fix: created custom error when env var for providers does not exist or unknown provider, moderator is attempted to be loaded

[lifeizhou-ap]

Why

Currently it throws a runtime error and print the stack trace when

• when the environment variable is not set for LLM apis,
• when unknown provider or moderator is trying to be loaded

This make it hard for users to understand what is going.

What

• Raise a MissingProviderEnvVariableError with attributes of provider, env_variable
• Raise a LoadExchangeAttributeError for unknown provider or moderator

With the above custom erros we can construct user friendly messages in goose

How will it be used
The errors will be caught in goose. goose PR

Note
This PR is for early feedback, and it won't be merged into this repo as @baxen is going to move this repo to goose. I will move the PR change to goose accordingly afterwards

[codefromthecrypt]

looks like progress

[codefromthecrypt]

maybe rename instructions to instructions_url?

Also, not sure style, but if ok with multi-line error we could \n this and then make the asserts multi-line

Suggested change

	self.message += f". {instructions}"
	self.message += f". See {instructions} for instructions."

[lifeizhou-ap]

Thanks @codefromthecrypt for your feedback!

Yes, we can rename to instructions_url so that we will have consistent message pattern. On the other hand, I am thinking whether we shall have some flexibility for the instructions. For example, people can put instructions without url. It is hard to know the future :).

I'll change as you suggested, in case we want to have more flexibility, it is fine because the function signature won't change anyway.

Also I've added the error message for unknown configuration values and put the screenshot in block/goose#103. I would like to get your feedback too! Thank you!

[lifeizhou-ap]

moved the change to block/goose#103