FEAT: Report the documents a storage listing cannot use (part 1)

pyrit/backend/middleware/error_handlers.py

@@ -15,6 +15,7 @@
 from pyrit.backend.models import FieldError, ProblemDetail
 from pyrit.exceptions.analytics_exception import AnalyticsException
 from pyrit.exceptions.dataset_exception import DatasetAPIException
+from pyrit.registry.file_document_storage import DocumentStorageUnavailableError, DocumentUnreadableError
 
 logger = logging.getLogger(__name__)
 
@@ -192,6 +193,38 @@ async def permission_error_handler(  # pyrit-async-suffix-exempt
             content=problem.model_dump(exclude_none=True),
         )
 
+    @app.exception_handler(DocumentStorageUnavailableError)
+    async def storage_unavailable_handler(  # pyrit-async-suffix-exempt
+        request: Request,
+        exc: DocumentStorageUnavailableError,
+    ) -> JSONResponse:  # pyrit-keyword-only-exempt Starlette exception handler (request, exc)
+        """
+        Handle stored documents that can't be read or written as 503 Service Unavailable.
+
+        The log names the storage location, without credentials, and the reason. It leaves out the traceback,
+        since the error behind it can carry a SAS URL. The response never carries a path or URL.
+
+        Returns:
+            JSONResponse: RFC 7807 problem detail response with 503 status.
+        """
+        logger.warning(f"{exc} ({request.method} {request.url.path})")
+        if isinstance(exc, DocumentUnreadableError):
+            detail = f"{exc.label} '{exc.name}' can't be read: {exc.reason}"
+        else:
+            detail = f"{exc.label} storage is unavailable"
+        problem = ProblemDetail(
+            type="/errors/storage-unavailable",
+            title="Service Unavailable",
+            status=status.HTTP_503_SERVICE_UNAVAILABLE,
+            detail=detail,
+            instance=str(request.url.path),
+        )
+
+        return JSONResponse(
+            status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+            content=problem.model_dump(exclude_none=True),
+        )
+
     @app.exception_handler(NotImplementedError)
     async def not_implemented_handler(  # pyrit-async-suffix-exempt
         request: Request,

pyrit/backend/routes/initializers.py

@@ -14,7 +14,6 @@
     DELETE /api/initializers/{name}         — unregister an initializer
 """
 
-from azure.core.exceptions import AzureError
 from fastapi import APIRouter, Depends, HTTPException, Query, Request, status
 
 from pyrit.backend.middleware.auth import require_admin
@@ -35,19 +34,6 @@
 router = APIRouter(prefix="/initializers", tags=["initializers"])
 
 
-def _custom_storage_unavailable() -> HTTPException:
-    """
-    Create a sanitized response for unavailable custom initializer storage.
-
-    Returns:
-        HTTPException: A service-unavailable response without SDK details.
-    """
-    return HTTPException(
-        status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
-        detail="Custom initializer storage is temporarily unavailable",
-    )
-
-
 def _configured_initializers(request: Request) -> list[ConfiguredInitializerSetting]:
     """
     Read the initializer list captured from ``.pyrit_conf`` at backend startup.
@@ -138,10 +124,7 @@ async def list_custom_initializers(request: Request) -> CustomInitializerListRes
         CustomInitializerListResponse: The configured source and stored scripts.
     """
     _check_custom_initializers_allowed(request)
-    try:
-        return await get_initializer_service().list_custom_initializers_async()
-    except AzureError as exc:
-        raise _custom_storage_unavailable() from exc
+    return await get_initializer_service().list_custom_initializers_async()
 
 
 @router.get(
@@ -205,8 +188,6 @@ async def register_initializer(  # pyrit-async-suffix-exempt
 
     try:
         return await service.register_initializer_async(name=body.name, script_content=body.script_content)
-    except AzureError as exc:
-        raise _custom_storage_unavailable() from exc
     except ValueError as e:
         detail = str(e)
         if "already registered" in detail:
@@ -243,8 +224,6 @@ async def unregister_initializer(  # pyrit-async-suffix-exempt
 
     try:
         await service.unregister_initializer_async(initializer_name=initializer_name)
-    except AzureError as exc:
-        raise _custom_storage_unavailable() from exc
     except ValueError as e:
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,

pyrit/backend/routes/scenario_presets.py

@@ -19,7 +19,6 @@
 
 from typing import Annotated
 
-from azure.core.exceptions import AzureError
 from fastapi import APIRouter, Depends, HTTPException, Query, status
 
 from pyrit.backend.middleware.auth import current_user_name, require_admin
@@ -38,19 +37,6 @@
 router = APIRouter(prefix="/scenario-presets", tags=["scenario-presets"])
 
 
-def _storage_unavailable() -> HTTPException:
-    """
-    Create a sanitized response for unavailable preset storage.
-
-    Returns:
-        HTTPException: A service-unavailable response without SDK details.
-    """
-    return HTTPException(
-        status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
-        detail="Scenario preset storage is temporarily unavailable",
-    )
-
-
 def _not_found(name: str) -> HTTPException:
     """
     Create a not-found response for a missing preset.
@@ -79,8 +65,6 @@ async def _load_preset_or_404_async(name: str) -> ScenarioPresetResponse:
     """
     try:
         preset = await get_scenario_preset_service().get_preset_async(name=name)
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
     except ValueError as exc:
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(exc)) from None
 
@@ -106,10 +90,7 @@ async def list_scenario_presets(  # pyrit-async-suffix-exempt
     Returns:
         ScenarioPresetListResponse: The configured source and the stored presets.
     """
-    try:
-        return await get_scenario_preset_service().list_presets_async(include_estimates=include_estimates)
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
+    return await get_scenario_preset_service().list_presets_async(include_estimates=include_estimates)
 
 
 @router.get(
@@ -177,8 +158,6 @@ async def create_scenario_preset(  # pyrit-async-suffix-exempt
             status_code=status.HTTP_409_CONFLICT,
             detail=f"Scenario preset '{preset.name}' already exists",
         ) from exc
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
 
 
 @router.put(
@@ -239,8 +218,6 @@ async def update_scenario_preset(  # pyrit-async-suffix-exempt
             status_code=status.HTTP_409_CONFLICT,
             detail=f"Scenario preset '{name}' changed since it was read; re-read it and retry",
         ) from exc
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
 
 
 @router.delete(
@@ -278,8 +255,6 @@ async def delete_scenario_preset(  # pyrit-async-suffix-exempt
             status_code=status.HTTP_409_CONFLICT,
             detail=f"Scenario preset '{name}' changed since it was read; re-read it and retry",
         ) from None
-    except (AzureError, TimeoutError) as exc:
-        raise _storage_unavailable() from exc
     except ValueError as exc:
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(exc)) from None
 
@@ -329,8 +304,6 @@ async def resolve_scenario_preset(  # pyrit-async-suffix-exempt
             status_code=status.HTTP_409_CONFLICT,
             detail=f"Scenario preset '{name}' changed since it was read; re-read it and retry",
         ) from None
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
     except ValueError as exc:
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(exc)) from None
 

pyrit/backend/routes/scenario_suites.py

@@ -6,7 +6,6 @@
 from typing import Annotated
 from uuid import UUID
 
-from azure.core.exceptions import AzureError
 from fastapi import APIRouter, Depends, HTTPException, Query, Request, status
 
 from pyrit.backend.middleware.auth import current_user_name, require_admin
@@ -37,19 +36,6 @@
 router = APIRouter(prefix="/scenario-suites", tags=["scenario-suites"])
 
 
-def _storage_unavailable() -> HTTPException:
-    """
-    Create a sanitized response for unavailable suite storage.
-
-    Returns:
-        HTTPException: A service-unavailable response without SDK details.
-    """
-    return HTTPException(
-        status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
-        detail="Scenario suite storage is temporarily unavailable",
-    )
-
-
 def _not_found(name: str) -> HTTPException:
     """
     Create a not-found response for a missing suite.
@@ -72,8 +58,6 @@ async def _load_suite_or_404_async(name: str) -> ScenarioSuiteResponse:
     """
     try:
         suite = await get_scenario_suite_service().get_suite_async(name=name)
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
     except ValueError as exc:
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(exc)) from None
 
@@ -90,10 +74,7 @@ async def list_scenario_suites() -> ScenarioSuiteListResponse:  # pyrit-async-su
     Returns:
         ScenarioSuiteListResponse: Suites from both origins with current entry statuses.
     """
-    try:
-        return await get_scenario_suite_service().list_suites_async()
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
+    return await get_scenario_suite_service().list_suites_async()
 
 
 @router.post("/{name}/readiness", response_model=ScenarioReadiness)
@@ -116,8 +97,6 @@ async def get_scenario_suite_readiness_async(
         raise HTTPException(status_code=409, detail=str(exc)) from None
     except ValueError as exc:
         raise HTTPException(status_code=400, detail=str(exc)) from None
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
 
 
 @router.get(
@@ -211,8 +190,6 @@ async def launch_scenario_suite(  # pyrit-async-suffix-exempt
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(exc)) from None
     except ScenarioSuiteExecutionCapacityError as exc:
         raise HTTPException(status_code=status.HTTP_429_TOO_MANY_REQUESTS, detail=str(exc)) from None
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
     except RuntimeError as exc:
         raise HTTPException(status_code=status.HTTP_503_SERVICE_UNAVAILABLE, detail=str(exc)) from None
 
@@ -269,8 +246,6 @@ async def create_scenario_suite(  # pyrit-async-suffix-exempt
             status_code=status.HTTP_409_CONFLICT,
             detail=f"Scenario suite '{suite.name}' already exists",
         ) from exc
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
 
 
 @router.put(
@@ -319,8 +294,6 @@ async def update_scenario_suite(  # pyrit-async-suffix-exempt
             status_code=status.HTTP_409_CONFLICT,
             detail=f"Scenario suite '{name}' changed since it was read; re-read it and retry",
         ) from exc
-    except AzureError as exc:
-        raise _storage_unavailable() from exc
 
 
 @router.delete(
@@ -350,8 +323,6 @@ async def delete_scenario_suite(  # pyrit-async-suffix-exempt
             status_code=status.HTTP_409_CONFLICT,
             detail=f"Scenario suite '{name}' changed since it was read; re-read it and retry",
         ) from None
-    except (AzureError, TimeoutError) as exc:
-        raise _storage_unavailable() from exc
     except ValueError as exc:
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(exc)) from None
 

pyrit/backend/services/scenario_suite_service.py

@@ -9,8 +9,6 @@
 from functools import lru_cache
 from typing import Literal
 
-from azure.core.exceptions import AzureError
-
 from pyrit.backend.models import (
     ScenarioSuiteEntryStatus,
     ScenarioSuiteListResponse,
@@ -25,6 +23,7 @@
     ScenarioSuiteRegistryEntry,
 )
 from pyrit.registry import ScenarioPresetRegistry, ScenarioSuiteRegistry
+from pyrit.registry.file_document_storage import DocumentStorageUnavailableError
 
 logger = logging.getLogger(__name__)
 
@@ -217,12 +216,9 @@ def _resolve_preset_projection(
                 )
 
             stored_version = registry.get_preset_version(name=preset_name)
-        except (AzureError, OSError):
-            logger.warning(
-                "Could not project scenario preset '%s' because preset storage is unavailable.",
-                preset_name,
-                exc_info=True,
-            )
+        except DocumentStorageUnavailableError as error:
+            # The message leaves out credentials; the error behind it, which a traceback would show, may not.
+            logger.warning(f"Could not project scenario preset '{preset_name}': {error}")
             return _PresetProjection(
                 origin=None,
                 version=None,

pyrit/registry/components/initializer_registry.py

@@ -420,6 +420,7 @@ def register_from_content(self, *, name: str, script_content: str) -> str:
             ValueError: If the source cannot be compiled, does not
                 define a concrete ``PyRITInitializer`` subclass, or
                 *name* collides with an existing entry.
+            DocumentStorageUnavailableError: If the script storage can't be written.
         """
         self._ensure_discovered()
 
@@ -429,11 +430,7 @@ def register_from_content(self, *, name: str, script_content: str) -> str:
             raise ValueError(f"Initializer '{name}' is already registered. Unregister it first to replace it.")
 
         discovered = self._load_custom_initializer_class(name=name, script_content=script_content)
-        storage = self._get_custom_storage()
-        try:
-            storage.save_script(name=name, content=script_content)
-        except OSError as error:
-            raise ValueError(f"Failed to write initializer script: {error}") from error
+        self._get_custom_storage().save_script(name=name, content=script_content)
 
         self.register_class(discovered, name=name)
         logger.info(f"Registered custom initializer: {name} ({discovered.__name__})")
@@ -484,10 +481,12 @@ def unregister_and_cleanup(self, name: str) -> None:
         Raises:
             KeyError: If no stored custom initializer has the requested name.
             ValueError: If the name refers to a built-in initializer.
+            DocumentStorageUnavailableError: If the script storage can't be used.
         """
         self._ensure_discovered()
         storage = self._get_custom_storage()
-        stored_names = storage.list_scripts().keys()
+        # A script that can't be read or decoded is still deleted, so the API can clean it up.
+        stored_names = sorted(storage.list_script_names())
         if name not in stored_names:
             if name in self._builtin_names:
                 raise ValueError(f"Cannot remove built-in initializer '{name}'.")

pyrit/registry/custom_initializer_storage.py

@@ -67,6 +67,16 @@ def list_scripts(self) -> dict[str, str]:
                 logger.warning(f"Skipping stored initializer '{name}': it is not valid UTF-8 text.")
         return scripts
 
+    def list_script_names(self) -> set[str]:
+        """
+        List the names of every stored script, including ones that can't be read or decoded.
+
+        Returns:
+            set[str]: The registry names of the stored scripts.
+        """
+        listing = self._list_documents_with_issues()
+        return set(listing.documents) | {unreadable.name for unreadable in listing.unreadable}
+
     def save_script(self, *, name: str, content: str) -> None:
         """Persist one custom initializer script."""
         self._save_document(name=name, content=content.encode("utf-8"))