chore: upgrade to Python 3.11

aidial_adapter_openai/audio_api/transcribe/adapter.py

@@ -43,7 +43,7 @@ def _get_usage(
 ) -> TokenUsage | None:
     # NOTE: whisper has completely different API for its responses
     duration: Any | None = getattr(chunk, "duration", None)
-    if duration is not None and isinstance(duration, (float, int)):
+    if duration is not None and isinstance(duration, float | int):
         return TokenUsage(prompt_tokens=int(duration))
 
     usage_dict: dict | None = getattr(chunk, "usage", None)  # type: ignore

aidial_adapter_openai/audio_api/transcribe/prompt.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 import mimetypes
-from typing import Any, List
+from typing import Any
 
 from aidial_sdk.exceptions import RequestValidationError
 from pydantic import BaseModel
@@ -50,7 +50,7 @@ async def from_request(
 
         system_message = _collect_system_messages(messages)
 
-        audios: List[FileResource] = []
+        audios: list[FileResource] = []
 
         for message in result:
             for file in message.files:

aidial_adapter_openai/chat_completions/gpt.py

@@ -1,4 +1,4 @@
-from typing import AsyncIterator, Callable, Coroutine, List, Mapping, Tuple
+from collections.abc import AsyncIterator, Callable, Coroutine, Mapping
 
 from aidial_sdk.exceptions import InvalidRequestError
 from openai import AsyncAzureOpenAI, AsyncOpenAI, AsyncStream
@@ -29,10 +29,10 @@
 
 async def multi_modal_truncate_prompt(
     request: dict,
-    messages: List[MultiModalMessage],
+    messages: list[MultiModalMessage],
     max_prompt_tokens: int,
     tokenizer: Tokenizer,
-) -> Tuple[List[MultiModalMessage], DiscardedMessages, TruncatedTokens]:
+) -> tuple[list[MultiModalMessage], DiscardedMessages, TruncatedTokens]:
     return await truncate_prompt(
         messages=messages,
         message_tokens=tokenizer.tokenize_request_message,
@@ -62,9 +62,9 @@ def _extract_max_prompt_tokens(request: dict) -> int | None:
 
 
 async def _truncate_messages(
-    request: dict, messages: List[MultiModalMessage], tokenizer: Tokenizer
-) -> Tuple[
-    List[MultiModalMessage],
+    request: dict, messages: list[MultiModalMessage], tokenizer: Tokenizer
+) -> tuple[
+    list[MultiModalMessage],
     DiscardedMessages | None,
     Callable[[], Coroutine[None, None, TruncatedTokens]],
 ]:
@@ -109,7 +109,7 @@ async def chat_completion(
     eliminate_empty_choices: bool,
 ) -> ResponseWithHeaders[AsyncIterator[dict] | dict]:
     n: int = request.get("n") or 1
-    messages: List[dict] = request["messages"]
+    messages: list[dict] = request["messages"]
     model_name = request["model"]
 
     multi_modal_messages = await ResourceProcessor(

aidial_adapter_openai/chat_completions/gpt_audio.py

@@ -6,7 +6,8 @@
 DIAL-compatible format with attachments and stages.
 """
 
-from typing import AsyncIterator, TypeVar
+from collections.abc import AsyncIterator
+from typing import TypeVar
 
 from pydantic import BaseModel
 

aidial_adapter_openai/chat_completions/gpt_oss.py

@@ -18,15 +18,16 @@
 https://cookbook.openai.com/articles/gpt-oss/handle-raw-cot
 """
 
-from typing import AsyncIterator, Set, TypeVar
+from collections.abc import AsyncIterator
+from typing import TypeVar
 
 from pydantic import BaseModel
 
 from aidial_adapter_openai.utils.streaming import map_stream
 
 
 class _ResponseTransformer(BaseModel):
-    opened_reasoning_stages: Set[int] = set()
+    opened_reasoning_stages: set[int] = set()
     """Indices of choices where a reasoning stage was open.
     """
 

aidial_adapter_openai/chat_completions/non_gpt.py

@@ -1,4 +1,4 @@
-from typing import AsyncIterator
+from collections.abc import AsyncIterator
 
 from openai import AsyncAzureOpenAI, AsyncOpenAI, AsyncStream
 from openai.types.chat.chat_completion import ChatCompletion

aidial_adapter_openai/chat_completions/transformation.py

@@ -1,5 +1,5 @@
 from dataclasses import dataclass
-from typing import List, Set, assert_never
+from typing import assert_never
 
 from aidial_sdk.exceptions import InvalidRequestError
 from openai.types.chat import (
@@ -46,15 +46,15 @@ class Error:
 
 class MessageTransformer:
     file_storage: FileStorage | None
-    errors: Set[Error]
-    images: List[ImageResource]
-    files: List[FileResource]
+    errors: set[Error]
+    images: list[ImageResource]
+    files: list[FileResource]
 
     def __init__(
         self,
         *,
         file_storage: FileStorage | None,
-        errors: Set[Error] | None = None,
+        errors: set[Error] | None = None,
     ):
         self.file_storage = file_storage
         self.errors = set() if errors is None else errors
@@ -81,12 +81,12 @@ async def try_download_resource(
             return None
 
     async def download_attachments(
-        self, attachments: List[dict]
-    ) -> List[ChatCompletionContentPartImageParam | File]:
+        self, attachments: list[dict]
+    ) -> list[ChatCompletionContentPartImageParam | File]:
         if attachments:
             logger.debug(f"original attachments: {attachments}")
 
-        ret: List[ChatCompletionContentPartImageParam | File] = []
+        ret: list[ChatCompletionContentPartImageParam | File] = []
         for attachment in attachments:
             if result := await self.download_attachment(attachment):
                 ret.append(result)
@@ -145,15 +145,15 @@ async def download_content(
         self,
         content: (
             str
-            | List[ChatCompletionContentPartParam | ContentArrayOfContentPart]
+            | list[ChatCompletionContentPartParam | ContentArrayOfContentPart]
         ),
-    ) -> List[ChatCompletionContentPartParam | ContentArrayOfContentPart]:
+    ) -> list[ChatCompletionContentPartParam | ContentArrayOfContentPart]:
         if isinstance(content, str):
             parts = [create_text_content_part(content)]
         else:
             parts = content
 
-        ret: List[
+        ret: list[
             ChatCompletionContentPartParam | ContentArrayOfContentPart
         ] = []
         for part in parts:
@@ -192,9 +192,9 @@ class ResourceProcessor(BaseModel):
     file_storage: FileStorage | None
 
     async def transform_messages(
-        self, messages: List[dict]
-    ) -> List[MultiModalMessage]:
-        errors: Set[Error] = set()
+        self, messages: list[dict]
+    ) -> list[MultiModalMessage]:
+        errors: set[Error] = set()
         transformations = [
             await MessageTransformer(
                 file_storage=self.file_storage, errors=errors

aidial_adapter_openai/completions.py

@@ -1,4 +1,5 @@
-from typing import Any, AsyncIterator, Dict
+from collections.abc import AsyncIterator
+from typing import Any
 
 from aidial_sdk.exceptions import RequestValidationError
 from openai import AsyncAzureOpenAI, AsyncOpenAI, AsyncStream
@@ -18,7 +19,7 @@ def sanitize_text(text: str) -> str:
 
 def convert_to_chat_completions_response(
     chunk: Completion, is_stream: bool
-) -> Dict[str, Any]:
+) -> dict[str, Any]:
     converted_chunk = build_chunk(
         id=chunk.id,
         model=chunk.model,
@@ -37,7 +38,7 @@ def convert_to_chat_completions_response(
 
 async def chat_completion(
     *,
-    request: Dict[str, Any],
+    request: dict[str, Any],
     client: AsyncAzureOpenAI | AsyncOpenAI,
     prompt_template: str | None,
 ) -> AsyncIterator[dict] | dict:

aidial_adapter_openai/configuration/app_config.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 import os
-from typing import Dict, List, assert_never
+from typing import assert_never
 
 from aidial_sdk.exceptions import InternalServerError
 
@@ -39,23 +39,23 @@ class DeploymentAPIType(ExtraForbidModel):
 
 
 class ApplicationConfig(ExtraForbidModel):
-    TIKTOKEN_MODEL_MAPPING: Dict[str, str] = {}
+    TIKTOKEN_MODEL_MAPPING: dict[str, str] = {}
 
-    DALLE3_DEPLOYMENTS: List[str] = []
+    DALLE3_DEPLOYMENTS: list[str] = []
     DALLE3_AZURE_API_VERSION: str = "2024-02-01"
 
-    GPT_IMAGE_1_DEPLOYMENTS: List[str] = []
+    GPT_IMAGE_1_DEPLOYMENTS: list[str] = []
     GPT_IMAGE_1_AZURE_API_VERSION: str = "2025-04-01-preview"
 
-    MISTRAL_DEPLOYMENTS: List[str] = []
-    DATABRICKS_DEPLOYMENTS: List[str] = []
-    GPT4O_DEPLOYMENTS: List[str] = []
-    GPT4O_MINI_DEPLOYMENTS: List[str] = []
-    AZURE_AI_VISION_DEPLOYMENTS: List[str] = []
+    MISTRAL_DEPLOYMENTS: list[str] = []
+    DATABRICKS_DEPLOYMENTS: list[str] = []
+    GPT4O_DEPLOYMENTS: list[str] = []
+    GPT4O_MINI_DEPLOYMENTS: list[str] = []
+    AZURE_AI_VISION_DEPLOYMENTS: list[str] = []
 
-    API_VERSIONS_MAPPING: Dict[str, str] = {}
-    COMPLETION_DEPLOYMENTS_PROMPT_TEMPLATES: Dict[str, str] = {}
-    NON_STREAMING_DEPLOYMENTS: List[str] = []
+    API_VERSIONS_MAPPING: dict[str, str] = {}
+    COMPLETION_DEPLOYMENTS_PROMPT_TEMPLATES: dict[str, str] = {}
+    NON_STREAMING_DEPLOYMENTS: list[str] = []
     ELIMINATE_EMPTY_CHOICES: bool = False
 
     AUDIO_AZURE_API_VERSION: str = "2025-03-01-preview"

aidial_adapter_openai/dial_api/embedding_inputs.py

@@ -1,8 +1,5 @@
+from collections.abc import AsyncIterator, Callable, Coroutine
 from typing import (
-    AsyncIterator,
-    Callable,
-    Coroutine,
-    List,
     TypeVar,
     assert_never,
     cast,
@@ -15,7 +12,7 @@
 _T = TypeVar("_T")
 
 _Coro = Coroutine[None, None, _T]
-_Tokens = List[int]
+_Tokens = list[int]
 
 
 async def reject_tokens(tokens: _Tokens):
@@ -25,7 +22,7 @@ async def reject_tokens(tokens: _Tokens):
     )
 
 
-async def reject_mixed(input: List[str | Attachment]):
+async def reject_mixed(input: list[str | Attachment]):
     raise RequestValidationError(
         "Embedding inputs composed of multiple texts and/or attachments aren't supported"
     )
@@ -37,7 +34,7 @@ async def collect_embedding_inputs(
     on_text: Callable[[str], _Coro[_T]],
     on_attachment: Callable[[Attachment], _Coro[_T]],
     on_tokens: Callable[[_Tokens], _Coro[_T]] = reject_tokens,
-    on_mixed: Callable[[List[str | Attachment]], _Coro[_T]] = reject_mixed,
+    on_mixed: Callable[[list[str | Attachment]], _Coro[_T]] = reject_mixed,
 ) -> AsyncIterator[_T]:
     async def _on_str_or_attachment(input: str | Attachment) -> _T:
         if isinstance(input, str):
@@ -70,7 +67,7 @@ async def _on_str_or_attachment(input: str | Attachment) -> _T:
         return
 
     for input in request.custom_input:
-        if isinstance(input, (str, Attachment)):
+        if isinstance(input, str | Attachment):
             yield await _on_str_or_attachment(input)
         elif isinstance(input, list):
             if len(input) == 0:

aidial_adapter_openai/dial_api/request.py

@@ -1,12 +1,12 @@
-from typing import Any, Type, TypeVar
+from typing import Any, TypeVar
 
 from aidial_sdk.exceptions import RequestValidationError
 from pydantic import BaseModel, ValidationError
 
 _T = TypeVar("_T", bound=BaseModel)
 
 
-def parse_configuration(cls: Type[_T], data: Any) -> _T | None:
+def parse_configuration(cls: type[_T], data: Any) -> _T | None:
     if (cf := data.get("custom_fields")) is None:
         return None
 

aidial_adapter_openai/dial_api/resource.py

@@ -1,7 +1,6 @@
 import base64
 import mimetypes
 from abc import ABC, abstractmethod
-from typing import List
 
 from aidial_sdk.chat_completion import Attachment
 from pydantic import BaseModel, model_validator
@@ -25,17 +24,17 @@ class MissingContentTypeError(ValidationError):
 
 class UnsupportedContentTypeError(ValidationError):
     type: str
-    supported_types: List[str]
+    supported_types: list[str]
 
-    def __init__(self, *, message: str, type: str, supported_types: List[str]):
+    def __init__(self, *, message: str, type: str, supported_types: list[str]):
         self.type = type
         self.supported_types = supported_types
         super().__init__(message)
 
 
 class DialResource(ABC, BaseModel):
     entity_name: str | None = None
-    supported_types: List[str] | None = None
+    supported_types: list[str] | None = None
 
     @abstractmethod
     async def download(self, storage: FileStorage | None) -> Resource: ...

aidial_adapter_openai/dial_api/sdk_adapter.py

@@ -1,4 +1,4 @@
-from typing import Callable, Coroutine
+from collections.abc import Callable, Coroutine
 
 import fastapi
 from aidial_sdk.chat_completion import Request as DIALRequest

aidial_adapter_openai/dial_api/storage.py

@@ -1,7 +1,7 @@
 import base64
 import hashlib
 import mimetypes
-from typing import Mapping, Optional
+from collections.abc import Mapping
 from urllib.parse import unquote, urljoin
 
 from pydantic import BaseModel, SecretStr
@@ -28,7 +28,7 @@ class FileStorage(BaseModel):
     dial_url: str
     api_key: SecretStr
 
-    bucket: Optional[Bucket] = None
+    bucket: Bucket | None = None
 
     @property
     def headers(self) -> Mapping[str, str]:
@@ -129,14 +129,14 @@ def _compute_hash_digest(file_content: str | bytes) -> str:
 
 DIAL_USE_FILE_STORAGE = get_env_bool("DIAL_USE_FILE_STORAGE", False)
 
-DIAL_URL: Optional[str] = None
+DIAL_URL: str | None = None
 if DIAL_USE_FILE_STORAGE:
     DIAL_URL = get_env(
         "DIAL_URL", "DIAL_URL must be set to use the DIAL file storage"
     )
 
 
-def create_file_storage(headers: Mapping[str, str]) -> Optional[FileStorage]:
+def create_file_storage(headers: Mapping[str, str]) -> FileStorage | None:
     if not DIAL_USE_FILE_STORAGE or DIAL_URL is None:
         return None