try.py

import boto3
import json
import re
import ldclient
from ldclient.config import Config
import os


sdk_key = os.getenv("LAUNCHDARKLY_SDK_KEY")

feature_flag_key = "lengthy-vs-brief"

bedrock_runtime = boto3.client('bedrock-runtime', region_name='us-west-2')


def getResult(prompt):
    kwargs = {
        "modelId": "anthropic.claude-3-haiku-20240307-v1:0",
        "contentType": "application/json",
        "accept": "application/json",
        "body": json.dumps({
            "anthropic_version": "bedrock-2023-05-31",
            "max_tokens": 1000,
            "messages": [
                {
                    "role": "user",
                    "content": [

                        {
                            "type": "text",
                            "text": prompt
                        }
                    ]
                }
            ]
        })
    }

    response = bedrock_runtime.invoke_model(**kwargs)

    body = json.loads(response['body'].read())

    edit_body = body['content'][0]['text']
    matches = re.findall(r'\*\*(.*?)\*\*', edit_body)
    new_match = ""
    for match in matches:
        new_match = new_match + ", " + match

    new_prompt = "Give me the github repo link to all those packages: " + new_match

    new_kwargs = {
        "modelId": "anthropic.claude-3-haiku-20240307-v1:0",
        "contentType": "application/json",
        "accept": "application/json",
        "body": json.dumps({
            "anthropic_version": "bedrock-2023-05-31",
            "max_tokens": 1000,
            "messages": [
                {
                    "role": "user",
                    "content": [

                        {
                            "type": "text",
                            "text": new_prompt
                        }
                    ]
                }
            ]
        })
    }

    new_response = bedrock_runtime.invoke_model(**new_kwargs)

    new_body = json.loads(new_response['body'].read())

    new_edit_body = new_body['content'][0]['text']

    return new_edit_body


print(getResult("I need an NLP model that detect user sentiments"))