nvidia-nat-langchain 1.4.0a20251031__py3-none-any.whl → 1.4.0a20251220__py3-none-any.whl

nat/plugins/langchain/embedder.py

@@ -28,7 +28,13 @@ async def azure_openai_langchain(embedder_config: AzureOpenAIEmbedderModelConfig
 
     from langchain_openai import AzureOpenAIEmbeddings
 
-    client = AzureOpenAIEmbeddings(**embedder_config.model_dump(exclude={"type"}, by_alias=True, exclude_none=True))
+    client = AzureOpenAIEmbeddings(
+        **embedder_config.model_dump(exclude={"type", "api_version"},
+                                     by_alias=True,
+                                     exclude_none=True,
+                                     exclude_unset=True),
+        api_version=embedder_config.api_version,
+    )
 
     if isinstance(embedder_config, RetryMixin):
         client = patch_with_retry(client,
@@ -44,7 +50,8 @@ async def nim_langchain(embedder_config: NIMEmbedderModelConfig, builder: Builde
 
     from langchain_nvidia_ai_endpoints import NVIDIAEmbeddings
 
-    client = NVIDIAEmbeddings(**embedder_config.model_dump(exclude={"type"}, by_alias=True, exclude_none=True))
+    client = NVIDIAEmbeddings(
+        **embedder_config.model_dump(exclude={"type"}, by_alias=True, exclude_none=True, exclude_unset=True))
 
     if isinstance(embedder_config, RetryMixin):
         client = patch_with_retry(client,
@@ -60,7 +67,8 @@ async def openai_langchain(embedder_config: OpenAIEmbedderModelConfig, builder:
 
     from langchain_openai import OpenAIEmbeddings
 
-    client = OpenAIEmbeddings(**embedder_config.model_dump(exclude={"type"}, by_alias=True, exclude_none=True))
+    client = OpenAIEmbeddings(
+        **embedder_config.model_dump(exclude={"type"}, by_alias=True, exclude_none=True, exclude_unset=True))
 
     if isinstance(embedder_config, RetryMixin):
         client = patch_with_retry(client,

nat/plugins/langchain/llm.py

@@ -27,6 +27,8 @@ from nat.data_models.retry_mixin import RetryMixin
 from nat.data_models.thinking_mixin import ThinkingMixin
 from nat.llm.aws_bedrock_llm import AWSBedrockModelConfig
 from nat.llm.azure_openai_llm import AzureOpenAIModelConfig
+from nat.llm.dynamo_llm import DynamoModelConfig
+from nat.llm.dynamo_llm import create_httpx_client_with_dynamo_hooks
 from nat.llm.litellm_llm import LiteLlmModelConfig
 from nat.llm.nim_llm import NIMModelConfig
 from nat.llm.openai_llm import OpenAIModelConfig
@@ -122,6 +124,7 @@ async def aws_bedrock_langchain(llm_config: AWSBedrockModelConfig, _builder: Bui
         exclude={"type", "context_size", "thinking", "api_type"},
         by_alias=True,
         exclude_none=True,
+        exclude_unset=True,
     ))
 
     yield _patch_llm_based_on_config(client, llm_config)
@@ -135,7 +138,12 @@ async def azure_openai_langchain(llm_config: AzureOpenAIModelConfig, _builder: B
     validate_no_responses_api(llm_config, LLMFrameworkEnum.LANGCHAIN)
 
     client = AzureChatOpenAI(
-        **llm_config.model_dump(exclude={"type", "thinking", "api_type"}, by_alias=True, exclude_none=True))
+        **llm_config.model_dump(exclude={"type", "thinking", "api_type", "api_version"},
+                                by_alias=True,
+                                exclude_none=True,
+                                exclude_unset=True),
+        api_version=llm_config.api_version,
+    )
 
     yield _patch_llm_based_on_config(client, llm_config)
 
@@ -149,9 +157,12 @@ async def nim_langchain(llm_config: NIMModelConfig, _builder: Builder):
 
     # prefer max_completion_tokens over max_tokens
     client = ChatNVIDIA(
-        **llm_config.model_dump(exclude={"type", "max_tokens", "thinking", "api_type"},
-                                by_alias=True,
-                                exclude_none=True),
+        **llm_config.model_dump(
+            exclude={"type", "max_tokens", "thinking", "api_type"},
+            by_alias=True,
+            exclude_none=True,
+            exclude_unset=True,
+        ),
         max_completion_tokens=llm_config.max_tokens,
     )
 
@@ -171,6 +182,7 @@ async def openai_langchain(llm_config: OpenAIModelConfig, _builder: Builder):
                                 exclude={"type", "thinking", "api_type"},
                                 by_alias=True,
                                 exclude_none=True,
+                                exclude_unset=True,
                             ))
     else:
         # If stream_usage is specified, it will override the default value of True.
@@ -179,11 +191,65 @@ async def openai_langchain(llm_config: OpenAIModelConfig, _builder: Builder):
                                 exclude={"type", "thinking", "api_type"},
                                 by_alias=True,
                                 exclude_none=True,
+                                exclude_unset=True,
                             ))
 
     yield _patch_llm_based_on_config(client, llm_config)
 
 
+@register_llm_client(config_type=DynamoModelConfig, wrapper_type=LLMFrameworkEnum.LANGCHAIN)
+async def dynamo_langchain(llm_config: DynamoModelConfig, _builder: Builder):
+    """
+    Create a LangChain ChatOpenAI client for Dynamo with automatic prefix header injection.
+
+    This client injects Dynamo prefix headers at the HTTP transport level using httpx event hooks,
+    enabling KV cache optimization and request routing.
+    """
+    from langchain_openai import ChatOpenAI
+
+    # Build config dict excluding Dynamo-specific and NAT-specific fields
+    config_dict = llm_config.model_dump(
+        exclude={"type", "thinking", "api_type", *DynamoModelConfig.get_dynamo_field_names()},
+        by_alias=True,
+        exclude_none=True,
+        exclude_unset=True,
+    )
+
+    # Initialize http_async_client to None for proper cleanup
+    http_async_client = None
+
+    try:
+        # If prefix_template is set, create a custom httpx client with Dynamo hooks
+        if llm_config.prefix_template is not None:
+            http_async_client = create_httpx_client_with_dynamo_hooks(
+                prefix_template=llm_config.prefix_template,
+                total_requests=llm_config.prefix_total_requests,
+                osl=llm_config.prefix_osl,
+                iat=llm_config.prefix_iat,
+                timeout=llm_config.request_timeout,
+            )
+            config_dict["http_async_client"] = http_async_client
+            logger.info(
+                "Dynamo prefix headers enabled: template=%s, total_requests=%d, osl=%s, iat=%s",
+                llm_config.prefix_template,
+                llm_config.prefix_total_requests,
+                llm_config.prefix_osl,
+                llm_config.prefix_iat,
+            )
+
+        # Create the ChatOpenAI client
+        if llm_config.api_type == APITypeEnum.RESPONSES:
+            client = ChatOpenAI(stream_usage=True, use_responses_api=True, use_previous_response_id=True, **config_dict)
+        else:
+            client = ChatOpenAI(stream_usage=True, **config_dict)
+
+        yield _patch_llm_based_on_config(client, llm_config)
+    finally:
+        # Ensure the httpx client is properly closed to avoid resource leaks
+        if http_async_client is not None:
+            await http_async_client.aclose()
+
+
 @register_llm_client(config_type=LiteLlmModelConfig, wrapper_type=LLMFrameworkEnum.LANGCHAIN)
 async def litellm_langchain(llm_config: LiteLlmModelConfig, _builder: Builder):
 
@@ -191,7 +257,7 @@ async def litellm_langchain(llm_config: LiteLlmModelConfig, _builder: Builder):
 
     validate_no_responses_api(llm_config, LLMFrameworkEnum.LANGCHAIN)
 
-    client = ChatLiteLLM(
-        **llm_config.model_dump(exclude={"type", "thinking", "api_type"}, by_alias=True, exclude_none=True))
+    client = ChatLiteLLM(**llm_config.model_dump(
+        exclude={"type", "thinking", "api_type"}, by_alias=True, exclude_none=True, exclude_unset=True))
 
     yield _patch_llm_based_on_config(client, llm_config)

nat/plugins/langchain/tools/tavily_internet_search.py

@@ -30,7 +30,8 @@ class TavilyInternetSearchToolConfig(FunctionBaseConfig, name="tavily_internet_s
     Requires a TAVILY_API_KEY.
     """
     max_results: int = 3
-    api_key: SerializableSecretStr = Field(default="", description="The API key for the Tavily service.")
+    api_key: SerializableSecretStr = Field(default_factory=lambda: SerializableSecretStr(""),
+                                           description="The API key for the Tavily service.")
 
 
 @register_function(config_type=TavilyInternetSearchToolConfig)

{nvidia_nat_langchain-1.4.0a20251031.dist-info → nvidia_nat_langchain-1.4.0a20251220.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nvidia-nat-langchain
-Version: 1.4.0a20251031
+Version: 1.4.0a20251220
 Summary: Subpackage for LangChain/LangGraph integration in NeMo Agent toolkit
 Author: NVIDIA Corporation
 Maintainer: NVIDIA Corporation
@@ -16,7 +16,7 @@ Requires-Python: <3.14,>=3.11
 Description-Content-Type: text/markdown
 License-File: LICENSE-3rd-party.txt
 License-File: LICENSE.md
-Requires-Dist: nvidia-nat==v1.4.0a20251031
+Requires-Dist: nvidia-nat==v1.4.0a20251220
 Requires-Dist: langchain~=0.3.27
 Requires-Dist: langchain-aws~=0.2.31
 Requires-Dist: langchain-core~=0.3.75

{nvidia_nat_langchain-1.4.0a20251031.dist-info → nvidia_nat_langchain-1.4.0a20251220.dist-info}/RECORD

@@ -1,19 +1,19 @@
 nat/meta/pypi.md,sha256=T_KFtTXVxhFM8Y6K3OlNByA5sTXLQuqqUpHgNOCvZBU,1120
 nat/plugins/langchain/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-nat/plugins/langchain/embedder.py,sha256=ZSESaazyz7y3F0GSSsWRe_xfvxOe0Mwd45wEAkQ2jJk,3339
-nat/plugins/langchain/llm.py,sha256=6e-Ug6iahZdOJRLCJmVmQTgSYxu0xmslyANaUQG_qPY,8213
+nat/plugins/langchain/embedder.py,sha256=Ie-J4N4lvygW0zNKklKZVSYxYFcRW6p_QlRdcz0WxcE,3607
+nat/plugins/langchain/llm.py,sha256=JBeYfhD0IGS6W-Jl6vfDqUR8Pdji5zYua2cp3fr_vyw,10950
 nat/plugins/langchain/register.py,sha256=jgq6wSJoGQIZFJhS8RbUs25cLgNJjCkFu4M6qaWJS_4,906
 nat/plugins/langchain/retriever.py,sha256=SWbXXOezEUuPACnmSSU497NAmEVEMj2SrFJGodkRg34,2644
 nat/plugins/langchain/tool_wrapper.py,sha256=Zgb2_XB4bEhjPPeqS-ZH_OJT_pcQmteX7u03N_qCLfc,2121
 nat/plugins/langchain/tools/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 nat/plugins/langchain/tools/code_generation_tool.py,sha256=f5pna0WMOx3QOS4WnaMFKD7tBZ1-tS0PfI0IMYobtTQ,2723
 nat/plugins/langchain/tools/register.py,sha256=uemxqLxcNk1bGX4crV52oMphLTZWonStzkXwTZeG2Rw,889
-nat/plugins/langchain/tools/tavily_internet_search.py,sha256=W5sdZ9hobPc3xbnWPSbtFBClIn14EM8xT0XUVF2HpWo,2928
+nat/plugins/langchain/tools/tavily_internet_search.py,sha256=tXWYPE9nRVWZVy8lZAAFfOwy_zOm4iU5HxAFltNzE70,3010
 nat/plugins/langchain/tools/wikipedia_search.py,sha256=431YwLsjoC_mdvMZ_gY0Q37Uqaue2ASnAHpwr4jWCaU,2197
-nvidia_nat_langchain-1.4.0a20251031.dist-info/licenses/LICENSE-3rd-party.txt,sha256=fOk5jMmCX9YoKWyYzTtfgl-SUy477audFC5hNY4oP7Q,284609
-nvidia_nat_langchain-1.4.0a20251031.dist-info/licenses/LICENSE.md,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
-nvidia_nat_langchain-1.4.0a20251031.dist-info/METADATA,sha256=IMGSc_w655owDdGmTDmqkUsI_uhAVG9YBthllbyh45g,2263
-nvidia_nat_langchain-1.4.0a20251031.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-nvidia_nat_langchain-1.4.0a20251031.dist-info/entry_points.txt,sha256=4deXsMn97I012HhDw0UjoqcZ8eEoZ7BnqaRx5QmzebY,123
-nvidia_nat_langchain-1.4.0a20251031.dist-info/top_level.txt,sha256=8-CJ2cP6-f0ZReXe5Hzqp-5pvzzHz-5Ds5H2bGqh1-U,4
-nvidia_nat_langchain-1.4.0a20251031.dist-info/RECORD,,
+nvidia_nat_langchain-1.4.0a20251220.dist-info/licenses/LICENSE-3rd-party.txt,sha256=fOk5jMmCX9YoKWyYzTtfgl-SUy477audFC5hNY4oP7Q,284609
+nvidia_nat_langchain-1.4.0a20251220.dist-info/licenses/LICENSE.md,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
+nvidia_nat_langchain-1.4.0a20251220.dist-info/METADATA,sha256=q81NJfzfriPKA-AWIoRBDNwthsF3XbS_dv-RlQPJyxQ,2263
+nvidia_nat_langchain-1.4.0a20251220.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+nvidia_nat_langchain-1.4.0a20251220.dist-info/entry_points.txt,sha256=4deXsMn97I012HhDw0UjoqcZ8eEoZ7BnqaRx5QmzebY,123
+nvidia_nat_langchain-1.4.0a20251220.dist-info/top_level.txt,sha256=8-CJ2cP6-f0ZReXe5Hzqp-5pvzzHz-5Ds5H2bGqh1-U,4
+nvidia_nat_langchain-1.4.0a20251220.dist-info/RECORD,,