feat: AmazonBedrockChatGenerator - migrate Anthropic chat models to use messaging API

[vblagoje]

Fixes deepset-ai/haystack#7324

• Supports new Anthropic messaging API
• Supports old Claude 2.1 chat models and new Claude 3 chat models
• Add support for system messages

[vblagoje]

@anakin87 not a priority yet, let's sync internally first

[anakin87]

In general, I feel that our Bedrock Generators can be simplified somewhat...
It is not always easy to understand what is going on.

About this PR, I have an idea:
can we start adding integration tests, as we do for other integrations?
That way, we can be sure that our integration is stable and working...

[vblagoje]

@anakin87 in 8683af0 I've added model parameter tracking. I thought it lead to a better UX. If user uses unknown parameter rather than getting exception from AWS she'll get a warning.

Instead of:

>>> client.run(messages, generation_kwargs={"max_tokens": 512, "semi": ["Sprache"]})
Traceback (most recent call last):
  File "/Users/vblagoje/workspace/haystack-core-integrations/integrations/amazon_bedrock/src/haystack_integrations/components/generators/amazon_bedrock/chat/chat_generator.py", line 161, in invoke
    response = self.client.invoke_model_with_response_stream(
  File "/Users/vblagoje/miniconda3/envs/haystack/lib/python3.10/site-packages/botocore/client.py", line 553, in _api_call
    return self._make_api_call(operation_name, kwargs)
  File "/Users/vblagoje/miniconda3/envs/haystack/lib/python3.10/site-packages/botocore/client.py", line 1009, in _make_api_call
    raise error_class(parsed_response, operation_name)
botocore.errorfactory.ValidationException: An error occurred (ValidationException) when calling the InvokeModelWithResponseStream operation: Malformed input request: #: subject must not be valid against schema {"required":["messages"]}#: extraneous key [semi] is not permitted, please reformat your input and try again.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/vblagoje/workspace/haystack-core-integrations/integrations/amazon_bedrock/src/haystack_integrations/components/generators/amazon_bedrock/chat/chat_generator.py", line 190, in run
    return {"replies": self.invoke(messages=messages, **(generation_kwargs or {}))}
  File "/Users/vblagoje/workspace/haystack-core-integrations/integrations/amazon_bedrock/src/haystack_integrations/components/generators/amazon_bedrock/chat/chat_generator.py", line 176, in invoke
    raise AmazonBedrockInferenceError(msg) from exception
haystack_integrations.common.amazon_bedrock.errors.AmazonBedrockInferenceError: Could not inference Amazon Bedrock model anthropic.claude-3-sonnet-20240229-v1:0 due: An error occurred (ValidationException) when calling the InvokeModelWithResponseStream operation: Malformed input request: #: subject must not be valid against schema {"required":["messages"]}#: extraneous key [semi] is not permitted, please reformat your input and try again.

user will see:

>>> client.run(messages, generation_kwargs={"max_tokens": 512, "semi": ["Sprache"]})
Parameter 'semi' is not allowed and will be ignored.
Natürliche Sprachverarbeitung (NLP) ist ein Bereich der künstlichen Intelligenz, der sich mit der Entwicklung von Computersystemen befasst, die menschliche Sprache in all ihren Formen verstehen und verarbeiten können. Das Ziel ist es, Computern beizubringen, natürliche Sprache zu analysieren, zu interpretieren und zu generieren, so dass eine effektive Kommunikation zwischen Menschen und Maschinen möglich wird.

This comes with cost of tracking these parameters. LMK if you think it is worth it. We can rollback the last commit if you disagree.

[anakin87]

In general, I would like to simplify this integration a bit in the future.
I still see some legacy of the Prompt Node 😃

Anyway, your PR looks good and I appreciated the extra effort of adding integration tests.
I think they will help us in the future...