release: v1.0.0 (#32567)

Co-authored-by: Mohammad Mohtashim <45242107+keenborder786@users.noreply.github.com>
Co-authored-by: Caspar Broekhuizen <caspar@langchain.dev>
Co-authored-by: ccurme <chester.curme@gmail.com>
Co-authored-by: Christophe Bornet <cbornet@hotmail.com>
Co-authored-by: Eugene Yurtsev <eyurtsev@gmail.com>
Co-authored-by: Sadra Barikbin <sadraqazvin1@yahoo.com>
Co-authored-by: Vadym Barda <vadim.barda@gmail.com>

libs/partners/ollama/tests/integration_tests/chat_models/test_chat_models.py

@@ -7,9 +7,9 @@ from unittest.mock import MagicMock, patch
 
 import pytest
 from httpx import ConnectError
-from langchain_core.messages.ai import AIMessageChunk
+from langchain_core.messages.ai import AIMessage, AIMessageChunk
 from langchain_core.messages.human import HumanMessage
-from langchain_core.messages.tool import ToolCallChunk
+from langchain_core.messages.tool import ToolCallChunk, ToolMessage
 from langchain_core.tools import tool
 from ollama import ResponseError
 from pydantic import BaseModel, Field, ValidationError
@@ -18,6 +18,7 @@ from typing_extensions import TypedDict
 from langchain_ollama import ChatOllama
 
 DEFAULT_MODEL_NAME = "llama3.1"
+REASONING_MODEL_NAME = "gpt-oss:20b"
 
 
 @tool
@@ -236,3 +237,59 @@ async def test_tool_astreaming(model: str) -> None:
         if chunk.get("id")
     )
     assert final_tool_call["id"] == tool_call_id
+
+
+@pytest.mark.parametrize(
+    ("model", "output_version"),
+    [(REASONING_MODEL_NAME, None), (REASONING_MODEL_NAME, "v1")],
+)
+def test_agent_loop(model: str, output_version: Optional[str]) -> None:
+    """Test agent loop with tool calling and message passing."""
+
+    @tool
+    def get_weather(location: str) -> str:
+        """Get the weather for a location."""
+        return "It's sunny and 75 degrees."
+
+    llm = ChatOllama(model=model, output_version=output_version, reasoning="low")
+    llm_with_tools = llm.bind_tools([get_weather])
+
+    input_message = HumanMessage("What is the weather in San Francisco, CA?")
+    tool_call_message = llm_with_tools.invoke([input_message])
+    assert isinstance(tool_call_message, AIMessage)
+
+    tool_calls = tool_call_message.tool_calls
+    assert len(tool_calls) == 1
+    tool_call = tool_calls[0]
+    assert tool_call["name"] == "get_weather"
+    assert "location" in tool_call["args"]
+
+    tool_message = get_weather.invoke(tool_call)
+    assert isinstance(tool_message, ToolMessage)
+    assert tool_message.content
+    assert isinstance(tool_message.content, str)
+    assert "sunny" in tool_message.content.lower()
+
+    resp_message = llm_with_tools.invoke(
+        [
+            input_message,
+            tool_call_message,
+            tool_message,
+        ]
+    )
+    follow_up = HumanMessage("Explain why that might be using a reasoning step.")
+    assert isinstance(resp_message, AIMessage)
+    assert len(resp_message.content) > 0
+
+    response = llm_with_tools.invoke(
+        [input_message, tool_call_message, tool_message, resp_message, follow_up]
+    )
+    assert isinstance(resp_message, AIMessage)
+    assert len(resp_message.content) > 0
+
+    if output_version == "v1":
+        content_blocks = response.content_blocks
+        assert content_blocks is not None
+        assert len(content_blocks) > 0
+        assert any(block["type"] == "text" for block in content_blocks)
+        assert any(block["type"] == "reasoning" for block in content_blocks)

libs/partners/ollama/tests/integration_tests/chat_models/test_chat_models_reasoning.py

@@ -7,8 +7,10 @@ from langchain_ollama import ChatOllama
 
 SAMPLE = "What is 3^3?"
 
+REASONING_MODEL_NAME = "deepseek-r1:1.5b"
 
-@pytest.mark.parametrize("model", ["deepseek-r1:1.5b"])
+
+@pytest.mark.parametrize("model", [REASONING_MODEL_NAME])
 @pytest.mark.parametrize("use_async", [False, True])
 async def test_stream_no_reasoning(model: str, use_async: bool) -> None:
     """Test streaming with ``reasoning=False``."""
@@ -41,7 +43,7 @@ async def test_stream_no_reasoning(model: str, use_async: bool) -> None:
     assert "reasoning_content" not in result.additional_kwargs
 
 
-@pytest.mark.parametrize("model", ["deepseek-r1:1.5b"])
+@pytest.mark.parametrize("model", [REASONING_MODEL_NAME])
 @pytest.mark.parametrize("use_async", [False, True])
 async def test_stream_reasoning_none(model: str, use_async: bool) -> None:
     """Test streaming with ``reasoning=None``."""
@@ -76,7 +78,7 @@ async def test_stream_reasoning_none(model: str, use_async: bool) -> None:
     assert "</think>" not in result.additional_kwargs.get("reasoning_content", "")
 
 
-@pytest.mark.parametrize("model", ["deepseek-r1:1.5b"])
+@pytest.mark.parametrize("model", [REASONING_MODEL_NAME])
 @pytest.mark.parametrize("use_async", [False, True])
 async def test_reasoning_stream(model: str, use_async: bool) -> None:
     """Test streaming with ``reasoning=True``."""
@@ -111,8 +113,16 @@ async def test_reasoning_stream(model: str, use_async: bool) -> None:
     assert "<think>" not in result.additional_kwargs["reasoning_content"]
     assert "</think>" not in result.additional_kwargs["reasoning_content"]
 
+    content_blocks = result.content_blocks
+    assert content_blocks is not None
+    assert len(content_blocks) > 0
+    reasoning_blocks = [
+        block for block in content_blocks if block.get("type") == "reasoning"
+    ]
+    assert len(reasoning_blocks) > 0
 
-@pytest.mark.parametrize("model", ["deepseek-r1:1.5b"])
+
+@pytest.mark.parametrize("model", [REASONING_MODEL_NAME])
 @pytest.mark.parametrize("use_async", [False, True])
 async def test_invoke_no_reasoning(model: str, use_async: bool) -> None:
     """Test invoke with ``reasoning=False``."""
@@ -128,7 +138,7 @@ async def test_invoke_no_reasoning(model: str, use_async: bool) -> None:
     assert "</think>" not in result.content
 
 
-@pytest.mark.parametrize("model", ["deepseek-r1:1.5b"])
+@pytest.mark.parametrize("model", [REASONING_MODEL_NAME])
 @pytest.mark.parametrize("use_async", [False, True])
 async def test_invoke_reasoning_none(model: str, use_async: bool) -> None:
     """Test invoke with ``reasoning=None``."""
@@ -146,7 +156,7 @@ async def test_invoke_reasoning_none(model: str, use_async: bool) -> None:
     assert "</think>" not in result.additional_kwargs.get("reasoning_content", "")
 
 
-@pytest.mark.parametrize("model", ["deepseek-r1:1.5b"])
+@pytest.mark.parametrize("model", [REASONING_MODEL_NAME])
 @pytest.mark.parametrize("use_async", [False, True])
 async def test_reasoning_invoke(model: str, use_async: bool) -> None:
     """Test invoke with ``reasoning=True``."""
@@ -164,8 +174,16 @@ async def test_reasoning_invoke(model: str, use_async: bool) -> None:
     assert "<think>" not in result.additional_kwargs["reasoning_content"]
     assert "</think>" not in result.additional_kwargs["reasoning_content"]
 
+    content_blocks = result.content_blocks
+    assert content_blocks is not None
+    assert len(content_blocks) > 0
+    reasoning_blocks = [
+        block for block in content_blocks if block.get("type") == "reasoning"
+    ]
+    assert len(reasoning_blocks) > 0
 
-@pytest.mark.parametrize("model", ["deepseek-r1:1.5b"])
+
+@pytest.mark.parametrize("model", [REASONING_MODEL_NAME])
 def test_think_tag_stripping_necessity(model: str) -> None:
     """Test that demonstrates why ``_strip_think_tags`` is necessary.