llm_function.py

from PIL import Image
import os
import anthropic
import base64
from dotenv import load_dotenv
import cv2
from mimetypes import MimeTypes 

load_dotenv() 
client = anthropic.Anthropic(api_key= os.getenv('ANTHROPIC_API_KEY'))


prompt = """Given 2 construction blueprints your task is to analyze carefully both blueprints and point out difference for following categories in markdown format -
1. Strcutural grid.
2. Layout Areas - rooms , balcony , porch , staircase , elevator etc.
3. Interior changes or optimization.
Summarize all the difference in paragraph concisely in Markdown format .
"""


def encode_image(image_path):
    with open(image_path, "rb") as image_file:
        return base64.b64encode(image_file.read()).decode("utf-8")
    # return base64.b64encode(image_path.getvalue()).decode("utf-8")


def chat_claude(prompt , image1 , image2 ) :
    # print("image 1"  , image1)
    image1_data = encode_image(image1) 
    # print("image 1 data" , image1_data)
    image2_data = encode_image(image2) 
    mime = MimeTypes() 
    image1_media_type =mime.guess_type(image1)[0]
    image2_media_type =mime.guess_type(image2)[0]
    
     

    message = client.messages.create(
        model="claude-3-opus-20240229",
        max_tokens = 4096,
        messages=[
            {
                "role": "user",
                "content": [
                    {
                        "type": "text",
                        "text": "Image 1:"
                    },
                    {
                        "type": "image",
                        "source": {
                            "type": "base64",
                            # "media_type": "image/jpeg",
                            "media_type": image1_media_type,

                            "data": image1_data,
                        },
                    },
                    {
                        "type": "text",
                        "text": "Image 2:"
                    },
                    {
                        "type": "image",
                        "source": {
                            "type": "base64",
                            # "media_type": "image/jpeg",
                            "media_type": image2_media_type,
                            "data": image2_data,
                        },
                    },
                    {
                        "type": "text",
                        "text": f"{prompt}"
                    }
                ],
            }
        ],
    )
    return message.content[0].text