✨config ModelEngine Service

Author: Zhi-a

backend/services/model_management_service.py

@@ -48,8 +48,6 @@ async def create_model_for_tenant(user_id: str, tenant_id: str, model_data: Dict
         model_data['ssl_verify'] = True
         if "open/router" in model_base_url:
             model_data['ssl_verify'] = False
-
-
         # Split model_name into repo and name
         model_repo, model_name = split_repo_name(
             model_data["model_name"]) if model_data.get("model_name") else ("", "")

backend/services/model_provider_service.py

@@ -85,11 +85,9 @@ async def get_models(self, provider_config: Dict) -> List[Dict]:
             List of models with canonical fields
         """
         try:
-            # Allow overriding host and api key via provider_config (from frontend).
-            # Fall back to environment-configured values.
             model_type: str = provider_config.get("model_type", "")
-            host = provider_config.get("base_url") or MODEL_ENGINE_HOST
-            api_key = provider_config.get("api_key") or MODEL_ENGINE_APIKEY
+            host = provider_config.get("base_url")
+            api_key = provider_config.get("api_key")
 
             if not host or not api_key:
                 logger.warning("ModelEngine host or api key not configured")
@@ -135,7 +133,6 @@ async def get_models(self, provider_config: Dict) -> List[Dict]:
                         "model_type": internal_type,
                         "model_tag": me_type,
                         "max_tokens": DEFAULT_LLM_MAX_TOKENS if internal_type in ("llm", "vlm") else 0,
-                        # ModelEngine models will get base_url and api_key from provider_config (or env)
                         "base_url": host,
                         "api_key": api_key,
                     })

test/backend/services/test_model_health_service.py

@@ -1,5 +1,3 @@
-from consts.exceptions import TimeoutException
-import asyncio
 import os
 import sys
 from unittest import mock
@@ -792,5 +790,3 @@ async def test_embedding_dimension_check_wrapper_value_error():
         mock_logger.error.assert_called_once_with(
             "Error checking embedding dimension: Unsupported model type"
         )
-
-

test/backend/services/test_model_management_service.py

@@ -307,6 +307,7 @@ async def test_create_model_for_tenant_success_llm():
             "base_url": "http://localhost:8000",
             "model_type": "llm",
         }
+        model_data['ssl_verify'] = False
 
         await svc.create_model_for_tenant(user_id, tenant_id, model_data)
 
@@ -316,6 +317,32 @@ async def test_create_model_for_tenant_success_llm():
         assert mock_create.call_count == 1
 
 
+@pytest.mark.asyncio
+async def test_create_model_for_tenant_open_router_disables_ssl():
+    """When base_url contains 'open/router' ssl_verify should be set to False."""
+    svc = import_svc()
+
+    with mock.patch.object(svc, "get_model_by_display_name", return_value=None), \
+            mock.patch.object(svc, "create_model_record") as mock_create, \
+            mock.patch.object(svc, "split_repo_name", return_value=("modelengine", "m")):
+
+        user_id = "u1"
+        tenant_id = "t1"
+        model_data = {
+            "model_name": "modelengine/m",
+            "display_name": None,
+            "base_url": "https://api.example.com/open/router/v1",
+            "model_type": "llm",
+        }
+
+        await svc.create_model_for_tenant(user_id, tenant_id, model_data)
+
+        # Ensure a single record created and ssl_verify was disabled
+        assert mock_create.call_count == 1
+        create_args = mock_create.call_args[0][0]
+        assert create_args["ssl_verify"] is False
+
+
 @pytest.mark.asyncio
 async def test_create_model_for_tenant_conflict_raises():
     svc = import_svc()
@@ -459,7 +486,7 @@ async def test_create_model_for_tenant_multi_embedding_sets_default_chunk_batch(
         mock_dim.assert_awaited_once()
         # Should create two records: multi_embedding and its embedding variant
         assert mock_create.call_count == 2
-        
+
         # Verify chunk_batch was set to 10 for both records
         create_calls = mock_create.call_args_list
         # First call is for multi_embedding
@@ -519,7 +546,7 @@ async def test_batch_create_models_for_tenant_other_provider():
     if not hasattr(svc.ProviderEnum, 'MODELENGINE'):
         modelengine_item = _EnumItem("modelengine")
         svc.ProviderEnum.MODELENGINE = modelengine_item
-    
+
     with mock.patch.object(svc, "get_models_by_tenant_factory_type", return_value=[]), \
             mock.patch.object(svc, "delete_model_record"), \
             mock.patch.object(svc, "split_repo_name", return_value=("openai", "gpt-4")), \
@@ -529,7 +556,7 @@ async def test_batch_create_models_for_tenant_other_provider():
             mock.patch.object(svc, "create_model_record", return_value=True):
 
         await svc.batch_create_models_for_tenant("u1", "t1", batch_payload)
-        
+
         # Verify prepare_model_dict was called with empty model_url for non-Silicon/ModelEngine provider
         call_args = svc.prepare_model_dict.call_args
         assert call_args[1]["model_url"] == ""  # Should be empty for other providers
@@ -618,7 +645,7 @@ def get_by_display(display_name, tenant_id):
         update_calls = [call for call in mock_update.call_args_list if call[0][0] == "id1"]
         if update_calls:
             assert update_calls[0][0][1] == {"max_tokens": 8192}
-        
+
         # Should NOT update model2 (max_tokens same) or model3 (new max_tokens is None)
         # Verify model2 and model3 were not updated
         model2_calls = [call for call in mock_update.call_args_list if call[0][0] == "id2"]

test/backend/services/test_model_provider_service.py

@@ -776,11 +776,9 @@ async def test_modelengine_get_models_llm_success():
     """ModelEngine provider should return LLM models with correct type mapping."""
     from backend.services.model_provider_service import ModelEngineProvider
 
-    provider_config = {"model_type": "llm"}
+    provider_config = {"model_type": "llm", "base_url": "https://model-engine.com", "api_key": "test-key"}
 
-    with mock.patch("backend.services.model_provider_service.MODEL_ENGINE_HOST", "https://model-engine.com"), \
-         mock.patch("backend.services.model_provider_service.MODEL_ENGINE_APIKEY", "test-key"), \
-         mock.patch("backend.services.model_provider_service.aiohttp.ClientSession") as mock_session_class, \
+    with mock.patch("backend.services.model_provider_service.aiohttp.ClientSession") as mock_session_class, \
          mock.patch("backend.services.model_provider_service.aiohttp.ClientTimeout"), \
          mock.patch("backend.services.model_provider_service.aiohttp.TCPConnector"):
 
@@ -825,11 +823,9 @@ async def test_modelengine_get_models_embedding_success():
     """ModelEngine provider should return embedding models with correct type mapping."""
     from backend.services.model_provider_service import ModelEngineProvider
 
-    provider_config = {"model_type": "embedding"}
+    provider_config = {"model_type": "embedding", "base_url": "https://model-engine.com", "api_key": "test-key"}
 
-    with mock.patch("backend.services.model_provider_service.MODEL_ENGINE_HOST", "https://model-engine.com"), \
-         mock.patch("backend.services.model_provider_service.MODEL_ENGINE_APIKEY", "test-key"), \
-         mock.patch("backend.services.model_provider_service.aiohttp.ClientSession") as mock_session_class, \
+    with mock.patch("backend.services.model_provider_service.aiohttp.ClientSession") as mock_session_class, \
          mock.patch("backend.services.model_provider_service.aiohttp.ClientTimeout"), \
          mock.patch("backend.services.model_provider_service.aiohttp.TCPConnector"):
 
@@ -871,11 +867,9 @@ async def test_modelengine_get_models_all_types():
     """ModelEngine provider should return all models when no type filter specified."""
     from backend.services.model_provider_service import ModelEngineProvider
 
-    provider_config = {}  # No model_type filter
+    provider_config = {"base_url": "https://model-engine.com", "api_key": "test-key"}  # No model_type filter
 
-    with mock.patch("backend.services.model_provider_service.MODEL_ENGINE_HOST", "https://model-engine.com"), \
-         mock.patch("backend.services.model_provider_service.MODEL_ENGINE_APIKEY", "test-key"), \
-         mock.patch("backend.services.model_provider_service.aiohttp.ClientSession") as mock_session_class, \
+    with mock.patch("backend.services.model_provider_service.aiohttp.ClientSession") as mock_session_class, \
          mock.patch("backend.services.model_provider_service.aiohttp.ClientTimeout"), \
          mock.patch("backend.services.model_provider_service.aiohttp.TCPConnector"):
 

test/backend/test_model_consts.py

@@ -0,0 +1,30 @@
+import pytest
+from pydantic import ValidationError
+
+from backend.consts import model as model_consts
+
+
+def test_model_connect_status_enum_defaults_and_get_value():
+    assert model_consts.ModelConnectStatusEnum.get_default() == "not_detected"
+    assert model_consts.ModelConnectStatusEnum.get_value("") == "not_detected"
+    assert model_consts.ModelConnectStatusEnum.get_value(None) == "not_detected"
+    assert model_consts.ModelConnectStatusEnum.get_value("available") == "available"
+
+
+def test_model_request_and_validation():
+    # Basic construction
+    mr = model_consts.ModelRequest(model_name="mymodel", model_type="llm")
+    assert mr.model_name == "mymodel"
+    assert mr.model_type == "llm"
+
+    # Chunk create request requires non-empty content
+    with pytest.raises(ValidationError):
+        model_consts.ChunkCreateRequest(content="")
+
+    # Valid chunk create
+    req = model_consts.ChunkCreateRequest(content="a", title="t", filename="f")
+    assert req.content == "a"
+    assert req.title == "t"
+    assert req.filename == "f"
+
+