiKey(key=API_KEY, name="Test")]
+    require_api_key = create_api_key_dependency(keys)
+    app = FastAPI()
+    app.include_router(create_models_router(registry, require_api_key))
+    return app
+
+
+@pytest.fixture
+def client(app):
+    return TestClient(app)
+
+
+@pytest.fixture
+def auth_headers():
+    return {"Authorization": f"Bearer {API_KEY}"}
+
+
+def test_list_models_returns_16(client, auth_headers):
+    resp = client.get("/v1/models", headers=auth_headers)
+    assert resp.status_code == 200
+    body = resp.json()
+    assert body["object"] == "list"
+    assert len(body["data"]) == 16
+
+
+def test_list_models_contains_expected_names(client, auth_headers):
+    resp = client.get("/v1/models", headers=auth_headers)
+    names = [m["id"] for m in resp.json()["data"]]
+    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_list_models_requires_auth(client):
+    resp = client.get("/v1/models")
+    assert resp.status_code == 401