Resolve path function conflicts and fix proxy health check

app/api/api.py

@@ -273,7 +273,7 @@ def _get_app(
     add_exception_handlers(app)
 
     instrumentator = Instrumentator(
-        excluded_handlers=["/docs", "/redoc", "/metrics", "/openapi.json", "/favicon.ico", "none"]
+        excluded_handlers=["/docs", "/redoc", "/metrics", "/openapi.json", "/favicon.png", "none"]
     ).instrument(app)
 
     if msd_overwritten is not None:
@@ -309,7 +309,7 @@ async def swagger_doc(req: Request) -> HTMLResponse:
             title="CogStack ModelServe",
             oauth2_redirect_url=oauth2_redirect_url,
             init_oauth=app.swagger_ui_init_oauth,
-            swagger_favicon_url="/static/images/favicon.ico",
+            swagger_favicon_url="/static/images/favicon.png",
             swagger_ui_parameters=app.swagger_ui_parameters,
         )
 
@@ -320,7 +320,7 @@ async def redoc_doc(req: Request) -> HTMLResponse:
         return get_redoc_html(
             openapi_url=openapi_url,
             title="CogStack ModelServe",
-            redoc_favicon_url="/static/images/favicon.ico",
+            redoc_favicon_url="/static/images/favicon.png",
         )
 
     @app.get("/", include_in_schema=False)

app/api/routers/invocation.py

@@ -45,7 +45,7 @@
 PATH_PROCESS_BULK_FILE = "/process_bulk_file"
 PATH_REDACT = "/redact"
 PATH_REDACT_WITH_ENCRYPTION = "/redact_with_encryption"
-PATH_OPENAI_EMBEDDINGS = "/v1/embeddings"
+PATH_EMBEDDINGS = "/embeddings"
 
 router = APIRouter()
 config = get_settings()
@@ -359,13 +359,13 @@ def get_redacted_text_with_encryption(
 
 
 @router.post(
-    PATH_OPENAI_EMBEDDINGS,
-    tags=[Tags.OpenAICompatible.name],
+    PATH_EMBEDDINGS,
+    tags=[Tags.Embeddings.name],
     response_model=None,
     dependencies=[Depends(cms_globals.props.current_active_user)],
-    description="Create embeddings based on text(s), similar to OpenAI's /v1/embeddings endpoint",
+    description="Create embeddings based on text(s)",
 )
-def embed_texts(
+def get_text_embeddings(
     request: Request,
     request_data: Annotated[OpenAIEmbeddingsRequest, Body(
         description="Text(s) to be embedded", media_type="application/json"
@@ -374,7 +374,7 @@ def embed_texts(
     model_service: AbstractModelService = Depends(cms_globals.model_service_dep)
 ) -> JSONResponse:
     """
-    Embeds text or a list of texts, mimicking OpenAI's /v1/embeddings endpoint.
+    Embeds text or a list of texts.
 
     Args:
         request (Request): The request object.

app/domain.py

@@ -24,6 +24,7 @@ class Tags(str, Enum):
     Redaction = "Redact the extracted NER entities"
     Rendering = "Preview embeddable annotation snippet in HTML"
     Training = "Trigger model training on input annotations"
+    Embeddings = "Generate embeddings for input texts"
     Evaluating = "Evaluate the deployed model with trainer export"
     Authentication = "Authenticate registered users"
     Generative = "Generate text based on the input prompt"

app/mcp/server.py

@@ -140,7 +140,7 @@ def create_server() -> Starlette:
                     oauth_manager=oauth_manager,
                     public_paths=[
                         "/authorize",
-                        "/favicon.ico",
+                        "/favicon.png",
                         "/oauth/",
                         "/.well-known",
                         "/.well-known/",

docker-compose-proxy.yml

@@ -33,7 +33,7 @@ services:
       - 28202:28202 # grafana
       - 28203:28203 # graylog
     healthcheck:
-      test: ["CMD", "curl", "-f", "http://localhost/health"]
+      test: ["CMD", "curl", "-f", "-k", "https://localhost/health"]
       interval: 60s
       timeout: 20s
       retries: 3

tests/integration/features/serving.feature

@@ -93,6 +93,14 @@ Feature:
       | /redact_with_encryption | {"text": "Spinal stenosis", "public_key_pem": "-----BEGIN PUBLIC KEY-----\nMIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEA3ITkTP8Tm/5FygcwY2EQ7LgVsuCF0OH7psUqvlXnOPNCfX86CobHBiSFjG9o5ZeajPtTXaf1thUodgpJZVZSqpVTXwGKo8r0COMO87IcwYigkZZgG/WmZgoZART+AA0+JvjFGxflJAxSv7puGlf82E+u5Wz2psLBSDO5qrnmaDZTvPh5eX84cocahVVI7X09/kI+sZiKauM69yoy1bdx16YIIeNm0M9qqS3tTrjouQiJfZ8jUKSZ44Na/81LMVw5O46+5GvwD+OsR43kQ0TexMwgtHxQQsiXLWHCDNy2ZzkzukDYRwA3V2lwVjtQN0WjxHg24BTBDBM+v7iQ7cbweQIDAQAB\n-----END PUBLIC KEY-----"} |  application/json   |
     Then the response should contain encrypted labels
 
+  @embeddings
+  Scenario: Generate embeddings for a single text
+    Given CMS app is up and running
+    When I send a POST request with the following content
+      | endpoint        | data                                                | content_type      |
+      | /embeddings     | {"input": "Spinal stenosis", "model": "cms-model"}  | application/json  |
+    Then the response should contain embeddings
+
   @preview
   Scenario: Extract and preview entities
     Given CMS app is up and running

tests/integration/steps/test_steps.py

@@ -284,3 +284,15 @@ def check_response_annotation_stats(context):
     assert len(response_lines) > 1
     assert "concept,anno_count,anno_unique_counts,anno_ignorance_counts" == response_lines[0]
     context["response"].close()
+
+@then("the response should contain embeddings")
+def check_response_embeddings(context):
+    assert context["response"].status_code == 200
+    assert context["response"].headers["Content-Type"] == "application/json"
+    response_json = context["response"].json()
+    assert "data" in response_json
+    assert isinstance(response_json["data"], list)
+    assert len(response_json["data"]) > 0
+    assert "embedding" in response_json["data"][0]
+    assert isinstance(response_json["data"][0]["embedding"], list)
+    context["response"].close()