Keep getting 400 error

[jyotip11]

BadRequestError: Error code: 400 - {'detail': "Invalid value for 'content': expected a string, got null."}
It seems like the tool/sub agent works and then the supervisor is stuck and does not generate anything. How to fix the issue? Anyone facing the same?

[hwchase17]

are you able to share langsmith trace?

[jyotip11]

@tool
def multiply(a: int, b: int) -> int:
"""Multiply two numbers."""
return a*b

from langchain_openai import ChatOpenAI

tools = [multiply]

chat = ChatOpenAI(model="gpt-3.5-turbo-1106", temperature=0)

from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages(
[
(
"system",
"You are a helpful assistant. You may not need to use tools for every query - the user may just want to chat!",
),
("placeholder", "{messages}"),
("placeholder", "{agent_scratchpad}"),
]
)

from langchain.agents import AgentExecutor, create_tool_calling_agent

agent = create_tool_calling_agent(chat, tools, prompt)

agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

agent_executor.invoke({"messages": [HumanMessage(content="multiply 1*2")]})

Trace:

Entering new AgentExecutor chain...

Invoking: multiply with {'a': 2, 'b': 3}

6

BadRequestError Traceback (most recent call last)
Cell In[42], line 1
----> 1 conversational_agent_executor.invoke(
2 {"messages": [HumanMessage("Help me with 2*3")]},
3 {"configurable": {"session_id": "unused"}},
4 )

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:5057, in RunnableBindingBase.invoke(self, input, config, **kwargs)
5051 def invoke(
5052 self,
5053 input: Input,
5054 config: Optional[RunnableConfig] = None,
5055 **kwargs: Optional[Any],
5056 ) -> Output:
-> 5057 return self.bound.invoke(
5058 input,
5059 self._merge_configs(config),
5060 **{**self.kwargs, **kwargs},
5061 )

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:5057, in RunnableBindingBase.invoke(self, input, config, **kwargs)
5051 def invoke(
5052 self,
5053 input: Input,
5054 config: Optional[RunnableConfig] = None,
5055 **kwargs: Optional[Any],
5056 ) -> Output:
-> 5057 return self.bound.invoke(
5058 input,
5059 self._merge_configs(config),
5060 **{**self.kwargs, **kwargs},
5061 )

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:2875, in RunnableSequence.invoke(self, input, config, **kwargs)
2873 input = step.invoke(input, config, **kwargs)
2874 else:
-> 2875 input = step.invoke(input, config)
2876 # finish the root run
2877 except BaseException as e:

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/branch.py:239, in RunnableBranch.invoke(self, input, config, **kwargs)
237 break
238 else:
--> 239 output = self.default.invoke(
240 input,
241 config=patch_config(
242 config, callbacks=run_manager.get_child(tag="branch:default")
243 ),
244 **kwargs,
245 )
246 except BaseException as e:
247 run_manager.on_chain_error(e)

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:5057, in RunnableBindingBase.invoke(self, input, config, **kwargs)
5051 def invoke(
5052 self,
5053 input: Input,
5054 config: Optional[RunnableConfig] = None,
5055 **kwargs: Optional[Any],
5056 ) -> Output:
-> 5057 return self.bound.invoke(
5058 input,
5059 self._merge_configs(config),
5060 **{**self.kwargs, **kwargs},
5061 )

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain/chains/base.py:166, in Chain.invoke(self, input, config, **kwargs)
164 except BaseException as e:
165 run_manager.on_chain_error(e)
--> 166 raise e
167 run_manager.on_chain_end(outputs)
169 if include_run_info:

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain/chains/base.py:156, in Chain.invoke(self, input, config, **kwargs)
153 try:
154 self._validate_inputs(inputs)
155 outputs = (
--> 156 self._call(inputs, run_manager=run_manager)
157 if new_arg_supported
158 else self._call(inputs)
159 )
161 final_outputs: Dict[str, Any] = self.prep_outputs(
162 inputs, outputs, return_only_outputs
163 )
164 except BaseException as e:

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain/agents/agent.py:1612, in AgentExecutor._call(self, inputs, run_manager)
1610 # We now enter the agent loop (until it returns something).
1611 while self._should_continue(iterations, time_elapsed):
-> 1612 next_step_output = self._take_next_step(
1613 name_to_tool_map,
1614 color_mapping,
1615 inputs,
1616 intermediate_steps,
1617 run_manager=run_manager,
1618 )
1619 if isinstance(next_step_output, AgentFinish):
1620 return self._return(
1621 next_step_output, intermediate_steps, run_manager=run_manager
1622 )

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain/agents/agent.py:1318, in AgentExecutor._take_next_step(self, name_to_tool_map, color_mapping, inputs, intermediate_steps, run_manager)
1309 def _take_next_step(
1310 self,
1311 name_to_tool_map: Dict[str, BaseTool],
(...)
1315 run_manager: Optional[CallbackManagerForChainRun] = None,
1316 ) -> Union[AgentFinish, List[Tuple[AgentAction, str]]]:
1317 return self._consume_next_step(
-> 1318 [
1319 a
1320 for a in self._iter_next_step(
1321 name_to_tool_map,
1322 color_mapping,
1323 inputs,
1324 intermediate_steps,
1325 run_manager,
1326 )
1327 ]
1328 )

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain/agents/agent.py:1318, in (.0)
1309 def _take_next_step(
1310 self,
1311 name_to_tool_map: Dict[str, BaseTool],
(...)
1315 run_manager: Optional[CallbackManagerForChainRun] = None,
1316 ) -> Union[AgentFinish, List[Tuple[AgentAction, str]]]:
1317 return self._consume_next_step(
-> 1318 [
1319 a
1320 for a in self._iter_next_step(
1321 name_to_tool_map,
1322 color_mapping,
1323 inputs,
1324 intermediate_steps,
1325 run_manager,
1326 )
1327 ]
1328 )

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain/agents/agent.py:1346, in AgentExecutor._iter_next_step(self, name_to_tool_map, color_mapping, inputs, intermediate_steps, run_manager)
1343 intermediate_steps = self._prepare_intermediate_steps(intermediate_steps)
1345 # Call the LLM to see what to do.
-> 1346 output = self.agent.plan(
1347 intermediate_steps,
1348 callbacks=run_manager.get_child() if run_manager else None,
1349 **inputs,
1350 )
1351 except OutputParserException as e:
1352 if isinstance(self.handle_parsing_errors, bool):

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain/agents/agent.py:580, in RunnableMultiActionAgent.plan(self, intermediate_steps, callbacks, **kwargs)
572 final_output: Any = None
573 if self.stream_runnable:
574 # Use streaming to make sure that the underlying LLM is invoked in a
575 # streaming
(...)
578 # Because the response from the plan is not a generator, we need to
579 # accumulate the output into final output and return that.
--> 580 for chunk in self.runnable.stream(inputs, config={"callbacks": callbacks}):
581 if final_output is None:
582 final_output = chunk

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:3253, in RunnableSequence.stream(self, input, config, **kwargs)
3247 def stream(
3248 self,
3249 input: Input,
3250 config: Optional[RunnableConfig] = None,
3251 **kwargs: Optional[Any],
3252 ) -> Iterator[Output]:
-> 3253 yield from self.transform(iter([input]), config, **kwargs)

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:3240, in RunnableSequence.transform(self, input, config, **kwargs)
3234 def transform(
3235 self,
3236 input: Iterator[Input],
3237 config: Optional[RunnableConfig] = None,
3238 **kwargs: Optional[Any],
3239 ) -> Iterator[Output]:
-> 3240 yield from self._transform_stream_with_config(
3241 input,
3242 self._transform,
3243 patch_config(config, run_name=(config or {}).get("run_name") or self.name),
3244 **kwargs,
3245 )

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:2053, in Runnable._transform_stream_with_config(self, input, transformer, config, run_type, **kwargs)
2051 try:
2052 while True:
-> 2053 chunk: Output = context.run(next, iterator) # type: ignore
2054 yield chunk
2055 if final_output_supported:

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:3202, in RunnableSequence._transform(self, input, run_manager, config, **kwargs)
3199 else:
3200 final_pipeline = step.transform(final_pipeline, config)
-> 3202 for output in final_pipeline:
3203 yield output

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:1271, in Runnable.transform(self, input, config, **kwargs)
1268 final: Input
1269 got_first_val = False
-> 1271 for ichunk in input:
1272 # The default implementation of transform is to buffer input and
1273 # then call stream.
1274 # It'll attempt to gather all input into a single chunk using
1275 # the + operator.
1276 # If the input is not addable, then we'll assume that we can
1277 # only operate on the last chunk,
1278 # and we'll iterate until we get to the last chunk.
1279 if not got_first_val:
1280 final = ichunk

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:5264, in RunnableBindingBase.transform(self, input, config, **kwargs)
5258 def transform(
5259 self,
5260 input: Iterator[Input],
5261 config: Optional[RunnableConfig] = None,
5262 **kwargs: Any,
5263 ) -> Iterator[Output]:
-> 5264 yield from self.bound.transform(
5265 input,
5266 self._merge_configs(config),
5267 **{**self.kwargs, **kwargs},
5268 )

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/runnables/base.py:1289, in Runnable.transform(self, input, config, **kwargs)
1286 final = ichunk
1288 if got_first_val:
-> 1289 yield from self.stream(final, config, **kwargs)

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/language_models/chat_models.py:365, in BaseChatModel.stream(self, input, config, stop, **kwargs)
358 except BaseException as e:
359 run_manager.on_llm_error(
360 e,
361 response=LLMResult(
362 generations=[[generation]] if generation else []
363 ),
364 )
--> 365 raise e
366 else:
367 run_manager.on_llm_end(LLMResult(generations=[[generation]]))

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_core/language_models/chat_models.py:345, in BaseChatModel.stream(self, input, config, stop, **kwargs)
343 generation: Optional[ChatGenerationChunk] = None
344 try:
--> 345 for chunk in self._stream(messages, stop=stop, **kwargs):
346 if chunk.message.id is None:
347 chunk.message.id = f"run-{run_manager.run_id}"

File /opt/zillow/.venv/lib/python3.9/site-packages/langchain_openai/chat_models/base.py:520, in BaseChatOpenAI._stream(self, messages, stop, run_manager, **kwargs)
518 base_generation_info = {"headers": dict(raw_response.headers)}
519 else:
--> 520 response = self.client.create(**payload)
521 base_generation_info = {}
522 with response:

File /opt/zillow/.venv/lib/python3.9/site-packages/openai/_utils/_utils.py:277, in required_args..inner..wrapper(*args, **kwargs)
275 msg = f"Missing required argument: {quote(missing[0])}"
276 raise TypeError(msg)
--> 277 return func(*args, **kwargs)

File /opt/zillow/.venv/lib/python3.9/site-packages/openai/resources/chat/completions.py:646, in Completions.create(self, messages, model, frequency_penalty, function_call, functions, logit_bias, logprobs, max_tokens, n, parallel_tool_calls, presence_penalty, response_format, seed, service_tier, stop, stream, stream_options, temperature, tool_choice, tools, top_logprobs, top_p, user, extra_headers, extra_query, extra_body, timeout)
612 @required_args(["messages", "model"], ["messages", "model", "stream"])
613 def create(
614 self,
(...)
644 timeout: float | httpx.Timeout | None | NotGiven = NOT_GIVEN,
645 ) -> ChatCompletion | Stream[ChatCompletionChunk]:
--> 646 return self._post(
647 "/chat/completions",
648 body=maybe_transform(
649 {
650 "messages": messages,
651 "model": model,
652 "frequency_penalty": frequency_penalty,
653 "function_call": function_call,
654 "functions": functions,
655 "logit_bias": logit_bias,
656 "logprobs": logprobs,
657 "max_tokens": max_tokens,
658 "n": n,
659 "parallel_tool_calls": parallel_tool_calls,
660 "presence_penalty": presence_penalty,
661 "response_format": response_format,
662 "seed": seed,
663 "service_tier": service_tier,
664 "stop": stop,
665 "stream": stream,
666 "stream_options": stream_options,
667 "temperature": temperature,
668 "tool_choice": tool_choice,
669 "tools": tools,
670 "top_logprobs": top_logprobs,
671 "top_p": top_p,
672 "user": user,
673 },
674 completion_create_params.CompletionCreateParams,
675 ),
676 options=make_request_options(
677 extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
678 ),
679 cast_to=ChatCompletion,
680 stream=stream or False,
681 stream_cls=Stream[ChatCompletionChunk],
682 )

File /opt/zillow/.venv/lib/python3.9/site-packages/openai/_base_client.py:1266, in SyncAPIClient.post(self, path, cast_to, body, options, files, stream, stream_cls)
1252 def post(
1253 self,
1254 path: str,
(...)
1261 stream_cls: type[_StreamT] | None = None,
1262 ) -> ResponseT | _StreamT:
1263 opts = FinalRequestOptions.construct(
1264 method="post", url=path, json_data=body, files=to_httpx_files(files), **options
1265 )
-> 1266 return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))

File /opt/zillow/.venv/lib/python3.9/site-packages/openai/_base_client.py:942, in SyncAPIClient.request(self, cast_to, options, remaining_retries, stream, stream_cls)
933 def request(
934 self,
935 cast_to: Type[ResponseT],
(...)
940 stream_cls: type[_StreamT] | None = None,
941 ) -> ResponseT | _StreamT:
--> 942 return self._request(
943 cast_to=cast_to,
944 options=options,
945 stream=stream,
946 stream_cls=stream_cls,
947 remaining_retries=remaining_retries,
948 )

File /opt/zillow/.venv/lib/python3.9/site-packages/openai/_base_client.py:1046, in SyncAPIClient._request(self, cast_to, options, remaining_retries, stream, stream_cls)
1043 err.response.read()
1045 log.debug("Re-raising status error")
-> 1046 raise self._make_status_error_from_response(err.response) from None
1048 return self._process_response(
1049 cast_to=cast_to,
1050 options=options,
(...)
1053 stream_cls=stream_cls,
1054 )

BadRequestError: Error code: 400 - {'detail': "Invalid value for 'content': expected a string, got null."}

Seems like tool gets called but it just stops at that