docs/troubleshooting/errors/INVALID_TOOL_RESULTS/

[giscus[bot]]

docs/troubleshooting/errors/INVALID_TOOL_RESULTS/

You are passing too many, too few, or mismatched ToolMessages to a model.

https://python.langchain.com/docs/troubleshooting/errors/INVALID_TOOL_RESULTS/

[ejubber20]

Is it not possible for a tool to call an LLM then?

For example, imagine I have Node A, which calls Tool A. Tool A has an LLM call with structured output, and is supposed to return the results back to Node A. It seems like the relationship between ToolCall/ToolMessage would make this impossible, or is there a way to do it?