Fix CI test collection failure caused by eager KG client initialisation

Move the KG client initialisation from module-level in data_models.py to a
FastAPI lifespan handler in main.py, so test collection in CI does not require
KG credentials, while the service still fails fast on startup if they are missing.

The term_cache population and Enum construction are also made resilient to missing
credentials: get_term_cache() is called at module level but wrapped in a try/except
catching ValueError and AuthenticationError, and all Enum definitions use .get()
so they produce empty Enums rather than raising KeyError when the cache is empty.
In production, credentials are present so the module-level call succeeds and the
cache is fully populated before the server begins serving requests.

Author: apdavison

.gitlab-ci.yml

@@ -12,7 +12,7 @@ run_tests:
   script:
     - cd validation_service_api
     - pip install --no-cache-dir -r requirements.txt.lock
-    - pip install pytest
+    - pip install pytest pytest-asyncio
     - pytest validation_service/tests
   tags:
     - docker-runner

validation_service_api/validation_service/data_models.py

@@ -31,7 +31,6 @@
 from .auth import get_kg_client_for_service_account
 
 
-kg_service_client = get_kg_client_for_service_account()
 term_cache = {}
 
 logger = logging.getLogger("validation_service_api")
@@ -76,7 +75,7 @@ def filter_study_targets(study_targets):
     brain_regions = []
     species = []
     for item in as_list(study_targets):
-        item = item.resolve(kg_service_client, release_status="any")
+        item = item.resolve(get_kg_client_for_service_account(), release_status="any")
         if isinstance(item, omterms.CellType):
             cell_types.append(item.name)
         elif isinstance(item, omterms.UBERONParcellation):
@@ -111,14 +110,18 @@ def get_term_cache():
             omterms.Service,
             omterms.ActionStatusType
         ):
-            objects = cls.list(kg_service_client, api="core", release_status="any", size=10000)
+            objects = cls.list(get_kg_client_for_service_account(), api="core", release_status="any", size=10000)
             term_cache[cls.__name__] = {
                 "names": {obj.name: obj for obj in objects},
                 "ids": {obj.id: obj for obj in objects}
             }
     return term_cache
 
-get_term_cache()
+
+try:
+    get_term_cache()
+except (ValueError, AuthenticationError):
+    pass  # term_cache stays empty; Enums will have no members until the server starts
 
 
 def get_term(cls_name, attr):
@@ -143,7 +146,7 @@ def get_term_name_from_id(cls_name, attr):
     "Species",
     [
         (name.replace(" ", "_"), name)
-        for name in sorted(term_cache["Species"]["names"])
+        for name in sorted(term_cache.get("Species", {}).get("names", {}))
     ]
 )
 
@@ -152,7 +155,7 @@ def get_term_name_from_id(cls_name, attr):
     "BrainRegion",
     [
         (name.replace(" ", "_"), name)
-        for name in sorted(term_cache["UBERONParcellation"]["names"])
+        for name in sorted(term_cache.get("UBERONParcellation", {}).get("names", {}))
     ]
 )
 
@@ -161,7 +164,7 @@ def get_term_name_from_id(cls_name, attr):
     "ModelScope",
     [
         (name.replace(" ", "_").replace(":", "__"), name)
-        for name in sorted(term_cache["ModelScope"]["names"])
+        for name in sorted(term_cache.get("ModelScope", {}).get("names", {}))
     ],
 )
 
@@ -170,7 +173,7 @@ def get_term_name_from_id(cls_name, attr):
     "AbstractionLevel",
     [
         (name.replace(" ", "_").replace(":", "__"), name)
-        for name in sorted(term_cache["ModelAbstractionLevel"]["names"])
+        for name in sorted(term_cache.get("ModelAbstractionLevel", {}).get("names", {}))
     ],
 )
 
@@ -179,7 +182,7 @@ def get_term_name_from_id(cls_name, attr):
     "CellType",
     [
         (name.replace(" ", "_"), name)
-        for name in sorted(term_cache["CellType"]["names"])
+        for name in sorted(term_cache.get("CellType", {}).get("names", {}))
     ]
 )
 
@@ -193,7 +196,7 @@ def get_identifier(iri, prefix):
     "ContentType",
     [
         (get_identifier(obj.uuid, "ct"), obj.name)
-        for obj in sorted(term_cache["ContentType"]["names"].values(), key=lambda obj: obj.name)
+        for obj in sorted(term_cache.get("ContentType", {}).get("names", {}).values(), key=lambda obj: obj.name)
     ]
 )
 
@@ -209,7 +212,7 @@ class ImplementationStatus(str, Enum):
     "RecordingModality",
     [
         (name.replace(" ", "_"), name)
-        for name in sorted(term_cache["Technique"]["names"])
+        for name in sorted(term_cache.get("Technique", {}).get("names", {}))
         # or could use ExperimentalApproach
         # ideally should filter techniques to include only those are recording techniques
     ]
@@ -229,7 +232,7 @@ class ImplementationStatus(str, Enum):
     "ScoreType",
     [
         (name.replace(" ", "_"), name)
-        for name in sorted(term_cache["DifferenceMeasure"]["names"])
+        for name in sorted(term_cache.get("DifferenceMeasure", {}).get("names", {}))
     ]
 )
 
@@ -238,7 +241,7 @@ class ImplementationStatus(str, Enum):
     "License",
     [
         (name.replace(" ", "_"), name)
-        for name in sorted(term_cache["License"]["names"])
+        for name in sorted(term_cache.get("License", {}).get("names", {}))
     ]
 )
 
@@ -247,7 +250,7 @@ class ImplementationStatus(str, Enum):
     "ActionStatusType",
     [
         (name.replace(" ", "_"), name)
-        for name in sorted(term_cache["ActionStatusType"]["names"])
+        for name in sorted(term_cache.get("ActionStatusType", {}).get("names", {}))
     ]
 )
 

validation_service_api/validation_service/main.py

@@ -1,9 +1,12 @@
+from contextlib import asynccontextmanager
+
 from fastapi import FastAPI
 from starlette.middleware.sessions import SessionMiddleware
 from starlette.middleware.cors import CORSMiddleware
 
 from .resources import models, tests, vocab, results, auth, comments
 from . import settings
+from .auth import get_kg_client_for_service_account
 
 
 description = """
@@ -31,7 +34,13 @@
     description = warning_message + description
 
 
-app = FastAPI(title="EBRAINS Model Validation Service", description=description, version="3beta")
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    get_kg_client_for_service_account()  # fail fast if credentials are missing
+    yield
+
+
+app = FastAPI(title="EBRAINS Model Validation Service", description=description, version="3beta", lifespan=lifespan)
 
 app.add_middleware(
     SessionMiddleware,