nvidia-nat-langchain 1.4.0a20251015__py3-none-any.whl → 1.4.0a20251022__py3-none-any.whl

nat/plugins/langchain/llm.py

@@ -12,13 +12,16 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+# pylint: disable=unused-argument
 
+import logging
 from collections.abc import Sequence
 from typing import TypeVar
 
 from nat.builder.builder import Builder
 from nat.builder.framework_enum import LLMFrameworkEnum
 from nat.cli.register_workflow import register_llm_client
+from nat.data_models.llm import APITypeEnum
 from nat.data_models.llm import LLMBaseConfig
 from nat.data_models.retry_mixin import RetryMixin
 from nat.data_models.thinking_mixin import ThinkingMixin
@@ -31,8 +34,11 @@ from nat.llm.utils.thinking import BaseThinkingInjector
 from nat.llm.utils.thinking import FunctionArgumentWrapper
 from nat.llm.utils.thinking import patch_with_thinking
 from nat.utils.exception_handlers.automatic_retries import patch_with_retry
+from nat.utils.responses_api import validate_no_responses_api
 from nat.utils.type_utils import override
 
+logger = logging.getLogger(__name__)
+
 ModelType = TypeVar("ModelType")
 
 
@@ -65,20 +71,22 @@ def _patch_llm_based_on_config(client: ModelType, llm_config: LLMBaseConfig) ->
             Raises:
                 ValueError: If the messages are not a valid type for LanguageModelInput.
             """
-            system_message = SystemMessage(content=self.system_prompt)
-            if isinstance(messages, BaseMessage):
-                new_messages = [system_message, messages]
-                return FunctionArgumentWrapper(new_messages, *args, **kwargs)
-            elif isinstance(messages, PromptValue):
-                new_messages = [system_message, *messages.to_messages()]
-                return FunctionArgumentWrapper(new_messages, *args, **kwargs)
+            if isinstance(messages, PromptValue):
+                messages = messages.to_messages()
             elif isinstance(messages, str):
-                new_messages = [system_message, HumanMessage(content=messages)]
-                return FunctionArgumentWrapper(new_messages, *args, **kwargs)
-            elif isinstance(messages, Sequence):
-                if all(isinstance(m, BaseMessage) for m in messages):
-                    new_messages = [system_message, *list(messages)]
-                    return FunctionArgumentWrapper(new_messages, *args, **kwargs)
+                messages = [HumanMessage(content=messages)]
+
+            if isinstance(messages, Sequence) and all(isinstance(m, BaseMessage) for m in messages):
+                for i, message in enumerate(messages):
+                    if isinstance(message, SystemMessage):
+                        if self.system_prompt not in str(message.content):
+                            messages = list(messages)
+                            messages[i] = SystemMessage(content=f"{message.content}\n{self.system_prompt}")
+                        break
+                else:
+                    messages = list(messages)
+                    messages.insert(0, SystemMessage(content=self.system_prompt))
+                return FunctionArgumentWrapper(messages, *args, **kwargs)
             raise ValueError(f"Unsupported message type: {type(messages)}")
 
     if isinstance(llm_config, RetryMixin):
@@ -108,8 +116,10 @@ async def aws_bedrock_langchain(llm_config: AWSBedrockModelConfig, _builder: Bui
 
     from langchain_aws import ChatBedrockConverse
 
+    validate_no_responses_api(llm_config, LLMFrameworkEnum.LANGCHAIN)
+
     client = ChatBedrockConverse(**llm_config.model_dump(
-        exclude={"type", "context_size", "thinking"},
+        exclude={"type", "context_size", "thinking", "api_type"},
         by_alias=True,
         exclude_none=True,
     ))
@@ -122,7 +132,10 @@ async def azure_openai_langchain(llm_config: AzureOpenAIModelConfig, _builder: B
 
     from langchain_openai import AzureChatOpenAI
 
-    client = AzureChatOpenAI(**llm_config.model_dump(exclude={"type", "thinking"}, by_alias=True, exclude_none=True))
+    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))
 
     yield _patch_llm_based_on_config(client, llm_config)
 
@@ -132,9 +145,13 @@ async def nim_langchain(llm_config: NIMModelConfig, _builder: Builder):
 
     from langchain_nvidia_ai_endpoints import ChatNVIDIA
 
+    validate_no_responses_api(llm_config, LLMFrameworkEnum.LANGCHAIN)
+
     # prefer max_completion_tokens over max_tokens
     client = ChatNVIDIA(
-        **llm_config.model_dump(exclude={"type", "max_tokens", "thinking"}, by_alias=True, exclude_none=True),
+        **llm_config.model_dump(exclude={"type", "max_tokens", "thinking", "api_type"},
+                                by_alias=True,
+                                exclude_none=True),
         max_completion_tokens=llm_config.max_tokens,
     )
 
@@ -146,13 +163,23 @@ async def openai_langchain(llm_config: OpenAIModelConfig, _builder: Builder):
 
     from langchain_openai import ChatOpenAI
 
-    # If stream_usage is specified, it will override the default value of True.
-    client = ChatOpenAI(stream_usage=True,
-                        **llm_config.model_dump(
-                            exclude={"type", "thinking"},
-                            by_alias=True,
-                            exclude_none=True,
-                        ))
+    if llm_config.api_type == APITypeEnum.RESPONSES:
+        client = ChatOpenAI(stream_usage=True,
+                            use_responses_api=True,
+                            use_previous_response_id=True,
+                            **llm_config.model_dump(
+                                exclude={"type", "thinking", "api_type"},
+                                by_alias=True,
+                                exclude_none=True,
+                            ))
+    else:
+        # If stream_usage is specified, it will override the default value of True.
+        client = ChatOpenAI(stream_usage=True,
+                            **llm_config.model_dump(
+                                exclude={"type", "thinking", "api_type"},
+                                by_alias=True,
+                                exclude_none=True,
+                            ))
 
     yield _patch_llm_based_on_config(client, llm_config)
 
@@ -162,6 +189,9 @@ async def litellm_langchain(llm_config: LiteLlmModelConfig, _builder: Builder):
 
     from langchain_litellm import ChatLiteLLM
 
-    client = ChatLiteLLM(**llm_config.model_dump(exclude={"type", "thinking"}, by_alias=True, exclude_none=True))
+    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))
 
     yield _patch_llm_based_on_config(client, llm_config)

{nvidia_nat_langchain-1.4.0a20251015.dist-info → nvidia_nat_langchain-1.4.0a20251022.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nvidia-nat-langchain
-Version: 1.4.0a20251015
+Version: 1.4.0a20251022
 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==v1.4.0a20251015
+Requires-Dist: nvidia-nat==v1.4.0a20251022
 Requires-Dist: langchain-aws~=0.2.31
 Requires-Dist: langchain-core~=0.3.75
 Requires-Dist: langchain-litellm~=0.2.3

{nvidia_nat_langchain-1.4.0a20251015.dist-info → nvidia_nat_langchain-1.4.0a20251022.dist-info}/RECORD

@@ -1,7 +1,7 @@
 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=vlR_4bMLmmpXTv4kp3xFOB1eFeJ-wX1_DqnsQpPunAo,7110
+nat/plugins/langchain/llm.py,sha256=6e-Ug6iahZdOJRLCJmVmQTgSYxu0xmslyANaUQG_qPY,8213
 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
@@ -10,10 +10,10 @@ nat/plugins/langchain/tools/code_generation_tool.py,sha256=f5pna0WMOx3QOS4WnaMFK
 nat/plugins/langchain/tools/register.py,sha256=uemxqLxcNk1bGX4crV52oMphLTZWonStzkXwTZeG2Rw,889
 nat/plugins/langchain/tools/tavily_internet_search.py,sha256=UFMP1xh_kC3fydMQBeV-oDZ-M7jnLcs5OkMSzgm7mng,2653
 nat/plugins/langchain/tools/wikipedia_search.py,sha256=431YwLsjoC_mdvMZ_gY0Q37Uqaue2ASnAHpwr4jWCaU,2197
-nvidia_nat_langchain-1.4.0a20251015.dist-info/licenses/LICENSE-3rd-party.txt,sha256=fOk5jMmCX9YoKWyYzTtfgl-SUy477audFC5hNY4oP7Q,284609
-nvidia_nat_langchain-1.4.0a20251015.dist-info/licenses/LICENSE.md,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
-nvidia_nat_langchain-1.4.0a20251015.dist-info/METADATA,sha256=_AMrQwRxIW-FKwgiA2O7pzJGct-thF7gHXE8Q6iyMVA,2230
-nvidia_nat_langchain-1.4.0a20251015.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-nvidia_nat_langchain-1.4.0a20251015.dist-info/entry_points.txt,sha256=4deXsMn97I012HhDw0UjoqcZ8eEoZ7BnqaRx5QmzebY,123
-nvidia_nat_langchain-1.4.0a20251015.dist-info/top_level.txt,sha256=8-CJ2cP6-f0ZReXe5Hzqp-5pvzzHz-5Ds5H2bGqh1-U,4
-nvidia_nat_langchain-1.4.0a20251015.dist-info/RECORD,,
+nvidia_nat_langchain-1.4.0a20251022.dist-info/licenses/LICENSE-3rd-party.txt,sha256=fOk5jMmCX9YoKWyYzTtfgl-SUy477audFC5hNY4oP7Q,284609
+nvidia_nat_langchain-1.4.0a20251022.dist-info/licenses/LICENSE.md,sha256=QwcOLU5TJoTeUhuIXzhdCEEDDvorGiC6-3YTOl4TecE,11356
+nvidia_nat_langchain-1.4.0a20251022.dist-info/METADATA,sha256=yHhMK5r1Ps37cn4w_3jR9sdY6W5Coz-n1TKTL3E5CqM,2230
+nvidia_nat_langchain-1.4.0a20251022.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+nvidia_nat_langchain-1.4.0a20251022.dist-info/entry_points.txt,sha256=4deXsMn97I012HhDw0UjoqcZ8eEoZ7BnqaRx5QmzebY,123
+nvidia_nat_langchain-1.4.0a20251022.dist-info/top_level.txt,sha256=8-CJ2cP6-f0ZReXe5Hzqp-5pvzzHz-5Ds5H2bGqh1-U,4
+nvidia_nat_langchain-1.4.0a20251022.dist-info/RECORD,,