Skip to content

Replace params: Any in stream/generate with generalized types #150

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
elprans wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

Replace params: Any in stream/generate with generalized types #150

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 1 addition & 4 deletions examples/coding_agent_minimal.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,6 @@
import asyncio
import json
import sys
import typing

import ai

Expand All @@ -16,9 +15,7 @@
with sed, make sure to double check the result.
"""

STREAM_PARAMS: dict[str, typing.Any] = {
"providerOptions": {"gateway": {"caching": "auto"}},
}
STREAM_PARAMS = ai.InferenceRequestParams(cache=ai.CacheParams(mode="auto"))


@ai.tool
Expand Down
12 changes: 6 additions & 6 deletions examples/model_params.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,12 +13,12 @@


async def main() -> None:
params = {
"providerOptions": {
"gateway": {"sort": "cost"},
"anthropic": {"speed": "fast"},
}
}
params = ai.InferenceRequestParams(
routing=ai.RoutingParams(
provider_ranking=ai.ProviderRankingStrategy.COST
),
extra_body={"providerOptions": {"anthropic": {"speed": "fast"}}},
Copy link
Copy Markdown
Contributor

@msullivan msullivan May 22, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

extra_body feels clunky... but maybe only because it has like two more levels of ai gateway wrapping on it?

)
async with ai.stream(model, messages, params=params) as stream:
async for event in stream:
if isinstance(event, ai.events.TextDelta):
Expand Down
2 changes: 1 addition & 1 deletion examples/prompt_caching.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -142,7 +142,7 @@ async def _run(user_text: str) -> ai.types.usage.Usage | None:
ai.system_message(SYSTEM_PROMPT),
ai.user_message(user_text),
]
params = {"providerOptions": {"gateway": {"caching": "auto"}}}
params = ai.InferenceRequestParams(cache=ai.CacheParams(mode="auto"))
async with agent.run(model, messages, params=params) as stream:
async for _event in stream:
pass
Expand Down
67 changes: 60 additions & 7 deletions src/ai/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,11 +48,41 @@
UnsupportedProviderError,
)
from .models import (
DEFAULT,
GLOBAL,
RANDOM,
UNSET,
CacheParams,
CloudRegion,
ContextManagementParams,
GeoRegion,
ImageParams,
InferenceRequestParams,
MinPSamplerParams,
Model,
ModelProviderDefault,
OutputParams,
Provider,
ProviderProtocol,
ProviderRankingStrategy,
ProviderServiceParams,
RandomSeed,
ReasoningParams,
RepetitionPenaltyParams,
RoutingParams,
RoutingTarget,
RoutingTargetChain,
SeedSamplerParams,
Stream,
TemperatureSamplerParams,
TokenThreshold,
ToolCallingParams,
ToolChoiceMode,
ToolRef,
ToolSelection,
TopKSamplerParams,
TopPSamplerParams,
Unset,
VideoParams,
generate,
get_model,
Expand All @@ -72,18 +102,27 @@
)

__all__ = [
# Models (from models/)
"DEFAULT",
"GLOBAL",
"RANDOM",
"UNSET",
"AIError",
# Agents — primary API
"Agent",
# Agents — tools
"AgentTool",
"CacheParams",
"CloudRegion",
"ConfigurationError",
"Context",
"ContextManagementParams",
"GeoRegion",
"HTTPErrorContext",
"ImageParams",
"InferenceRequestParams",
"InstallationError",
"MinPSamplerParams",
"Model",
"ModelProviderDefault",
"OutputParams",
"Provider",
"ProviderAPIError",
"ProviderAuthenticationError",
Expand All @@ -99,34 +138,50 @@
"ProviderOverloadedError",
"ProviderPermissionDeniedError",
"ProviderProtocol",
"ProviderRankingStrategy",
"ProviderRateLimitError",
"ProviderRequestTooLargeError",
"ProviderResponseError",
"ProviderServiceParams",
"ProviderServiceUnavailableError",
"ProviderStatusError",
"ProviderTimeoutError",
"ProviderUnprocessableEntityError",
"RandomSeed",
"ReasoningParams",
"RepetitionPenaltyParams",
"RoutingParams",
"RoutingTarget",
"RoutingTargetChain",
"SeedSamplerParams",
"Stream",
"StreamingStatusTool",
"StreamingTextTool",
"SubAgentTool",
"TemperatureSamplerParams",
"TokenThreshold",
"Tool",
"ToolCall",
"ToolCallingParams",
"ToolChoiceMode",
"ToolRef",
"ToolRunner",
"ToolSelection",
"TopKSamplerParams",
"TopPSamplerParams",
"Unset",
"UnsupportedProviderError",
"VideoParams",
"abort_pending_hook",
"agent",
"assistant_message",
"cancel_hook",
"errors",
# Submodules
"events",
"file_part",
"generate",
"get_model",
"get_provider",
# Agents — hooks
"hook",
"mcp",
"messages",
Expand All @@ -143,9 +198,7 @@
"tool_result",
"tool_result_part",
"tools",
# Builders (from types/builders)
"user_message",
"util",
# Agents — composition
"yield_from",
]
10 changes: 10 additions & 0 deletions src/ai/_types.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,10 @@
from collections.abc import Iterator
from typing import Protocol, TypeVar

_T_co = TypeVar("_T_co", covariant=True)


class Collection(Protocol[_T_co]):
def __contains__(self, value: object, /) -> bool: ...
def __iter__(self) -> Iterator[_T_co]: ...
def __len__(self) -> int: ...
3 changes: 2 additions & 1 deletion src/ai/agents/_middleware.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,7 @@
import pydantic

from ..models.core.model import Model
from ..models.core.params import GenerateParams
from ..types import events as events_
from ..types.tools import Tool
from .agent import Context
Expand Down Expand Up @@ -71,7 +72,7 @@ class GenerateContext:

model: Model
messages: list[messages_.Message]
params: Any
params: GenerateParams

def __post_init__(self) -> None:
object.__setattr__(self, "messages", list(self.messages))
Expand Down
12 changes: 7 additions & 5 deletions src/ai/agents/agent.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -813,7 +813,9 @@ class Context(pydantic.BaseModel):
output_type: type[pydantic.BaseModel] | None = pydantic.Field(
default=None, exclude=True, repr=False
)
params: Any = pydantic.Field(default=None, exclude=True, repr=False)
params: models.InferenceRequestParams | None = pydantic.Field(
default=None, exclude=True, repr=False
)

_agent_tools_by_name: dict[str, AgentTool] = pydantic.PrivateAttr(
default_factory=dict
Expand Down Expand Up @@ -1178,7 +1180,7 @@ def run(
model: models.Model,
messages: list[types.messages.Message],
*,
params: Any = None,
params: models.InferenceRequestParams | None = None,
_middleware: list[middleware_._Middleware] | None = None,
) -> AbstractAsyncContextManager[AgentStream[str]]: ...
@overload
Expand All @@ -1188,7 +1190,7 @@ def run[T: pydantic.BaseModel](
messages: list[types.messages.Message],
*,
output_type: type[T],
params: Any = None,
params: models.InferenceRequestParams | None = None,
_middleware: list[middleware_._Middleware] | None = None,
) -> AbstractAsyncContextManager[AgentStream[T]]: ...
def run(
Expand All @@ -1197,7 +1199,7 @@ def run(
messages: list[types.messages.Message],
*,
output_type: type[pydantic.BaseModel] | None = None,
params: Any = None,
params: models.InferenceRequestParams | None = None,
_middleware: list[middleware_._Middleware] | None = None,
) -> AbstractAsyncContextManager[AgentStream[Any]]:
"""Run the agent loop, yielding events to the consumer.
Expand Down Expand Up @@ -1237,7 +1239,7 @@ async def _run(
messages: list[types.messages.Message],
*,
output_type: type[pydantic.BaseModel] | None,
params: Any,
params: models.InferenceRequestParams | None,
_middleware: list[middleware_._Middleware] | None,
) -> AsyncIterator[AgentStream[Any]]:
context = Context(
Expand Down
68 changes: 65 additions & 3 deletions src/ai/models/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,23 +45,85 @@
stream,
)
from .core.model import Model, get_model
from .core.params import GenerateParams, ImageParams, VideoParams
from .core.params import (
DEFAULT,
GLOBAL,
RANDOM,
UNSET,
CacheParams,
CloudRegion,
ContextManagementParams,
GenerateParams,
GeoRegion,
ImageParams,
InferenceRequestParams,
MinPSamplerParams,
ModelProviderDefault,
OutputParams,
ProviderRankingStrategy,
ProviderServiceParams,
RandomSeed,
ReasoningParams,
RepetitionPenaltyParams,
RoutingParams,
RoutingTarget,
RoutingTargetChain,
SeedSamplerParams,
TemperatureSamplerParams,
TokenThreshold,
ToolCallingParams,
ToolChoiceMode,
ToolRef,
ToolSelection,
TopKSamplerParams,
TopPSamplerParams,
Unset,
VideoParams,
)

__all__ = [
# Core types
"DEFAULT",
"GLOBAL",
"RANDOM",
"UNSET",
"CacheParams",
"CloudRegion",
"ContextManagementParams",
"Executor",
"GenerateExecutor",
"GenerateParams",
"GenerateRequest",
"GeoRegion",
"ImageParams",
"InferenceRequestParams",
"MinPSamplerParams",
"Model",
"ModelProviderDefault",
"OutputParams",
"Provider",
"ProviderProtocol",
"ProviderRankingStrategy",
"ProviderServiceParams",
"RandomSeed",
"ReasoningParams",
"RepetitionPenaltyParams",
"RoutingParams",
"RoutingTarget",
"RoutingTargetChain",
"SeedSamplerParams",
"Stream",
"StreamExecutor",
"StreamRequest",
"TemperatureSamplerParams",
"TokenThreshold",
"ToolCallingParams",
"ToolChoiceMode",
"ToolRef",
"ToolSelection",
"TopKSamplerParams",
"TopPSamplerParams",
"Unset",
"VideoParams",
# Public API
"generate",
"get_model",
"probe",
Expand Down
Loading
Loading