fix(react_agent): raise on truncated LLM output (finish_reason='length')

packages/nvidia_nat_langchain/src/nat/plugins/langchain/agent/react_agent/agent.py

@@ -38,6 +38,7 @@
 from pydantic import BaseModel
 from pydantic import Field
 
+from nat.llm.utils.constants import LLMFinishReason
 from nat.plugins.langchain.agent.base import AGENT_CALL_LOG_MESSAGE
 from nat.plugins.langchain.agent.base import AGENT_LOG_PREFIX
 from nat.plugins.langchain.agent.base import INPUT_SCHEMA_MESSAGE
@@ -309,6 +310,10 @@ async def agent_node(self, state: ReActGraphState, config: RunnableConfig | None
                         logger.debug("%s The agent's scratchpad (with tool result) was:\n%s",
                                      AGENT_LOG_PREFIX,
                                      agent_scratchpad)
+                # A truncated reply would be parsed as a complete answer or tool input (#2518)
+                if getattr(output_message, "response_metadata", {}).get("finish_reason") == LLMFinishReason.LENGTH:
+                    raise RuntimeError("LLM output truncated (finish_reason='length'). "
+                                       f"Truncated output: {str(output_message.content)[:500]!r}")
                 try:
                     # check if the agent has the final answer yet
                     logger.debug("%s Successfully obtained agent response. Parsing agent's response", AGENT_LOG_PREFIX)

packages/nvidia_nat_langchain/tests/agent/test_react.py

@@ -2442,3 +2442,31 @@ async def on_llm_end(self, *args: typing.Any, **kwargs: typing.Any) -> None:
                                                                          config={"callbacks": [CountLLMEnds()]})
     assert CountLLMEnds.calls == 2
     assert response.usage.total_tokens == 300
+
+
+async def test_agent_node_raises_on_truncated_output(mock_react_agent_no_raise):
+    from unittest.mock import AsyncMock
+    from unittest.mock import patch
+
+    truncated = AIMessage(content="Thought: I know this.\nFinal Answer: LangSmith is a platform for",
+                          response_metadata={"finish_reason": "length"})
+    with patch.object(mock_react_agent_no_raise, '_stream_llm', new_callable=AsyncMock) as mock_stream_llm:
+        mock_stream_llm.return_value = truncated
+        with pytest.raises(RuntimeError, match="finish_reason='length'"):
+            await mock_react_agent_no_raise.agent_node(ReActGraphState(messages=[HumanMessage(content="hi")]))
+
+
+async def test_agent_node_raises_on_truncated_output_in_agentic_cycle(mock_react_agent_no_raise):
+    from unittest.mock import AsyncMock
+    from unittest.mock import patch
+
+    truncated = AIMessage(content="Thought: search more.\nAction: Tool A\nAction Input: {\"query\": \"Lang",
+                          response_metadata={"finish_reason": "length"})
+    state = ReActGraphState(messages=[HumanMessage(content="hi")],
+                            agent_scratchpad=[AgentAction(tool="Tool A", tool_input="hello", log="test")],
+                            tool_responses=[ToolMessage(name="Tool A", tool_call_id="Tool A", content="result")])
+    with patch.object(mock_react_agent_no_raise, '_stream_llm', new_callable=AsyncMock) as mock_stream_llm:
+        mock_stream_llm.return_value = truncated
+        with pytest.raises(RuntimeError, match="finish_reason='length'"):
+            await mock_react_agent_no_raise.agent_node(state)
+        assert "agent_scratchpad" in mock_stream_llm.call_args.args[1]