langgraph/tutorials/reflexion/reflexion/

[giscus[bot]]

langgraph/tutorials/reflexion/reflexion/

Build language agents as graphs

https://langchain-ai.github.io/langgraph/tutorials/reflexion/reflexion/?h=tool+node

[wodecki]

Last call

events = graph.stream(
    [HumanMessage(content="How should we handle the climate crisis?")],
    stream_mode="values",
)
for i, step in enumerate(events):
    print(f"Step {i}")
    step[-1].pretty_print()

generates:

InvalidUpdateError: Expected dict, got [HumanMessage(content='How should we handle the climate crisis?')]

[isahers1]

Thank you for bringing this up, this PR should fix it, lmk if not: #1620

[wodecki]

For me the problem persists. In my opinion, the problem lies in the validator:
class ResponderWithRetries:
def init(self, runnable, validator):
self.runnable = runnable
self.validator = validator

def respond(self, state: list):
    response = []
    for attempt in range(3):
        response = self.runnable.invoke(
            {"messages": state}, {"tags": [f"attempt:{attempt}"]} #<<< that may generate the problem
        )

[nkimoto]

In the following class, the respond method takes the state variable as a list type, but it appears to be treated as a dictionary with expressions like state["messages"]. Could you please check this?

class ResponderWithRetries:
    def __init__(self, runnable, validator):
        self.runnable = runnable
        self.validator = validator

    def respond(self, state: list):
        response = []
        for attempt in range(3):
            response = self.runnable.invoke(
                {"messages": state["messages"]}, {"tags": [f"attempt:{attempt}"]}
            )
            try:
                self.validator.invoke(response)
                return {"messages": response}
            except ValidationError as e:
                state = state + [
                    response,
                    ToolMessage(
                        content=f"{repr(e)}\n\nPay close attention to the function schema.\n\n"
                        + self.validator.schema_json()
                        + " Respond by fixing all validation errors.",
                        tool_call_id=response.tool_calls[0]["id"],
                    ),
                ]
        return {"messages": response}

[haradatm]

I think it works with the following modification. Could you please check this?

• def respond(self, state: list):'' -> def respond(self, state: dict):''
• state = state + ['' -> state["messages"] = state["messages"] + [''

[vbarda]

thanks for spotting -- updated!

[haradatm]

@vbarda
Please fix this one too.

• state = state + ['' -> state["messages"] = state["messages"] + [''

[jerry0li]

Seems the respond function is wrong as the operator + between dict and list.

class ResponderWithRetries:
    def __init__(self, runnable, validator):
        self.runnable = runnable
        self.validator = validator

    def respond(self, state: dict):
        response = []
        for attempt in range(3):
            response = self.runnable.invoke({"messages": state["messages"]}, {"tags": [f"attempt:{attempt}"]})
            try:
                self.validator.invoke(response)
                return {"messages": response}
            except ValidationError as e:
                state = state + [
                    response,
                    ToolMessage(
                        content=f"{repr(e)}\n\nPay close attention to the function schema.\n\n"
                        + self.validator.schema_json()
                        + " Respond by fixing all validation errors.",
                        tool_call_id=response.tool_calls[0]["id"],
                    ),
                ]
        return {"messages": response}

• error msgs:

state = state + [
                    response,
                    ToolMessage(
                        content=f"{repr(e)}\n\nPay close attention to the function schema.\n\n"
                        + self.validator.schema_json()
                        + " Respond by fixing all validation errors.",
                        tool_call_id=response.tool_calls[0]["id"],
                    ),
                ]

TypeError: unsupported operand type(s) for +: 'dict' and 'list'

[haradatm]

See: #1287 (reply in thread)

[chrisyuuuuu]

is error? state = state -> state["messages"] = state["messages"]

[haradatm]

I think this is because the argument `state' of respond() is of type dict. You can check if you dare to rise ValidationError.

• Incorrect
  state = state + [
• Correct
  state["messages"] = state["messages"] + [