feat: add reasoning parameter to ChatAzureOpenAI() (#260)

* fix: add reasoning parameter to ChatAzureOpenAI for reasoning model detection

Azure OpenAI uses custom deployment IDs instead of model names, so the
existing `is_reasoning_model()` check (which looks for "o" or "gpt-5"
prefixes) doesn't work. This adds a `reasoning` parameter to
`ChatAzureOpenAI()` that users can set to enable reasoning features for
their o-series deployments.

Changes:
- Add `_is_reasoning_model` attribute to `OpenAIProvider`
- Use `self._is_reasoning_model` instead of `is_reasoning_model(self.model)`
  in `_chat_perform_args()` to determine whether to include reasoning content
- Add `reasoning` parameter to `ChatAzureOpenAI()` matching `ChatOpenAI()`
- Fix type annotations to use `ResponsesSubmitInputArgs` and `Response`
  (since Azure uses the responses API, not completions)

Fixes #242

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* fix: also detect reasoning models by deployment_id pattern

When deployment_id matches known reasoning model patterns (o*, gpt-5*),
automatically enable reasoning features. This maintains backward
compatibility with existing tests while still allowing explicit
override via the reasoning parameter for custom deployment names.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* Include encrypted content if reasoning is specified

This way, for Azure, an empty dictionary will lead to encrypted content being included

* Update changelog

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>

Author: cpsievert

CHANGELOG.md

@@ -20,6 +20,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
     * `tool_web_fetch()` is supported by Claude (requires beta header) and Google.
     * New content types `ContentToolRequestSearch`, `ContentToolResponseSearch`, `ContentToolRequestFetch`, and `ContentToolResponseFetch` capture web tool interactions.
 * `ChatOpenAI()` and `ChatAzureOpenAI()` gain a new `service_tier` parameter to request a specific service tier (e.g., `"flex"` for slower/cheaper or `"priority"` for faster/more expensive). (#204)
+* `ChatAzureOpenAI()` gains a `reasoning` parameter. (#260)
 * `Chat` and `Turn` now have a `_repr_markdown_` method and an overall improved `repr()` experience. (#245)
 * `ChatAuto()` now accepts `"claude"` as an alias for `"anthropic"`, reflecting Anthropic's rebranding of developer tools under the Claude name. (#239)
 
@@ -40,6 +41,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 * Fixed MCP tools not working with `ChatGoogle()`. (#257)
 * Tool functions parameters that are `typing.Annotated` with a `pydantic.Field` (e.g., `def add(x: Annotated[int, Field(description="First number")])`) are now handled correctly. (#251)
 
+
 ## [0.14.0] - 2025-12-09
 
 ### New features

chatlas/_provider_openai.py

@@ -279,7 +279,7 @@ def _chat_perform_args(
 
         # Request reasoning content for reasoning models
         include = []
-        if is_reasoning_model(self.model):
+        if "reasoning" in kwargs_full or is_reasoning_model(self.model):
             include.append("reasoning.encrypted_content")
 
         if "log_probs" in kwargs_full:

chatlas/_provider_openai_azure.py

@@ -11,7 +11,15 @@
 from ._utils import MISSING, MISSING_TYPE, is_testing, split_http_client_kwargs
 
 if TYPE_CHECKING:
-    from .types.openai import ChatAzureClientArgs, SubmitInputArgs
+    from openai.types.responses import Response
+    from openai.types.shared.reasoning_effort import ReasoningEffort
+    from openai.types.shared_params.reasoning import Reasoning
+
+    from .types.openai import (
+        ChatAzureClientArgs,
+        ResponsesSubmitInputArgs,
+        SubmitInputArgs,
+    )
 
 
 def ChatAzureOpenAI(
@@ -21,11 +29,12 @@ def ChatAzureOpenAI(
     api_version: str,
     api_key: Optional[str] = None,
     system_prompt: Optional[str] = None,
+    reasoning: "Optional[ReasoningEffort | Reasoning]" = None,
     service_tier: Optional[
         Literal["auto", "default", "flex", "scale", "priority"]
     ] = None,
     kwargs: Optional["ChatAzureClientArgs"] = None,
-) -> Chat["SubmitInputArgs", ChatCompletion]:
+) -> "Chat[ResponsesSubmitInputArgs, Response]":
     """
     Chat with a model hosted on Azure OpenAI.
 
@@ -65,6 +74,11 @@ def ChatAzureOpenAI(
         variable.
     system_prompt
         A system prompt to set the behavior of the assistant.
+    reasoning
+        The reasoning effort (e.g., `"low"`, `"medium"`, `"high"`) for
+        reasoning-capable models like the o and gpt-5 series. To use the default
+        reasoning settings in a way that will work for multi-turn conversations,
+        set this to an empty dictionary `{}`.
     service_tier
         Request a specific service tier. Options:
         - `"auto"` (default): uses the service tier configured in Project settings.
@@ -81,7 +95,13 @@ def ChatAzureOpenAI(
         A Chat object.
     """
 
-    kwargs_chat: "SubmitInputArgs" = {}
+    kwargs_chat: "ResponsesSubmitInputArgs" = {}
+
+    if reasoning is not None:
+        if isinstance(reasoning, str):
+            reasoning = {"effort": reasoning, "summary": "auto"}
+        kwargs_chat["reasoning"] = reasoning
+
     if service_tier is not None:
         kwargs_chat["service_tier"] = service_tier