aidial-sdk 0.47.0.dev8__tar.gz → 0.47.0.dev9__tar.gz

{aidial_sdk-0.47.0.dev8 → aidial_sdk-0.47.0.dev9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: aidial-sdk
-Version: 0.47.0.dev8
+Version: 0.47.0.dev9
 Summary: Framework to create applications and model adapters for AI DIAL
 License-Expression: Apache-2.0
 License-File: LICENSE

{aidial_sdk-0.47.0.dev8 → aidial_sdk-0.47.0.dev9}/aidial_sdk/application.py

@@ -24,7 +24,7 @@ from aidial_sdk.chat_completion.response import (
 from aidial_sdk.deployment.configuration import ConfigurationRequest
 from aidial_sdk.deployment.from_request_mixin import (
     FromRequestMixin,
-    resolve_deployment_id,
+    get_upstream_deployment_id,
 )
 from aidial_sdk.deployment.rate import RateRequest
 from aidial_sdk.deployment.tokenize import TokenizeRequest
@@ -288,7 +288,7 @@ class DIALApp(FastAPI):
 
     def _v1_dispatcher(self, endpoint: str) -> Handler:
         async def _handler(original_request: Request) -> Response:
-            deployment_id = resolve_deployment_id(
+            deployment_id = get_upstream_deployment_id(
                 original_request.headers, None
             )
             set_log_deployment(deployment_id)

{aidial_sdk-0.47.0.dev8 → aidial_sdk-0.47.0.dev9}/aidial_sdk/chat_completion/response.py

@@ -27,6 +27,7 @@ from aidial_sdk.deployment._headers import (
     DIAL_CACHE_EXPIRE_AT,
     DIAL_CACHE_EXTRA_METADATA,
 )
+from aidial_sdk.deployment.from_request_mixin import get_client_deployment_id
 from aidial_sdk.exceptions import HTTPException as DIALException
 from aidial_sdk.exceptions import RequestValidationError, RuntimeServerError
 from aidial_sdk.utils._cancel_scope import CancelScope
@@ -69,6 +70,7 @@ class Response:
                 if self.request.stream
                 else "chat.completion"
             ),
+            model=get_client_deployment_id(request.headers),
         )
 
         self._headers = []

{aidial_sdk-0.47.0.dev8 → aidial_sdk-0.47.0.dev9}/aidial_sdk/deployment/from_request_mixin.py

@@ -36,24 +36,27 @@ from aidial_sdk.utils.pydantic import ExtraAllowModel
 T = TypeVar("T", bound="FromRequestMixin")
 
 
-def resolve_deployment_id(
+def get_upstream_deployment_id(
     headers: Mapping[str, str], deployment_id: str | None
 ) -> str:
-    """The deployment id of a request comes either from the path it was
-    sent to or, for the deployment-agnostic paths, from its headers."""
+    """The upstream deployment id that a DIAL application is expected to call."""
 
-    effective_deployment_id = (
+    return (
         headers.get(DIAL_OVERRIDE_NAME)
         or deployment_id
-        or headers.get(DIAL_DEPLOYMENT_ID)
+        or get_client_deployment_id(headers)
     )
 
-    if effective_deployment_id is None:
+
+def get_client_deployment_id(headers: Mapping[str, str]) -> str:
+    """The client-facing deployment id that a DIAL application is expected
+    to return as a model parameter in the response."""
+
+    if (deployment_id := headers.get(DIAL_DEPLOYMENT_ID)) is None:
         raise InternalServerError(
             f"The request headers are missing {DIAL_DEPLOYMENT_ID} header."
         )
-
-    return effective_deployment_id
+    return deployment_id
 
 
 class FromRequestMixin(ABC, ExtraAllowModel):
@@ -235,7 +238,7 @@ class FromRequestDeploymentMixin(FromRequestMixin):
             bearer_token_secret=(
                 SecretStr(bearer_token) if bearer_token else None
             ),
-            deployment_id=resolve_deployment_id(headers, deployment_id),
+            deployment_id=get_upstream_deployment_id(headers, deployment_id),
             api_version=request.query_params.get("api-version"),
             headers=headers,
             original_request=request,

{aidial_sdk-0.47.0.dev8 → aidial_sdk-0.47.0.dev9}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "aidial-sdk"
-version = "0.47.0.dev8"
+version = "0.47.0.dev9"
 description = "Framework to create applications and model adapters for AI DIAL"
 authors = [{ name = "EPAM RAIL", email = "SpecialEPM-DIALDevTeam@epam.com" }]
 license = "Apache-2.0"