1

libs/core/langchain_core/language_models/base.py

@@ -70,6 +70,9 @@ class LangSmithParams(TypedDict, total=False):
     ls_stop: list[str] | None
     """Stop words for generation."""
 
+    versions: dict[str, str]
+    """Package versions for tracing (e.g., `{"langchain-anthropic": "1.3.3"}`)."""
+
 
 @cache  # Cache the tokenizer
 def get_tokenizer() -> Any:

libs/core/langchain_core/language_models/chat_models.py

@@ -63,7 +63,11 @@ from langchain_core.outputs.chat_generation import merge_chat_generation_chunks
 from langchain_core.prompt_values import ChatPromptValue, PromptValue, StringPromptValue
 from langchain_core.rate_limiters import BaseRateLimiter
 from langchain_core.runnables import RunnableMap, RunnablePassthrough
-from langchain_core.runnables.config import ensure_config, run_in_executor
+from langchain_core.runnables.config import (
+    _merge_metadata_dicts,
+    ensure_config,
+    run_in_executor,
+)
 from langchain_core.tracers._streaming import _StreamingCallbackHandler
 from langchain_core.utils.function_calling import (
     convert_to_json_schema,
@@ -503,10 +507,10 @@ class BaseChatModel(BaseLanguageModel[AIMessage], ABC):
 
             params = self._get_invocation_params(stop=stop, **kwargs)
             options = {"stop": stop, **kwargs, **ls_structured_output_format_dict}
-            inheritable_metadata = {
-                **(config.get("metadata") or {}),
-                **self._get_ls_params(stop=stop, **kwargs),
-            }
+            inheritable_metadata = _merge_metadata_dicts(
+                config.get("metadata") or {},
+                self._get_ls_params(stop=stop, **kwargs),
+            )
             callback_manager = CallbackManager.configure(
                 config.get("callbacks"),
                 self.callbacks,
@@ -631,10 +635,10 @@ class BaseChatModel(BaseLanguageModel[AIMessage], ABC):
 
         params = self._get_invocation_params(stop=stop, **kwargs)
         options = {"stop": stop, **kwargs, **ls_structured_output_format_dict}
-        inheritable_metadata = {
-            **(config.get("metadata") or {}),
-            **self._get_ls_params(stop=stop, **kwargs),
-        }
+        inheritable_metadata = _merge_metadata_dicts(
+            config.get("metadata") or {},
+            self._get_ls_params(stop=stop, **kwargs),
+        )
         callback_manager = AsyncCallbackManager.configure(
             config.get("callbacks"),
             self.callbacks,
@@ -895,10 +899,10 @@ class BaseChatModel(BaseLanguageModel[AIMessage], ABC):
 
         params = self._get_invocation_params(stop=stop, **kwargs)
         options = {"stop": stop, **ls_structured_output_format_dict}
-        inheritable_metadata = {
-            **(metadata or {}),
-            **self._get_ls_params(stop=stop, **kwargs),
-        }
+        inheritable_metadata = _merge_metadata_dicts(
+            metadata or {},
+            self._get_ls_params(stop=stop, **kwargs),
+        )
 
         callback_manager = CallbackManager.configure(
             callbacks,
@@ -1018,10 +1022,10 @@ class BaseChatModel(BaseLanguageModel[AIMessage], ABC):
 
         params = self._get_invocation_params(stop=stop, **kwargs)
         options = {"stop": stop, **ls_structured_output_format_dict}
-        inheritable_metadata = {
-            **(metadata or {}),
-            **self._get_ls_params(stop=stop, **kwargs),
-        }
+        inheritable_metadata = _merge_metadata_dicts(
+            metadata or {},
+            self._get_ls_params(stop=stop, **kwargs),
+        )
 
         callback_manager = AsyncCallbackManager.configure(
             callbacks,

libs/core/langchain_core/runnables/config.py

@@ -7,7 +7,15 @@ import asyncio
 # Cannot move uuid to TYPE_CHECKING as RunnableConfig is used in Pydantic models
 import uuid  # noqa: TC003
 import warnings
-from collections.abc import Awaitable, Callable, Generator, Iterable, Iterator, Sequence
+from collections.abc import (
+    Awaitable,
+    Callable,
+    Generator,
+    Iterable,
+    Iterator,
+    Mapping,
+    Sequence,
+)
 from concurrent.futures import Executor, Future, ThreadPoolExecutor
 from contextlib import contextmanager
 from contextvars import Context, ContextVar, Token, copy_context
@@ -354,6 +362,31 @@ def patch_config(
     return config
 
 
+def _merge_metadata_dicts(
+    base: Mapping[str, Any], incoming: Mapping[str, Any]
+) -> dict[str, Any]:
+    """Merge two metadata dicts with one extra level of depth.
+
+    If both sides have a dict value for the same key, the inner dicts are merged
+    (last-writer-wins within). Non-dict values use last-writer-wins at the
+    top level.
+
+    Args:
+        base: The base metadata dict.
+        incoming: The incoming metadata dict to merge on top.
+
+    Returns:
+        A new merged dict. Does not mutate inputs.
+    """
+    merged = {**base}
+    for key, value in incoming.items():
+        if key in merged and isinstance(merged[key], dict) and isinstance(value, dict):
+            merged[key] = {**merged[key], **value}
+        else:
+            merged[key] = value
+    return merged
+
+
 def merge_configs(*configs: RunnableConfig | None) -> RunnableConfig:
     """Merge multiple configs into one.
 
@@ -369,10 +402,10 @@ def merge_configs(*configs: RunnableConfig | None) -> RunnableConfig:
     for config in (ensure_config(c) for c in configs if c is not None):
         for key in config:
             if key == "metadata":
-                base["metadata"] = {
-                    **base.get("metadata", {}),
-                    **(config.get("metadata") or {}),
-                }
+                base["metadata"] = _merge_metadata_dicts(
+                    base.get("metadata", {}),
+                    config.get("metadata") or {},
+                )
             elif key == "tags":
                 base["tags"] = sorted(
                     set(base.get("tags", []) + (config.get("tags") or [])),

libs/core/tests/unit_tests/runnables/test_config.py

@@ -16,6 +16,7 @@ from langchain_core.callbacks.streaming_stdout import StreamingStdOutCallbackHan
 from langchain_core.runnables import RunnableBinding, RunnablePassthrough
 from langchain_core.runnables.config import (
     RunnableConfig,
+    _merge_metadata_dicts,
     _set_config_context,
     ensure_config,
     merge_configs,
@@ -161,3 +162,56 @@ async def test_run_in_executor() -> None:
 
     with pytest.raises(RuntimeError):
         await run_in_executor(None, raises_stop_iter)
+
+
+class TestMergeMetadataDicts:
+    """Tests for _merge_metadata_dicts deep-merge behavior."""
+
+    def test_deep_merge_preserves_both_nested_dicts(self) -> None:
+        base = {"versions": {"langchain-core": "0.3.1"}, "user_id": "abc"}
+        incoming = {"versions": {"langchain-anthropic": "1.3.3"}, "run": "x"}
+        result = _merge_metadata_dicts(base, incoming)
+        assert result == {
+            "versions": {
+                "langchain-core": "0.3.1",
+                "langchain-anthropic": "1.3.3",
+            },
+            "user_id": "abc",
+            "run": "x",
+        }
+
+    def test_last_writer_wins_within_nested_dicts(self) -> None:
+        base = {"versions": {"pkg": "1.0"}}
+        incoming = {"versions": {"pkg": "2.0"}}
+        result = _merge_metadata_dicts(base, incoming)
+        assert result == {"versions": {"pkg": "2.0"}}
+
+    def test_non_dict_overwrites_dict(self) -> None:
+        base = {"key": {"nested": "value"}}
+        incoming = {"key": "flat"}
+        result = _merge_metadata_dicts(base, incoming)
+        assert result == {"key": "flat"}
+
+    def test_dict_overwrites_non_dict(self) -> None:
+        base = {"key": "flat"}
+        incoming = {"key": {"nested": "value"}}
+        result = _merge_metadata_dicts(base, incoming)
+        assert result == {"key": {"nested": "value"}}
+
+    def test_no_mutation_of_inputs(self) -> None:
+        base = {"versions": {"a": "1"}}
+        incoming = {"versions": {"b": "2"}}
+        base_copy = {"versions": {"a": "1"}}
+        incoming_copy = {"versions": {"b": "2"}}
+        _merge_metadata_dicts(base, incoming)
+        assert base == base_copy
+        assert incoming == incoming_copy
+
+    def test_three_config_merge_accumulates(self) -> None:
+        c1: RunnableConfig = {"metadata": {"versions": {"a": "1"}}}
+        c2: RunnableConfig = {"metadata": {"versions": {"b": "2"}}}
+        c3: RunnableConfig = {"metadata": {"versions": {"c": "3"}}}
+        merged = merge_configs(c1, c2, c3)
+        assert merged["metadata"] == {
+            "versions": {"a": "1", "b": "2", "c": "3"},
+        }

libs/partners/anthropic/langchain_anthropic/chat_models.py

@@ -998,6 +998,7 @@ class ChatAnthropic(BaseChatModel):
             ls_params["ls_max_tokens"] = ls_max_tokens
         if ls_stop := stop or params.get("stop", None):
             ls_params["ls_stop"] = ls_stop
+        ls_params["versions"] = {"langchain-anthropic": __version__}
         return ls_params
 
     @model_validator(mode="before")

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

@@ -22,6 +22,7 @@ from pydantic import BaseModel, Field, SecretStr, ValidationError
 from pytest import CaptureFixture, MonkeyPatch
 
 from langchain_anthropic import ChatAnthropic
+from langchain_anthropic._version import __version__
 from langchain_anthropic.chat_models import (
     _create_usage_metadata,
     _format_image,
@@ -1677,12 +1678,21 @@ def test_anthropic_model_params() -> None:
         "ls_model_name": MODEL_NAME,
         "ls_max_tokens": 64000,
         "ls_temperature": None,
+        "versions": {"langchain-anthropic": __version__},
     }
 
     ls_params = llm._get_ls_params(model=MODEL_NAME)
     assert ls_params.get("ls_model_name") == MODEL_NAME
 
 
+def test_ls_params_versions_value() -> None:
+    """Test that _get_ls_params reports the correct langchain-anthropic version."""
+    llm = ChatAnthropic(model=MODEL_NAME)
+    ls_params = llm._get_ls_params()
+    assert "versions" in ls_params
+    assert ls_params["versions"] == {"langchain-anthropic": __version__}
+
+
 def test_streaming_cache_token_reporting() -> None:
     """Test that cache tokens are properly reported in streaming events."""
     from unittest.mock import MagicMock