Support for Bedrock, and Anthropic

Also, code to normalize response to OpenAI format.

Author: rohitprasad15

aisuitealt/providers/anthropic_provider.py

@@ -0,0 +1,39 @@
+import anthropic
+from provider import Provider
+
+class AnthropicProvider(Provider):
+    def __init__(self, **config):
+        """
+        Initialize the Anthropic provider with the given configuration.
+        Pass the entire configuration dictionary to the Anthropic client constructor.
+        """
+
+        self.client = anthropic.Anthropic(**config)
+
+    def chat_completions_create(self, model, messages, **kwargs):
+        # Check if the fist message is a system message
+        if messages[0]["role"] == "system":
+            system_message = messages[0]["content"]
+            messages = messages[1:]
+        else:
+            system_message = None
+
+        return self.normalize_response(self.client.messages.create(
+            model=model,
+            system=system_message,
+            messages=messages,
+            **kwargs
+        ))
+
+    def normalize_response(self, response):
+        """ Normalize the response from the Anthropic API to match OpenAI's response format. """
+        return {
+            "choices": [
+                {
+                    "message": {
+                        "role": response.get("role", "assistant"),
+                        "content": response.get("content", ""),
+                    }
+                }
+            ]
+        }

aisuitealt/providers/aws_bedrock_provider.py

@@ -1,4 +1,4 @@
-from anthropic import AnthropicBedrock
+import boto3
 from provider import Provider, LLMError
 
 class AWSBedrockProvider(Provider):
@@ -10,13 +10,51 @@ def __init__(self, **config):
         # Anthropic Bedrock client will use the default AWS credential providers, such as
         # using ~/.aws/credentials or the "AWS_SECRET_ACCESS_KEY" and "AWS_ACCESS_KEY_ID" environment variables.
         # Any overrides from the user is passed to the constructor.
-        self.client = AnthropicBedrock(**config)
+        self.client = boto3.client("bedrock-runtime", **config)
+        # Maintain a list of Inference Parameters which Bedrock supports.
+        # https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_InferenceConfiguration.html
+        self.inference_parameters = ['maxTokens', 'temperature', 'topP', 'stopSequences']
+
+    def normalize_response(self, response):
+        """Normalize the response from the Bedrock API to match OpenAI's response format."""
+        return {
+            "choices": [
+                {
+                    "message": {
+                        "content": response["output"]["message"]["content"] if response["output"].get("message") else "",
+                        "role": "assistant"
+                    },
+                }
+            ]
+        }
 
     def chat_completions_create(self, model, messages, **kwargs):
         # Any exception raised by Anthropic will be returned to the caller.
         # Maybe we should catch them and raise a custom LLMError.
-        return self.client.messages.create(
+        system_message = None
+        if messages[0]["role"] == "system":
+            system_message = [messages[0]["content"]]
+            messages = messages[1:]
+
+        # Maintain a list of Inference Parameters which Bedrock supports.
+        # These fields need to be passed using inferenceConfig.
+        # Rest all other fields are passed as additionalModelRequestFields.
+        inference_config = {}
+        additional_model_request_fields = {}
+
+        # Iterate over the kwargs and separate the inference parameters and additional model request fields.
+        for key, value in kwargs.items():
+            if key in self.inference_parameters:
+                inference_config[key] = value
+            else:
+                additional_model_request_fields[key] = value
+
+        # Call the Bedrock Converse API.
+        response = self.client.converse(
             model=model,
             messages=messages,
-            **kwargs # Pass any additional arguments to the Anthropic API. Eg: max_tokens.
+            system=system_message,
+            inferenceConfig=inference_config,
+            additionalModelRequestFields=additional_model_request_fields
         )
+        return self.normalize_response(response)