feat: add dimensions to embed request (#574)

Bumps [actions/setup-python](https://github.com/actions/setup-python) from 5 to 6.
- [Release notes](https://github.com/actions/setup-python/releases)
- [Commits](https://github.com/actions/setup-python/compare/v5...v6)

---
updated-dependencies:
- dependency-name: actions/setup-python
  dependency-version: '6'
  dependency-type: direct:production
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

.github/workflows/publish.yaml

@@ -13,8 +13,8 @@ jobs:
       id-token: write
       contents: write
     steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
       - uses: astral-sh/setup-uv@v5
         with:
           enable-cache: true

.github/workflows/test.yaml

@@ -10,7 +10,7 @@ jobs:
   test:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
       - uses: astral-sh/setup-uv@v5
         with:
           enable-cache: true
@@ -19,8 +19,8 @@ jobs:
   lint:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
       - uses: astral-sh/setup-uv@v5
         with:
           enable-cache: true

examples/README.md

@@ -28,8 +28,10 @@ See [ollama/docs/api.md](https://github.com/ollama/ollama/blob/main/docs/api.md)
 - [multi-tool.py](multi-tool.py) - Using multiple tools, with thinking enabled
 
  #### gpt-oss
-- [gpt-oss-tools.py](gpt-oss-tools.py) - Using tools with gpt-oss
-- [gpt-oss-tools-stream.py](gpt-oss-tools-stream.py) - Using tools with gpt-oss, with streaming enabled
+- [gpt-oss-tools.py](gpt-oss-tools.py)
+- [gpt-oss-tools-stream.py](gpt-oss-tools-stream.py) 
+- [gpt-oss-tools-browser.py](gpt-oss-tools-browser.py) - Using browser research tools with gpt-oss
+- [gpt-oss-tools-browser-stream.py](gpt-oss-tools-browser-stream.py) - Using browser research tools with gpt-oss, with streaming enabled
 
 
 ### Multimodal with Images - Chat with a multimodal (image chat) model
@@ -73,3 +75,6 @@ Requirement: `pip install tqdm`
 
 ### Thinking (generate) - Enable thinking mode for a model
 - [thinking-generate.py](thinking-generate.py)
+
+### Thinking (levels) - Choose the thinking level
+- [thinking-levels.py](thinking-levels.py)

examples/gpt-oss-tools-browser-stream.py

@@ -0,0 +1,198 @@
+# /// script
+# requires-python = ">=3.11"
+# dependencies = [
+#     "gpt-oss",
+#     "ollama",
+#     "rich",
+# ]
+# ///
+
+import asyncio
+import json
+from typing import Iterator, Optional
+
+from gpt_oss.tools.simple_browser import ExaBackend, SimpleBrowserTool
+from openai_harmony import Author, Role, TextContent
+from openai_harmony import Message as HarmonyMessage
+from rich import print
+
+from ollama import Client
+from ollama._types import ChatResponse
+
+_backend = ExaBackend(source='web')
+_browser_tool = SimpleBrowserTool(backend=_backend)
+
+
+def heading(text):
+  print(text)
+  print('=' * (len(text) + 3))
+
+
+async def _browser_search_async(query: str, topn: int = 10, source: str | None = None) -> str:
+  # map Ollama message to Harmony format
+  harmony_message = HarmonyMessage(
+    author=Author(role=Role.USER),
+    content=[TextContent(text=json.dumps({'query': query, 'topn': topn}))],
+    recipient='browser.search',
+  )
+
+  result_text: str = ''
+  async for response in _browser_tool._process(harmony_message):
+    if response.content:
+      for content in response.content:
+        if isinstance(content, TextContent):
+          result_text += content.text
+  return result_text or f'No results for query: {query}'
+
+
+async def _browser_open_async(id: int | str = -1, cursor: int = -1, loc: int = -1, num_lines: int = -1, *, view_source: bool = False, source: str | None = None) -> str:
+  payload = {'id': id, 'cursor': cursor, 'loc': loc, 'num_lines': num_lines, 'view_source': view_source, 'source': source}
+
+  harmony_message = HarmonyMessage(
+    author=Author(role=Role.USER),
+    content=[TextContent(text=json.dumps(payload))],
+    recipient='browser.open',
+  )
+
+  result_text: str = ''
+  async for response in _browser_tool._process(harmony_message):
+    if response.content:
+      for content in response.content:
+        if isinstance(content, TextContent):
+          result_text += content.text
+  return result_text or f'Could not open: {id}'
+
+
+async def _browser_find_async(pattern: str, cursor: int = -1) -> str:
+  payload = {'pattern': pattern, 'cursor': cursor}
+
+  harmony_message = HarmonyMessage(
+    author=Author(role=Role.USER),
+    content=[TextContent(text=json.dumps(payload))],
+    recipient='browser.find',
+  )
+
+  result_text: str = ''
+  async for response in _browser_tool._process(harmony_message):
+    if response.content:
+      for content in response.content:
+        if isinstance(content, TextContent):
+          result_text += content.text
+  return result_text or f'Pattern not found: {pattern}'
+
+
+def browser_search(query: str, topn: int = 10, source: Optional[str] = None) -> str:
+  return asyncio.run(_browser_search_async(query=query, topn=topn, source=source))
+
+
+def browser_open(id: int | str = -1, cursor: int = -1, loc: int = -1, num_lines: int = -1, *, view_source: bool = False, source: Optional[str] = None) -> str:
+  return asyncio.run(_browser_open_async(id=id, cursor=cursor, loc=loc, num_lines=num_lines, view_source=view_source, source=source))
+
+
+def browser_find(pattern: str, cursor: int = -1) -> str:
+  return asyncio.run(_browser_find_async(pattern=pattern, cursor=cursor))
+
+
+# Schema definitions for each browser tool
+browser_search_schema = {
+  'type': 'function',
+  'function': {
+    'name': 'browser.search',
+  },
+}
+
+browser_open_schema = {
+  'type': 'function',
+  'function': {
+    'name': 'browser.open',
+  },
+}
+
+browser_find_schema = {
+  'type': 'function',
+  'function': {
+    'name': 'browser.find',
+  },
+}
+
+available_tools = {
+  'browser.search': browser_search,
+  'browser.open': browser_open,
+  'browser.find': browser_find,
+}
+
+
+model = 'gpt-oss:20b'
+print('Model: ', model, '\n')
+
+prompt = 'What is Ollama?'
+print('You: ', prompt, '\n')
+messages = [{'role': 'user', 'content': prompt}]
+
+client = Client()
+
+# gpt-oss can call tools while "thinking"
+# a loop is needed to call the tools and get the results
+final = True
+while True:
+  response_stream: Iterator[ChatResponse] = client.chat(
+    model=model,
+    messages=messages,
+    tools=[browser_search_schema, browser_open_schema, browser_find_schema],
+    options={'num_ctx': 8192},  # 8192 is the recommended lower limit for the context window
+    stream=True,
+  )
+
+  tool_calls = []
+  thinking = ''
+  content = ''
+
+  for chunk in response_stream:
+    if chunk.message.tool_calls:
+      tool_calls.extend(chunk.message.tool_calls)
+
+    if chunk.message.content:
+      if not (chunk.message.thinking or chunk.message.thinking == '') and final:
+        heading('\n\nFinal result: ')
+        final = False
+      print(chunk.message.content, end='', flush=True)
+
+    if chunk.message.thinking:
+      thinking += chunk.message.thinking
+      print(chunk.message.thinking, end='', flush=True)
+
+  if thinking != '':
+    messages.append({'role': 'assistant', 'content': thinking, 'tool_calls': tool_calls})
+
+  print()
+
+  if tool_calls:
+    for tool_call in tool_calls:
+      tool_name = tool_call.function.name
+      args = tool_call.function.arguments or {}
+      function_to_call = available_tools.get(tool_name)
+
+      if function_to_call:
+        heading(f'\nCalling tool: {tool_name}')
+        if args:
+          print(f'Arguments: {args}')
+
+        try:
+          result = function_to_call(**args)
+          print(f'Tool result: {result[:200]}')
+          if len(result) > 200:
+            heading('... [truncated]')
+          print()
+
+          result_message = {'role': 'tool', 'content': result, 'tool_name': tool_name}
+          messages.append(result_message)
+
+        except Exception as e:
+          err = f'Error from {tool_name}: {e}'
+          print(err)
+          messages.append({'role': 'tool', 'content': err, 'tool_name': tool_name})
+      else:
+        print(f'Tool {tool_name} not found')
+  else:
+    # no more tool calls, we can stop the loop
+    break

examples/gpt-oss-tools-browser.py

@@ -0,0 +1,175 @@
+# /// script
+# requires-python = ">=3.11"
+# dependencies = [
+#     "gpt-oss",
+#     "ollama",
+#     "rich",
+# ]
+# ///
+
+import asyncio
+import json
+from typing import Optional
+
+from gpt_oss.tools.simple_browser import ExaBackend, SimpleBrowserTool
+from openai_harmony import Author, Role, TextContent
+from openai_harmony import Message as HarmonyMessage
+
+from ollama import Client
+
+_backend = ExaBackend(source='web')
+_browser_tool = SimpleBrowserTool(backend=_backend)
+
+
+def heading(text):
+  print(text)
+  print('=' * (len(text) + 3))
+
+
+async def _browser_search_async(query: str, topn: int = 10, source: str | None = None) -> str:
+  # map Ollama message to Harmony format
+  harmony_message = HarmonyMessage(
+    author=Author(role=Role.USER),
+    content=[TextContent(text=json.dumps({'query': query, 'topn': topn}))],
+    recipient='browser.search',
+  )
+
+  result_text: str = ''
+  async for response in _browser_tool._process(harmony_message):
+    if response.content:
+      for content in response.content:
+        if isinstance(content, TextContent):
+          result_text += content.text
+  return result_text or f'No results for query: {query}'
+
+
+async def _browser_open_async(id: int | str = -1, cursor: int = -1, loc: int = -1, num_lines: int = -1, *, view_source: bool = False, source: str | None = None) -> str:
+  payload = {'id': id, 'cursor': cursor, 'loc': loc, 'num_lines': num_lines, 'view_source': view_source, 'source': source}
+
+  harmony_message = HarmonyMessage(
+    author=Author(role=Role.USER),
+    content=[TextContent(text=json.dumps(payload))],
+    recipient='browser.open',
+  )
+
+  result_text: str = ''
+  async for response in _browser_tool._process(harmony_message):
+    if response.content:
+      for content in response.content:
+        if isinstance(content, TextContent):
+          result_text += content.text
+  return result_text or f'Could not open: {id}'
+
+
+async def _browser_find_async(pattern: str, cursor: int = -1) -> str:
+  payload = {'pattern': pattern, 'cursor': cursor}
+
+  harmony_message = HarmonyMessage(
+    author=Author(role=Role.USER),
+    content=[TextContent(text=json.dumps(payload))],
+    recipient='browser.find',
+  )
+
+  result_text: str = ''
+  async for response in _browser_tool._process(harmony_message):
+    if response.content:
+      for content in response.content:
+        if isinstance(content, TextContent):
+          result_text += content.text
+  return result_text or f'Pattern not found: {pattern}'
+
+
+def browser_search(query: str, topn: int = 10, source: Optional[str] = None) -> str:
+  return asyncio.run(_browser_search_async(query=query, topn=topn, source=source))
+
+
+def browser_open(id: int | str = -1, cursor: int = -1, loc: int = -1, num_lines: int = -1, *, view_source: bool = False, source: Optional[str] = None) -> str:
+  return asyncio.run(_browser_open_async(id=id, cursor=cursor, loc=loc, num_lines=num_lines, view_source=view_source, source=source))
+
+
+def browser_find(pattern: str, cursor: int = -1) -> str:
+  return asyncio.run(_browser_find_async(pattern=pattern, cursor=cursor))
+
+
+# Schema definitions for each browser tool
+browser_search_schema = {
+  'type': 'function',
+  'function': {
+    'name': 'browser.search',
+  },
+}
+
+browser_open_schema = {
+  'type': 'function',
+  'function': {
+    'name': 'browser.open',
+  },
+}
+
+browser_find_schema = {
+  'type': 'function',
+  'function': {
+    'name': 'browser.find',
+  },
+}
+
+available_tools = {
+  'browser.search': browser_search,
+  'browser.open': browser_open,
+  'browser.find': browser_find,
+}
+
+
+model = 'gpt-oss:20b'
+print('Model: ', model, '\n')
+
+prompt = 'What is Ollama?'
+print('You: ', prompt, '\n')
+messages = [{'role': 'user', 'content': prompt}]
+
+client = Client()
+while True:
+  response = client.chat(
+    model=model,
+    messages=messages,
+    tools=[browser_search_schema, browser_open_schema, browser_find_schema],
+    options={'num_ctx': 8192},  # 8192 is the recommended lower limit for the context window
+  )
+
+  if hasattr(response.message, 'thinking') and response.message.thinking:
+    heading('Thinking')
+    print(response.message.thinking.strip() + '\n')
+
+  if hasattr(response.message, 'content') and response.message.content:
+    heading('Assistant')
+    print(response.message.content.strip() + '\n')
+
+  # add message to chat history
+  messages.append(response.message)
+
+  if response.message.tool_calls:
+    for tool_call in response.message.tool_calls:
+      tool_name = tool_call.function.name
+      args = tool_call.function.arguments or {}
+      function_to_call = available_tools.get(tool_name)
+      if not function_to_call:
+        print(f'Unknown tool: {tool_name}')
+        continue
+
+      try:
+        result = function_to_call(**args)
+        heading(f'Tool: {tool_name}')
+        if args:
+          print(f'Arguments: {args}')
+        print(result[:200])
+        if len(result) > 200:
+          print('... [truncated]')
+        print()
+        messages.append({'role': 'tool', 'content': result, 'tool_name': tool_name})
+      except Exception as e:
+        err = f'Error from {tool_name}: {e}'
+        print(err)
+        messages.append({'role': 'tool', 'content': err, 'tool_name': tool_name})
+  else:
+    # break on no more tool calls
+    break

examples/gpt-oss-tools-stream.py

@@ -1,7 +1,17 @@
+# /// script
+# requires-python = ">=3.11"
+# dependencies = [
+#     "gpt-oss",
+#     "ollama",
+#     "rich",
+# ]
+# ///
 import random
 from typing import Iterator
 
-from ollama import chat
+from rich import print
+
+from ollama import Client
 from ollama._types import ChatResponse
 
 
@@ -40,37 +50,55 @@ available_tools = {'get_weather': get_weather, 'get_weather_conditions': get_wea
 
 messages = [{'role': 'user', 'content': 'What is the weather like in London? What are the conditions in Toronto?'}]
 
+client = Client(
+  # Ollama Turbo
+  # host="https://ollama.com", headers={'Authorization': (os.getenv('OLLAMA_API_KEY'))}
+)
 
 model = 'gpt-oss:20b'
 # gpt-oss can call tools while "thinking"
 # a loop is needed to call the tools and get the results
 final = True
 while True:
-  response_stream: Iterator[ChatResponse] = chat(model=model, messages=messages, tools=[get_weather, get_weather_conditions], stream=True)
+  response_stream: Iterator[ChatResponse] = client.chat(model=model, messages=messages, tools=[get_weather, get_weather_conditions], stream=True)
+  tool_calls = []
+  thinking = ''
+  content = ''
 
   for chunk in response_stream:
+    if chunk.message.tool_calls:
+      tool_calls.extend(chunk.message.tool_calls)
+
     if chunk.message.content:
       if not (chunk.message.thinking or chunk.message.thinking == '') and final:
-        print('\nFinal result: ')
+        print('\n\n' + '=' * 10)
+        print('Final result: ')
         final = False
       print(chunk.message.content, end='', flush=True)
+
     if chunk.message.thinking:
+      # accumulate thinking
+      thinking += chunk.message.thinking
       print(chunk.message.thinking, end='', flush=True)
 
+  if thinking != '' or content != '' or len(tool_calls) > 0:
+    messages.append({'role': 'assistant', 'thinking': thinking, 'content': content, 'tool_calls': tool_calls})
+
   print()
 
-  if chunk.message.tool_calls:
-    for tool_call in chunk.message.tool_calls:
+  if tool_calls:
+    for tool_call in tool_calls:
       function_to_call = available_tools.get(tool_call.function.name)
       if function_to_call:
-        print('\nCalling tool: ', tool_call.function.name, 'with arguments: ', tool_call.function.arguments)
+        print('\nCalling tool:', tool_call.function.name, 'with arguments: ', tool_call.function.arguments)
         result = function_to_call(**tool_call.function.arguments)
         print('Tool result: ', result + '\n')
 
-        messages.append(chunk.message)
-        messages.append({'role': 'tool', 'content': result, 'tool_name': tool_call.function.name})
+        result_message = {'role': 'tool', 'content': result, 'tool_name': tool_call.function.name}
+        messages.append(result_message)
       else:
         print(f'Tool {tool_call.function.name} not found')
+        messages.append({'role': 'tool', 'content': f'Tool {tool_call.function.name} not found', 'tool_name': tool_call.function.name})
 
   else:
     # no more tool calls, we can stop the loop

examples/gpt-oss-tools.py

@@ -1,6 +1,16 @@
+# /// script
+# requires-python = ">=3.11"
+# dependencies = [
+#     "gpt-oss",
+#     "ollama",
+#     "rich",
+# ]
+# ///
 import random
 
-from ollama import chat
+from rich import print
+
+from ollama import Client
 from ollama._types import ChatResponse
 
 
@@ -40,11 +50,15 @@ available_tools = {'get_weather': get_weather, 'get_weather_conditions': get_wea
 messages = [{'role': 'user', 'content': 'What is the weather like in London? What are the conditions in Toronto?'}]
 
 
+client = Client(
+  # Ollama Turbo
+  # host="https://ollama.com", headers={'Authorization': (os.getenv('OLLAMA_API_KEY'))}
+)
 model = 'gpt-oss:20b'
 # gpt-oss can call tools while "thinking"
 # a loop is needed to call the tools and get the results
 while True:
-  response: ChatResponse = chat(model=model, messages=messages, tools=[get_weather, get_weather_conditions])
+  response: ChatResponse = client.chat(model=model, messages=messages, tools=[get_weather, get_weather_conditions])
 
   if response.message.content:
     print('Content: ')
@@ -53,18 +67,18 @@ while True:
     print('Thinking: ')
     print(response.message.thinking + '\n')
 
+  messages.append(response.message)
+
   if response.message.tool_calls:
     for tool_call in response.message.tool_calls:
       function_to_call = available_tools.get(tool_call.function.name)
       if function_to_call:
         result = function_to_call(**tool_call.function.arguments)
         print('Result from tool call name: ', tool_call.function.name, 'with arguments: ', tool_call.function.arguments, 'result: ', result + '\n')
-
-        messages.append(response.message)
         messages.append({'role': 'tool', 'content': result, 'tool_name': tool_call.function.name})
       else:
         print(f'Tool {tool_call.function.name} not found')
-
+        messages.append({'role': 'tool', 'content': f'Tool {tool_call.function.name} not found', 'tool_name': tool_call.function.name})
   else:
     # no more tool calls, we can stop the loop
     break

examples/thinking-levels.py

@@ -0,0 +1,26 @@
+from ollama import chat
+
+
+def heading(text):
+  print(text)
+  print('=' * len(text))
+
+
+messages = [
+  {'role': 'user', 'content': 'What is 10 + 23?'},
+]
+
+# gpt-oss supports 'low', 'medium', 'high'
+levels = ['low', 'medium', 'high']
+for i, level in enumerate(levels):
+  response = chat('gpt-oss:20b', messages=messages, think=level)
+
+  heading(f'Thinking ({level})')
+  print(response.message.thinking)
+  print('\n')
+  heading('Response')
+  print(response.message.content)
+  print('\n')
+  if i < len(levels) - 1:
+    print('-' * 20)
+    print('\n')

ollama/_client.py

@@ -274,7 +274,7 @@ class Client(BaseClient):
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: Literal[False] = False,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
     options: Optional[Union[Mapping[str, Any], Options]] = None,
     keep_alive: Optional[Union[float, str]] = None,
@@ -288,7 +288,7 @@ class Client(BaseClient):
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: Literal[True] = True,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
     options: Optional[Union[Mapping[str, Any], Options]] = None,
     keep_alive: Optional[Union[float, str]] = None,
@@ -301,7 +301,7 @@ class Client(BaseClient):
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: bool = False,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
     options: Optional[Union[Mapping[str, Any], Options]] = None,
     keep_alive: Optional[Union[float, str]] = None,
@@ -363,6 +363,7 @@ class Client(BaseClient):
     truncate: Optional[bool] = None,
     options: Optional[Union[Mapping[str, Any], Options]] = None,
     keep_alive: Optional[Union[float, str]] = None,
+    dimensions: Optional[int] = None,
   ) -> EmbedResponse:
     return self._request(
       EmbedResponse,
@@ -374,6 +375,7 @@ class Client(BaseClient):
         truncate=truncate,
         options=options,
         keep_alive=keep_alive,
+        dimensions=dimensions,
       ).model_dump(exclude_none=True),
     )
 
@@ -702,7 +704,7 @@ class AsyncClient(BaseClient):
     template: str = '',
     context: Optional[Sequence[int]] = None,
     stream: Literal[False] = False,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     raw: bool = False,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
     images: Optional[Sequence[Union[str, bytes, Image]]] = None,
@@ -721,7 +723,7 @@ class AsyncClient(BaseClient):
     template: str = '',
     context: Optional[Sequence[int]] = None,
     stream: Literal[True] = True,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     raw: bool = False,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
     images: Optional[Sequence[Union[str, bytes, Image]]] = None,
@@ -739,7 +741,7 @@ class AsyncClient(BaseClient):
     template: Optional[str] = None,
     context: Optional[Sequence[int]] = None,
     stream: bool = False,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     raw: Optional[bool] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
     images: Optional[Sequence[Union[str, bytes, Image]]] = None,
@@ -785,7 +787,7 @@ class AsyncClient(BaseClient):
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: Literal[False] = False,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
     options: Optional[Union[Mapping[str, Any], Options]] = None,
     keep_alive: Optional[Union[float, str]] = None,
@@ -799,7 +801,7 @@ class AsyncClient(BaseClient):
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: Literal[True] = True,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
     options: Optional[Union[Mapping[str, Any], Options]] = None,
     keep_alive: Optional[Union[float, str]] = None,
@@ -812,7 +814,7 @@ class AsyncClient(BaseClient):
     *,
     tools: Optional[Sequence[Union[Mapping[str, Any], Tool, Callable]]] = None,
     stream: bool = False,
-    think: Optional[bool] = None,
+    think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None,
     format: Optional[Union[Literal['', 'json'], JsonSchemaValue]] = None,
     options: Optional[Union[Mapping[str, Any], Options]] = None,
     keep_alive: Optional[Union[float, str]] = None,
@@ -875,6 +877,7 @@ class AsyncClient(BaseClient):
     truncate: Optional[bool] = None,
     options: Optional[Union[Mapping[str, Any], Options]] = None,
     keep_alive: Optional[Union[float, str]] = None,
+    dimensions: Optional[int] = None,
   ) -> EmbedResponse:
     return await self._request(
       EmbedResponse,
@@ -886,6 +889,7 @@ class AsyncClient(BaseClient):
         truncate=truncate,
         options=options,
         keep_alive=keep_alive,
+        dimensions=dimensions,
       ).model_dump(exclude_none=True),
     )
 

ollama/_types.py

@@ -207,7 +207,7 @@ class GenerateRequest(BaseGenerateRequest):
   images: Optional[Sequence[Image]] = None
   'Image data for multimodal models.'
 
-  think: Optional[bool] = None
+  think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None
   'Enable thinking mode (for thinking models).'
 
 
@@ -357,7 +357,7 @@ class ChatRequest(BaseGenerateRequest):
   tools: Optional[Sequence[Tool]] = None
   'Tools to use for the chat.'
 
-  think: Optional[bool] = None
+  think: Optional[Union[bool, Literal['low', 'medium', 'high']]] = None
   'Enable thinking mode (for thinking models).'
 
 
@@ -382,6 +382,9 @@ class EmbedRequest(BaseRequest):
 
   keep_alive: Optional[Union[float, str]] = None
 
+  dimensions: Optional[int] = None
+  'Dimensions truncates the output embedding to the specified dimension.'
+
 
 class EmbedResponse(BaseGenerateResponse):
   """