fix(langchain_v1): remove non llm controllable params from tool message on invocation failure

[eyurtsev]

The LLM shouldn't be seeing parameters it cannot control in the ToolMessage error it gets when it invokes a tool with incorrect args.

This fixes the behavior within langchain to address immediate issue.

We may want to change the behavior in langchain_core as well to prevent validation of injected arguments. But this would be done in a separate change

[sydney-runkle]

• Changing how we inject args (where that happens) seems good ✅
• It seems like the right way to do this is actually to just skip validation for injected args (so that issues would never show up in validation errors)

One concern that I have at the moment is, what if only state for example fails validation? We don't want to be hitting the error branch in that case at all... what's the lift like to change this in core instead?

[eyurtsev]

| It seems like the right way to do this is actually to just skip validation for injected args (so that issues would never show up in validation errors)

Yeah -- this is the work in langchain_core. Let me know what you discover. Splitting into 2 separate PRs