release(core): 0.3.78 (#33253)

fix #33206

.github/workflows/_release.yml

@@ -431,7 +431,7 @@ jobs:
             git ls-remote --tags origin "langchain-${{ matrix.partner }}*" \
             | awk '{print $2}' \
             | sed 's|refs/tags/||' \
-            | grep -E '[0-9]+\.[0-9]+\.[0-9]+$' \
+            | grep -E '==0\.3\.[0-9]+$' \
             | sort -Vr \
             | head -n 1
           )"

docs/README.md

@@ -34,7 +34,7 @@ Currently, the build process roughly follows these steps:
 
     The docs site is served by Vercel. The Vercel deployment process copies the HTML
     files from the `langchain-api-docs-html` repository and deploys them to the live
-    site. Deployments are triggered on each new commit pushed to `master`.
+    site. Deployments are triggered on each new commit pushed to `v0.3`.
 
 #### Build Technical Details
 

docs/docusaurus.config.js

@@ -87,7 +87,7 @@ const config = {
       ({
         docs: {
           editUrl:
-            "https://github.com/langchain-ai/langchain/edit/master/docs/",
+            "https://github.com/langchain-ai/langchain/edit/v0.3/docs/",
           sidebarPath: require.resolve("./sidebars.js"),
           remarkPlugins: [
             [require("@docusaurus/remark-plugin-npm2yarn"), { sync: true }],

docs/ignore-step.sh

@@ -16,14 +16,13 @@ fi
 
 if  { \
         [ "$VERCEL_ENV" == "production" ] || \
-        [ "$VERCEL_GIT_COMMIT_REF" == "master" ] || \
         [ "$VERCEL_GIT_COMMIT_REF" == "v0.1" ] || \
         [ "$VERCEL_GIT_COMMIT_REF" == "v0.2" ] || \
         [ "$VERCEL_GIT_COMMIT_REF" == "v0.3rc" ]; \
     } && [ "$VERCEL_GIT_REPO_OWNER" == "langchain-ai" ]
 then
-     echo "✅ Production build - proceeding with build"
-     exit 1
+    echo "✅ Production build - proceeding with build"
+    exit 1
 fi
 
 

libs/core/langchain_core/messages/utils.py

@@ -1040,6 +1040,7 @@ def convert_to_openai_messages(
     messages: Union[MessageLikeRepresentation, Sequence[MessageLikeRepresentation]],
     *,
     text_format: Literal["string", "block"] = "string",
+    include_id: bool = False,
 ) -> Union[dict, list[dict]]:
     """Convert LangChain messages into OpenAI message dicts.
 
@@ -1057,6 +1058,8 @@ def convert_to_openai_messages(
                     If a message has a string content, this is turned into a list
                     with a single content block of type ``'text'``. If a message has
                     content blocks these are left as is.
+        include_id: Whether to include message ids in the openai messages, if they
+                    are present in the source messages.
 
     Raises:
         ValueError: if an unrecognized ``text_format`` is specified, or if a message
@@ -1145,6 +1148,8 @@ def convert_to_openai_messages(
             oai_msg["refusal"] = message.additional_kwargs["refusal"]
         if isinstance(message, ToolMessage):
             oai_msg["tool_call_id"] = message.tool_call_id
+        if include_id and message.id:
+            oai_msg["id"] = message.id
 
         if not message.content:
             content = "" if text_format == "string" else []

libs/core/langchain_core/version.py

@@ -1,3 +1,3 @@
 """langchain-core version information and utilities."""
 
-VERSION = "0.3.77"
+VERSION = "0.3.78"

libs/core/pyproject.toml

@@ -16,7 +16,7 @@ dependencies = [
     "pydantic>=2.7.4,<3.0.0",
 ]
 name = "langchain-core"
-version = "0.3.77"
+version = "0.3.78"
 description = "Building applications with LLMs through composability"
 readme = "README.md"
 

libs/core/tests/unit_tests/messages/test_utils.py

@@ -882,10 +882,21 @@ def test_convert_to_openai_messages_string() -> None:
 
 
 def test_convert_to_openai_messages_single_message() -> None:
-    message = HumanMessage(content="Hello")
+    message: BaseMessage = HumanMessage(content="Hello")
     result = convert_to_openai_messages(message)
     assert result == {"role": "user", "content": "Hello"}
 
+    # Test IDs
+    result = convert_to_openai_messages(message, include_id=True)
+    assert result == {"role": "user", "content": "Hello"}  # no ID
+
+    message = AIMessage(content="Hello", id="resp_123")
+    result = convert_to_openai_messages(message)
+    assert result == {"role": "assistant", "content": "Hello"}
+
+    result = convert_to_openai_messages(message, include_id=True)
+    assert result == {"role": "assistant", "content": "Hello", "id": "resp_123"}
+
 
 def test_convert_to_openai_messages_multiple_messages() -> None:
     messages = [

libs/core/uv.lock

@@ -958,7 +958,7 @@ wheels = [
 
 [[package]]
 name = "langchain-core"
-version = "0.3.77"
+version = "0.3.78"
 source = { editable = "." }
 dependencies = [
     { name = "jsonpatch" },
@@ -1050,7 +1050,7 @@ typing = [
 
 [[package]]
 name = "langchain-tests"
-version = "0.3.21"
+version = "0.3.22"
 source = { directory = "../standard-tests" }
 dependencies = [
     { name = "httpx" },

libs/partners/ollama/langchain_ollama/chat_models.py

@@ -648,26 +648,30 @@ class ChatOllama(BaseChatModel):
         if self.stop is not None:
             stop = self.stop
 
-        options_dict = kwargs.pop(
-            "options",
-            {
-                "mirostat": self.mirostat,
-                "mirostat_eta": self.mirostat_eta,
-                "mirostat_tau": self.mirostat_tau,
-                "num_ctx": self.num_ctx,
-                "num_gpu": self.num_gpu,
-                "num_thread": self.num_thread,
-                "num_predict": self.num_predict,
-                "repeat_last_n": self.repeat_last_n,
-                "repeat_penalty": self.repeat_penalty,
-                "temperature": self.temperature,
-                "seed": self.seed,
-                "stop": self.stop if stop is None else stop,
-                "tfs_z": self.tfs_z,
-                "top_k": self.top_k,
-                "top_p": self.top_p,
-            },
-        )
+        options_dict = kwargs.pop("options", None)
+        if options_dict is None:
+            # Only include parameters that are explicitly set (not None)
+            options_dict = {
+                k: v
+                for k, v in {
+                    "mirostat": self.mirostat,
+                    "mirostat_eta": self.mirostat_eta,
+                    "mirostat_tau": self.mirostat_tau,
+                    "num_ctx": self.num_ctx,
+                    "num_gpu": self.num_gpu,
+                    "num_thread": self.num_thread,
+                    "num_predict": self.num_predict,
+                    "repeat_last_n": self.repeat_last_n,
+                    "repeat_penalty": self.repeat_penalty,
+                    "temperature": self.temperature,
+                    "seed": self.seed,
+                    "stop": self.stop if stop is None else stop,
+                    "tfs_z": self.tfs_z,
+                    "top_k": self.top_k,
+                    "top_p": self.top_p,
+                }.items()
+                if v is not None
+            }
 
         params = {
             "messages": ollama_messages,

libs/partners/ollama/pyproject.toml

@@ -11,7 +11,7 @@ dependencies = [
     "langchain-core>=0.3.76,<2.0.0",
 ]
 name = "langchain-ollama"
-version = "0.3.9"
+version = "0.3.10"
 description = "An integration package connecting Ollama and LangChain"
 readme = "README.md"
 

libs/partners/ollama/tests/unit_tests/test_chat_models.py

@@ -24,7 +24,7 @@ MODEL_NAME = "llama3.1"
 
 @contextmanager
 def _mock_httpx_client_stream(
-    *args: Any, **kwargs: Any
+    *_args: Any, **_kwargs: Any
 ) -> Generator[Response, Any, Any]:
     yield Response(
         status_code=200,
@@ -310,3 +310,103 @@ def test_load_response_with_actual_content_is_not_skipped(
         assert result.content == "This is actual content"
         assert result.response_metadata.get("done_reason") == "load"
         assert not caplog.text
+
+
+def test_none_parameters_excluded_from_options() -> None:
+    """Test that None parameters are excluded from the options dict sent to Ollama."""
+    response = [
+        {
+            "model": "test-model",
+            "created_at": "2025-01-01T00:00:00.000000000Z",
+            "done": True,
+            "done_reason": "stop",
+            "message": {"role": "assistant", "content": "Hello!"},
+        }
+    ]
+
+    with patch("langchain_ollama.chat_models.Client") as mock_client_class:
+        mock_client = MagicMock()
+        mock_client_class.return_value = mock_client
+        mock_client.chat.return_value = response
+
+        # Create ChatOllama with only num_ctx set
+        llm = ChatOllama(model="test-model", num_ctx=4096)
+        llm.invoke([HumanMessage("Hello")])
+
+        # Verify that chat was called
+        assert mock_client.chat.called
+
+        # Get the options dict that was passed to chat
+        call_kwargs = mock_client.chat.call_args[1]
+        options = call_kwargs.get("options", {})
+
+        # Only num_ctx should be in options, not None parameters
+        assert "num_ctx" in options
+        assert options["num_ctx"] == 4096
+
+        # These parameters should NOT be in options since they were None
+        assert "mirostat" not in options
+        assert "mirostat_eta" not in options
+        assert "mirostat_tau" not in options
+        assert "tfs_z" not in options
+
+
+def test_all_none_parameters_results_in_empty_options() -> None:
+    """Test that when all parameters are None, options dict is empty."""
+    response = [
+        {
+            "model": "test-model",
+            "created_at": "2025-01-01T00:00:00.000000000Z",
+            "done": True,
+            "done_reason": "stop",
+            "message": {"role": "assistant", "content": "Hello!"},
+        }
+    ]
+
+    with patch("langchain_ollama.chat_models.Client") as mock_client_class:
+        mock_client = MagicMock()
+        mock_client_class.return_value = mock_client
+        mock_client.chat.return_value = response
+
+        # Create ChatOllama with no parameters set
+        llm = ChatOllama(model="test-model")
+        llm.invoke([HumanMessage("Hello")])
+
+        # Get the options dict that was passed to chat
+        call_kwargs = mock_client.chat.call_args[1]
+        options = call_kwargs.get("options", {})
+
+        # Options should be empty when no parameters are set
+        assert options == {}
+
+
+def test_explicit_options_dict_preserved() -> None:
+    """Test that explicitly provided options dict is preserved and not filtered."""
+    response = [
+        {
+            "model": "test-model",
+            "created_at": "2025-01-01T00:00:00.000000000Z",
+            "done": True,
+            "done_reason": "stop",
+            "message": {"role": "assistant", "content": "Hello!"},
+        }
+    ]
+
+    with patch("langchain_ollama.chat_models.Client") as mock_client_class:
+        mock_client = MagicMock()
+        mock_client_class.return_value = mock_client
+        mock_client.chat.return_value = response
+
+        llm = ChatOllama(model="test-model")
+        # Pass explicit options dict, including None values
+        llm.invoke(
+            [HumanMessage("Hello")],
+            options={"temperature": 0.5, "custom_param": None},
+        )
+
+        # Get the options dict that was passed to chat
+        call_kwargs = mock_client.chat.call_args[1]
+        options = call_kwargs.get("options", {})
+
+        # Explicit options should be preserved as-is
+        assert options == {"temperature": 0.5, "custom_param": None}

libs/partners/ollama/uv.lock

@@ -381,7 +381,7 @@ typing = [
 
 [[package]]
 name = "langchain-ollama"
-version = "0.3.9"
+version = "0.3.10"
 source = { editable = "." }
 dependencies = [
     { name = "langchain-core" },