types/client: relax think type to support model-defined thinking levels

ollama/_client.py

@@ -209,7 +209,7 @@ def generate(
     template: str = '',
     context: Optional[Sequence[int]] = None,
     stream: Literal[False] = False,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     raw: bool = False,
@@ -233,7 +233,7 @@ def generate(
     template: str = '',
     context: Optional[Sequence[int]] = None,
     stream: Literal[True] = True,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     raw: bool = False,
@@ -256,7 +256,7 @@ def generate(
     template: Optional[str] = None,
     context: Optional[Sequence[int]] = None,
     stream: bool = False,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     raw: Optional[bool] = None,
@@ -313,7 +313,7 @@ def chat(
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: Literal[False] = False,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
@@ -329,7 +329,7 @@ def chat(
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: Literal[True] = True,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
@@ -344,7 +344,7 @@ def chat(
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: bool = False,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
@@ -842,7 +842,7 @@ async def generate(
     template: str = '',
     context: Optional[Sequence[int]] = None,
     stream: Literal[False] = False,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     raw: bool = False,
@@ -866,7 +866,7 @@ async def generate(
     template: str = '',
     context: Optional[Sequence[int]] = None,
     stream: Literal[True] = True,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     raw: bool = False,
@@ -889,7 +889,7 @@ async def generate(
     template: Optional[str] = None,
     context: Optional[Sequence[int]] = None,
     stream: bool = False,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     raw: Optional[bool] = None,
@@ -945,7 +945,7 @@ async def chat(
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: Literal[False] = False,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
@@ -961,7 +961,7 @@ async def chat(
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: Literal[True] = True,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
@@ -976,7 +976,7 @@ async def chat(
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: bool = False,
-    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
+    think: Optional[Union[bool, str]] = None,
     logprobs: Optional[bool] = None,
     top_logprobs: Optional[int] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,

ollama/_types.py

@@ -207,7 +207,7 @@ class GenerateRequest(BaseGenerateRequest):
   images: Optional[Sequence[Image]] = None
   'Image data for multimodal models.'
 
-  think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None
+  think: Optional[Union[bool, str]] = None
   'Enable thinking mode (for thinking models).'
 
   logprobs: Optional[bool] = None
@@ -400,7 +400,7 @@ def serialize_model(self, nxt):
   tools: Optional[Sequence[Tool]] = None
   'Tools to use for the chat.'
 
-  think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None
+  think: Optional[Union[bool, str]] = None
   'Enable thinking mode (for thinking models).'
 
   logprobs: Optional[bool] = None

tests/test_client.py

@@ -1490,10 +1490,40 @@ async def test_async_client_context_manager():
 
 
 def test_generate_think_annotation_matches_chat():
-  # The `think` parameter accepts bool or the 'low'/'medium'/'high' string levels.
-  # Client.generate must keep the same annotation as Client.chat and
-  # AsyncClient.generate so passing a string level does not raise a false type
-  # error (regression guard for the sync generate overloads/implementation).
+  # The `think` parameter accepts bool or string thinking levels (e.g. 'low', 'medium', 'high', 'xhigh', 'max').
+  # Client.generate must keep the same annotation as Client.chat,
+  # AsyncClient.chat, and AsyncClient.generate so passing a string level does not
+  # raise a false type error (regression guard for the sync generate overloads/implementation).
   expected = inspect.signature(Client.chat).parameters['think'].annotation
   assert inspect.signature(Client.generate).parameters['think'].annotation == expected
+  assert inspect.signature(AsyncClient.chat).parameters['think'].annotation == expected
   assert inspect.signature(AsyncClient.generate).parameters['think'].annotation == expected
+
+
+def test_client_chat_with_think_level(httpserver: HTTPServer):
+  httpserver.expect_ordered_request(
+    '/api/chat',
+    method='POST',
+    json={
+      'model': 'qwen3.8:27b',
+      'messages': [{'role': 'user', 'content': 'Hello'}],
+      'tools': [],
+      'stream': False,
+      'think': 'xhigh',
+    },
+  ).respond_with_json(
+    {
+      'model': 'qwen3.8:27b',
+      'message': {
+        'role': 'assistant',
+        'content': 'Hi there.',
+        'thinking': 'Thinking deeply...',
+      },
+    }
+  )
+
+  client = Client(httpserver.url_for('/'))
+  response = client.chat('qwen3.8:27b', messages=[{'role': 'user', 'content': 'Hello'}], think='xhigh')
+  assert response['model'] == 'qwen3.8:27b'
+  assert response['message']['content'] == 'Hi there.'
+  assert response['message']['thinking'] == 'Thinking deeply...'

tests/test_type_serialization.py

@@ -4,7 +4,7 @@
 
 import pytest
 
-from ollama._types import CreateRequest, Image
+from ollama._types import ChatRequest, CreateRequest, GenerateRequest, Image
 
 
 def test_image_serialization_bytes():
@@ -105,3 +105,20 @@ def test_create_request_serialization_license_list():
   request = CreateRequest(model='test-model', license=['MIT', 'Apache-2.0'])
   serialized = request.model_dump()
   assert serialized['license'] == ['MIT', 'Apache-2.0']
+
+
+@pytest.mark.parametrize('level', ['low', 'medium', 'high', 'xhigh', 'max'])
+def test_think_model_defined_levels_serialization(level):
+  chat_req = ChatRequest(model='test-model', messages=[{'role': 'user', 'content': 'hi'}], think=level)
+  assert chat_req.think == level
+  assert chat_req.model_dump(exclude_none=True)['think'] == level
+
+  gen_req = GenerateRequest(model='test-model', think=level)
+  assert gen_req.think == level
+  assert gen_req.model_dump(exclude_none=True)['think'] == level
+
+
+def test_think_boolean_serialization():
+  assert ChatRequest(model='test-model', think=True).model_dump(exclude_none=True)['think'] is True
+  assert ChatRequest(model='test-model', think=False).model_dump(exclude_none=True)['think'] is False
+  assert 'think' not in ChatRequest(model='test-model', think=None).model_dump(exclude_none=True)