Fix: max_tokens typo in Mistral Chat

[vish9812]

ALLOWED_PARAMS has max_tokens but generation_kwargs was using max_gen_len.

Closes Issue : https://github.com/deepset-ai/haystack-core-integrations/issues/741

[CLAassistant]

All committers have signed the CLA.

[masci]

note: AWS-based tests are expected to fail when running from forks (to be fixed)

[julian-risch]

Hello Vishal @vish9812 and thank you for opening this pull request. Let me try it out locally once and then I will take care of merging this and releasing a new version of the amazon-bedrock-haystack PyPI package.

[vish9812]

@julian-risch I only tested the max_tokens and temperature, and was able to run inference successfully. But you're right ALLOWED_PARAMS don't match the suggested params in AWS.

{
    "prompt": string,
    "max_tokens" : int,
    "stop" : [string],    
    "temperature": float,
    "top_p": float,
    "top_k": int
}

Also, the docs can be updated to mention that Mistral model is also supported.