app.py

from langchain.agents import load_tools
from langchain.agents import initialize_agent
from langchain.agents import AgentType
from langchain.chat_models import AzureChatOpenAI
from langchain.llms import OpenAI
from langchain.chains.conversation.memory import ConversationBufferWindowMemory

import os


OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
OPENAI_API_BASE = os.getenv("OPENAI_API_BASE")
#llm = ChatOpenAI(openai_api_key=OPENAI_API_KEY, temperature=0, model_name='gpt-3.5-turbo',openai_api_base=OPENAI_API_BASE)
llm = AzureChatOpenAI(deployment_name="bitservice_chat_35",openai_api_base=OPENAI_API_BASE,openai_api_key=OPENAI_API_KEY,openai_api_version="2023-03-15-preview",model_name="gpt-3.5-turbo")


import torch
from transformers import BlipProcessor, BlipForConditionalGeneration

image_to_text_model = "Salesforce/blip-image-captioning-large"
device = 'cuda' if torch.cuda.is_available() else 'cpu'

processor = BlipProcessor.from_pretrained(image_to_text_model)
model = BlipForConditionalGeneration.from_pretrained(image_to_text_model).to(device)

from transformers.models.oneformer.modeling_oneformer import OneFormerModelOutput
import requests
from PIL import Image

def describeImage(image_url):
  image_object = Image.open(image_url).convert('RGB')
  # image
  inputs = processor(image_object, return_tensors="pt").to(device)
  outputs = model.generate(**inputs)
  return processor.decode(outputs[0], skip_special_tokens=True)


from langchain.tools import BaseTool

class DescribeImageTool(BaseTool):
    name = "Describe Image Tool"
    description = 'use this tool to describe an image.'

    def _run(self, url: str):
        description = describeImage(url)
        return description
    
    def _arun(self, query: str):
        raise NotImplementedError("Async operation not supported yet")

tools = [DescribeImageTool()]


agent = initialize_agent(
    agent='chat-conversational-react-description',
    tools=tools,
    llm=llm,
    verbose=True,
    max_iterations=3,
    early_stopping_method='generate',
    memory=ConversationBufferWindowMemory(
        memory_key='chat_history',
        k=5,
        return_messages=True
    )
)


import gradio as gr
def segment(image):
    #pass  # Implement your image segmentation model here...
    print(image)
    image_url = image
    
    return agent(f"Describe the following image:\n{image_url}").get('output').replace('The response to your last comment is','')

demo = gr.Interface(segment, gr.Image(type="filepath",shape=(200, 200)), "text")
demo.launch()