mirascope 2.0.0a2__py3-none-any.whl → 2.0.0a4__py3-none-any.whl

mirascope/llm/{clients/base/client.py → providers/base/base_provider.py}

@@ -4,10 +4,8 @@ from __future__ import annotations
 
 from abc import ABC, abstractmethod
 from collections.abc import Sequence
-from contextvars import ContextVar, Token
-from types import TracebackType
-from typing import Generic, overload
-from typing_extensions import Self, TypeVar, Unpack
+from typing import TYPE_CHECKING, Any, ClassVar, Generic, TypeAlias, overload
+from typing_extensions import TypeVar, Unpack
 
 from ...context import Context, DepsT
 from ...formatting import Format, FormattableT
@@ -34,51 +32,40 @@ from ...tools import (
 )
 from .params import Params
 
-ModelIdT = TypeVar("ModelIdT", bound=str)
+if TYPE_CHECKING:
+    from ..provider_id import ProviderId
+
 ProviderClientT = TypeVar("ProviderClientT")
 
-ClientT = TypeVar("ClientT", bound="BaseClient")
-"""Type variable for an LLM client."""
+Provider: TypeAlias = "BaseProvider[Any]"
+"""Type alias for `BaseProvider` with any client type."""
 
 
-class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
-    """Base abstract client for provider-specific implementations.
+class BaseProvider(Generic[ProviderClientT], ABC):
+    """Base abstract provider for LLM interactions.
 
     This class defines explicit methods for each type of call, eliminating
     the need for complex overloads in provider implementations.
     """
 
-    client: ProviderClientT
-    _token: Token | None = None
+    id: ClassVar[ProviderId]
+    """Provider identifier (e.g., "anthropic", "openai")."""
 
-    @property
-    @abstractmethod
-    def _context_var(self) -> ContextVar:
-        """The ContextVar for this client type."""
-        ...
+    default_scope: ClassVar[str | list[str]]
+    """Default scope(s) for this provider when explicitly registered.
 
-    def __enter__(self) -> Self:
-        """Sets the client context and stores the token."""
-        self._token = self._context_var.set(self)
-        return self
+    Can be a single scope string or a list of scopes. For example:
+    - "anthropic/" - Single scope
+    - ["anthropic/", "openai/"] - Multiple scopes (e.g., for AWS Bedrock)
+    """
 
-    def __exit__(
-        self,
-        exc_type: type[BaseException] | None,
-        exc_val: BaseException | None,
-        exc_tb: TracebackType | None,
-    ) -> None:
-        """Restores the client context to the token from the last setting."""
-        if self._token is not None:
-            self._context_var.reset(self._token)
-            self._token = None
+    client: ProviderClientT
 
     @overload
-    @abstractmethod
     def call(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool] | Toolkit | None = None,
         format: None = None,
@@ -88,11 +75,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     def call(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool] | Toolkit | None = None,
         format: type[FormattableT] | Format[FormattableT],
@@ -102,11 +88,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     def call(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool] | Toolkit | None = None,
         format: type[FormattableT] | Format[FormattableT] | None,
@@ -115,11 +100,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         """Generate an `llm.Response` with an optional response format."""
         ...
 
-    @abstractmethod
     def call(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool] | Toolkit | None = None,
         format: type[FormattableT] | Format[FormattableT] | None = None,
@@ -137,15 +121,33 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         Returns:
             An `llm.Response` object containing the LLM-generated content.
         """
+        return self._call(
+            model_id=model_id,
+            messages=messages,
+            tools=tools,
+            format=format,
+            **params,
+        )
+
+    @abstractmethod
+    def _call(
+        self,
+        *,
+        model_id: str,
+        messages: Sequence[Message],
+        tools: Sequence[Tool] | Toolkit | None = None,
+        format: type[FormattableT] | Format[FormattableT] | None = None,
+        **params: Unpack[Params],
+    ) -> Response | Response[FormattableT]:
+        """Implementation for call(). Subclasses override this method."""
         ...
 
     @overload
-    @abstractmethod
     def context_call(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool | ContextTool[DepsT]]
         | ContextToolkit[DepsT]
@@ -157,12 +159,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     def context_call(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool | ContextTool[DepsT]]
         | ContextToolkit[DepsT]
@@ -174,12 +175,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     def context_call(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool | ContextTool[DepsT]]
         | ContextToolkit[DepsT]
@@ -190,12 +190,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         """Generate an `llm.ContextResponse` with an optional response format."""
         ...
 
-    @abstractmethod
     def context_call(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool | ContextTool[DepsT]]
         | ContextToolkit[DepsT]
@@ -216,14 +215,36 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         Returns:
             An `llm.ContextResponse` object containing the LLM-generated content.
         """
+        return self._context_call(
+            ctx=ctx,
+            model_id=model_id,
+            messages=messages,
+            tools=tools,
+            format=format,
+            **params,
+        )
+
+    @abstractmethod
+    def _context_call(
+        self,
+        *,
+        ctx: Context[DepsT],
+        model_id: str,
+        messages: Sequence[Message],
+        tools: Sequence[Tool | ContextTool[DepsT]]
+        | ContextToolkit[DepsT]
+        | None = None,
+        format: type[FormattableT] | Format[FormattableT] | None = None,
+        **params: Unpack[Params],
+    ) -> ContextResponse[DepsT, None] | ContextResponse[DepsT, FormattableT]:
+        """Implementation for context_call(). Subclasses override this method."""
         ...
 
     @overload
-    @abstractmethod
     async def call_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
         format: None = None,
@@ -233,11 +254,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     async def call_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
         format: type[FormattableT] | Format[FormattableT],
@@ -247,11 +267,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     async def call_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
         format: type[FormattableT] | Format[FormattableT] | None,
@@ -260,11 +279,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         """Generate an `llm.AsyncResponse` with an optional response format."""
         ...
 
-    @abstractmethod
     async def call_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
         format: type[FormattableT] | Format[FormattableT] | None = None,
@@ -282,15 +300,33 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         Returns:
             An `llm.AsyncResponse` object containing the LLM-generated content.
         """
+        return await self._call_async(
+            model_id=model_id,
+            messages=messages,
+            tools=tools,
+            format=format,
+            **params,
+        )
+
+    @abstractmethod
+    async def _call_async(
+        self,
+        *,
+        model_id: str,
+        messages: Sequence[Message],
+        tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
+        format: type[FormattableT] | Format[FormattableT] | None = None,
+        **params: Unpack[Params],
+    ) -> AsyncResponse | AsyncResponse[FormattableT]:
+        """Implementation for call_async(). Subclasses override this method."""
         ...
 
     @overload
-    @abstractmethod
     async def context_call_async(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
         | AsyncContextToolkit[DepsT]
@@ -302,12 +338,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     async def context_call_async(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
         | AsyncContextToolkit[DepsT]
@@ -319,12 +354,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     async def context_call_async(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
         | AsyncContextToolkit[DepsT]
@@ -335,12 +369,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         """Generate an `llm.AsyncContextResponse` with an optional response format."""
         ...
 
-    @abstractmethod
     async def context_call_async(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
         | AsyncContextToolkit[DepsT]
@@ -361,14 +394,36 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         Returns:
             An `llm.AsyncContextResponse` object containing the LLM-generated content.
         """
+        return await self._context_call_async(
+            ctx=ctx,
+            model_id=model_id,
+            messages=messages,
+            tools=tools,
+            format=format,
+            **params,
+        )
+
+    @abstractmethod
+    async def _context_call_async(
+        self,
+        *,
+        ctx: Context[DepsT],
+        model_id: str,
+        messages: Sequence[Message],
+        tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
+        | AsyncContextToolkit[DepsT]
+        | None = None,
+        format: type[FormattableT] | Format[FormattableT] | None = None,
+        **params: Unpack[Params],
+    ) -> AsyncContextResponse[DepsT, None] | AsyncContextResponse[DepsT, FormattableT]:
+        """Implementation for context_call_async(). Subclasses override this method."""
         ...
 
     @overload
-    @abstractmethod
     def stream(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool] | Toolkit | None = None,
         format: None = None,
@@ -378,11 +433,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     def stream(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool] | Toolkit | None = None,
         format: type[FormattableT] | Format[FormattableT],
@@ -392,11 +446,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     def stream(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool] | Toolkit | None = None,
         format: type[FormattableT] | Format[FormattableT] | None,
@@ -405,11 +458,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         """Stream an `llm.StreamResponse` with an optional response format."""
         ...
 
-    @abstractmethod
     def stream(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool] | Toolkit | None = None,
         format: type[FormattableT] | Format[FormattableT] | None = None,
@@ -427,15 +479,33 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         Returns:
             An `llm.StreamResponse` object for iterating over the LLM-generated content.
         """
+        return self._stream(
+            model_id=model_id,
+            messages=messages,
+            tools=tools,
+            format=format,
+            **params,
+        )
+
+    @abstractmethod
+    def _stream(
+        self,
+        *,
+        model_id: str,
+        messages: Sequence[Message],
+        tools: Sequence[Tool] | Toolkit | None = None,
+        format: type[FormattableT] | Format[FormattableT] | None = None,
+        **params: Unpack[Params],
+    ) -> StreamResponse | StreamResponse[FormattableT]:
+        """Implementation for stream(). Subclasses override this method."""
         ...
 
     @overload
-    @abstractmethod
     def context_stream(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool | ContextTool[DepsT]]
         | ContextToolkit[DepsT]
@@ -447,12 +517,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     def context_stream(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool | ContextTool[DepsT]]
         | ContextToolkit[DepsT]
@@ -464,12 +533,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     def context_stream(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool | ContextTool[DepsT]]
         | ContextToolkit[DepsT]
@@ -482,12 +550,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         """Stream an `llm.ContextStreamResponse` with an optional response format."""
         ...
 
-    @abstractmethod
     def context_stream(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[Tool | ContextTool[DepsT]]
         | ContextToolkit[DepsT]
@@ -510,14 +577,38 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         Returns:
             An `llm.ContextStreamResponse` object for iterating over the LLM-generated content.
         """
+        return self._context_stream(
+            ctx=ctx,
+            model_id=model_id,
+            messages=messages,
+            tools=tools,
+            format=format,
+            **params,
+        )
+
+    @abstractmethod
+    def _context_stream(
+        self,
+        *,
+        ctx: Context[DepsT],
+        model_id: str,
+        messages: Sequence[Message],
+        tools: Sequence[Tool | ContextTool[DepsT]]
+        | ContextToolkit[DepsT]
+        | None = None,
+        format: type[FormattableT] | Format[FormattableT] | None = None,
+        **params: Unpack[Params],
+    ) -> (
+        ContextStreamResponse[DepsT, None] | ContextStreamResponse[DepsT, FormattableT]
+    ):
+        """Implementation for context_stream(). Subclasses override this method."""
         ...
 
     @overload
-    @abstractmethod
     async def stream_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
         format: None = None,
@@ -527,11 +618,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     async def stream_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
         format: type[FormattableT] | Format[FormattableT],
@@ -541,11 +631,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     async def stream_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
         format: type[FormattableT] | Format[FormattableT] | None,
@@ -554,11 +643,10 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         """Stream an `llm.AsyncStreamResponse` with an optional response format."""
         ...
 
-    @abstractmethod
     async def stream_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
         format: type[FormattableT] | Format[FormattableT] | None = None,
@@ -576,15 +664,33 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         Returns:
             An `llm.AsyncStreamResponse` object for asynchronously iterating over the LLM-generated content.
         """
+        return await self._stream_async(
+            model_id=model_id,
+            messages=messages,
+            tools=tools,
+            format=format,
+            **params,
+        )
+
+    @abstractmethod
+    async def _stream_async(
+        self,
+        *,
+        model_id: str,
+        messages: Sequence[Message],
+        tools: Sequence[AsyncTool] | AsyncToolkit | None = None,
+        format: type[FormattableT] | Format[FormattableT] | None = None,
+        **params: Unpack[Params],
+    ) -> AsyncStreamResponse | AsyncStreamResponse[FormattableT]:
+        """Implementation for stream_async(). Subclasses override this method."""
         ...
 
     @overload
-    @abstractmethod
     async def context_stream_async(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
         | AsyncContextToolkit[DepsT]
@@ -596,12 +702,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     async def context_stream_async(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
         | AsyncContextToolkit[DepsT]
@@ -613,12 +718,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         ...
 
     @overload
-    @abstractmethod
     async def context_stream_async(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
         | AsyncContextToolkit[DepsT]
@@ -632,12 +736,11 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         """Stream an `llm.AsyncContextStreamResponse` with an optional response format."""
         ...
 
-    @abstractmethod
     async def context_stream_async(
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         messages: Sequence[Message],
         tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
         | AsyncContextToolkit[DepsT]
@@ -661,13 +764,39 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         Returns:
             An `llm.AsyncContextStreamResponse` object for asynchronously iterating over the LLM-generated content.
         """
+        return await self._context_stream_async(
+            ctx=ctx,
+            model_id=model_id,
+            messages=messages,
+            tools=tools,
+            format=format,
+            **params,
+        )
+
+    @abstractmethod
+    async def _context_stream_async(
+        self,
+        *,
+        ctx: Context[DepsT],
+        model_id: str,
+        messages: Sequence[Message],
+        tools: Sequence[AsyncTool | AsyncContextTool[DepsT]]
+        | AsyncContextToolkit[DepsT]
+        | None = None,
+        format: type[FormattableT] | Format[FormattableT] | None = None,
+        **params: Unpack[Params],
+    ) -> (
+        AsyncContextStreamResponse[DepsT, None]
+        | AsyncContextStreamResponse[DepsT, FormattableT]
+    ):
+        """Implementation for context_stream_async(). Subclasses override this method."""
         ...
 
     @overload
     def resume(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: Response,
         content: UserContent,
         **params: Unpack[Params],
@@ -679,7 +808,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     def resume(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: Response[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -691,7 +820,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     def resume(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: Response | Response[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -702,7 +831,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     def resume(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: Response | Response[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -736,7 +865,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     async def resume_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncResponse,
         content: UserContent,
         **params: Unpack[Params],
@@ -748,7 +877,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     async def resume_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncResponse[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -760,7 +889,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     async def resume_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncResponse | AsyncResponse[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -771,7 +900,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     async def resume_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncResponse | AsyncResponse[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -806,7 +935,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: ContextResponse[DepsT, None],
         content: UserContent,
         **params: Unpack[Params],
@@ -819,7 +948,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: ContextResponse[DepsT, FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -832,7 +961,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: ContextResponse[DepsT, None] | ContextResponse[DepsT, FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -844,7 +973,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: ContextResponse[DepsT, None] | ContextResponse[DepsT, FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -881,7 +1010,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncContextResponse[DepsT, None],
         content: UserContent,
         **params: Unpack[Params],
@@ -894,7 +1023,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncContextResponse[DepsT, FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -907,7 +1036,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncContextResponse[DepsT, None]
         | AsyncContextResponse[DepsT, FormattableT],
         content: UserContent,
@@ -920,7 +1049,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncContextResponse[DepsT, None]
         | AsyncContextResponse[DepsT, FormattableT],
         content: UserContent,
@@ -957,7 +1086,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     def resume_stream(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: StreamResponse,
         content: UserContent,
         **params: Unpack[Params],
@@ -969,7 +1098,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     def resume_stream(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: StreamResponse[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -981,7 +1110,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     def resume_stream(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: StreamResponse | StreamResponse[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -992,7 +1121,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     def resume_stream(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: StreamResponse | StreamResponse[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -1026,7 +1155,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     async def resume_stream_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncStreamResponse,
         content: UserContent,
         **params: Unpack[Params],
@@ -1038,7 +1167,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     async def resume_stream_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncStreamResponse[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -1050,7 +1179,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     async def resume_stream_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncStreamResponse | AsyncStreamResponse[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -1061,7 +1190,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
     async def resume_stream_async(
         self,
         *,
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncStreamResponse | AsyncStreamResponse[FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -1096,7 +1225,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: ContextStreamResponse[DepsT, None],
         content: UserContent,
         **params: Unpack[Params],
@@ -1109,7 +1238,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: ContextStreamResponse[DepsT, FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -1122,7 +1251,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: ContextStreamResponse[DepsT, None]
         | ContextStreamResponse[DepsT, FormattableT],
         content: UserContent,
@@ -1137,7 +1266,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: ContextStreamResponse[DepsT, None]
         | ContextStreamResponse[DepsT, FormattableT],
         content: UserContent,
@@ -1177,7 +1306,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncContextStreamResponse[DepsT, None],
         content: UserContent,
         **params: Unpack[Params],
@@ -1190,7 +1319,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncContextStreamResponse[DepsT, FormattableT],
         content: UserContent,
         **params: Unpack[Params],
@@ -1203,7 +1332,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncContextStreamResponse[DepsT, None]
         | AsyncContextStreamResponse[DepsT, FormattableT],
         content: UserContent,
@@ -1219,7 +1348,7 @@ class BaseClient(Generic[ModelIdT, ProviderClientT], ABC):
         self,
         *,
         ctx: Context[DepsT],
-        model_id: ModelIdT,
+        model_id: str,
         response: AsyncContextStreamResponse[DepsT, None]
         | AsyncContextStreamResponse[DepsT, FormattableT],
         content: UserContent,