TavilySearchResults breaks langgraph

[woraphonedge]

from langchain_community.tools.tavily_search import TavilySearchResults

tavily_search = TavilySearchResults(max_results=2)

tools = [
    get_weather,
    other_tool,
    tavily_search
]

from langgraph.prebuilt import ToolNode
tool_node = ToolNode(tools=tools)
graph_builder.add_node("tools", tool_node)

...
graph_builder.add_conditional_edges(
    "chatbot",
    tools_condition,
    {"tools": "tools", "__end__": "__end__"},
)
...

for chunk in graph.stream(
    {"messages": [("human", "what is node in langgraph")]}, stream_mode="values"
):
    chunk["messages"][-1].pretty_print()

If graph agent decides to use tavily_search, it can perform search, then throw error
BadRequestError: Error code: 400 - {'error': {'message': "Missing required parameter: 'messages[2].content[0].type'.", 'type': 'invalid_request_error', 'param': 'messages[2].content[0].type', 'code': 'missing_required_parameter'}}

Other tools work fine though.

[hinthornw]

What LLM are you using?

[woraphonedge]

gpt-4o-mini and anthropic. Same result.

Thing is, if I replace tools_condition with

class BasicToolNode:
    """A node that runs the tools requested in the last AIMessage."""

    def __init__(self, tools: list) -> None:
        self.tools_by_name = {tool.name: tool for tool in tools}

    def __call__(self, inputs: dict):
        if messages := inputs.get("messages", []):
            message = messages[-1]
        else:
            raise ValueError("No message found in input")
        outputs = []
        for tool_call in message.tool_calls:
            tool_result = self.tools_by_name[tool_call["name"]].invoke(
                tool_call["args"]
            )
            outputs.append(
                ToolMessage(
                    content=json.dumps(tool_result),
                    name=tool_call["name"],
                    tool_call_id=tool_call["id"],
                )
            )
        return {"messages": outputs}

It will work ok.

[hwchase17]

I believe you just need to upgrade to version 0.1.12 and this should be fixed - can you try this and confirm? if yes, we should yank this version (which had a bug in it)

[woraphonedge]

Yes, upgrading to 0.1.14 does fix it!

Thank you very much.