Assistant vs SimpleRequest

Community
1

Hi.
I have an app that allows me to ask questions about some images to GPT. I made the feature some time ago so i decided to clean my code a little.
At the begining i made my comm interface with the request method.

# Load each image from the paths
            images = []
            for path in image_paths:
                image = cv2.imread(path)
                if image is None:
                    raise FileNotFoundError(f"Unable to read image from {path}")
                
                images.append(image)

            payload = compose_payload(images=images, prompt=prompt)
            response = requests.post("https://api.openai.com/v1/chat/completions", headers=headers, json=payload)
            response = response.json()

Something like this. Basicly send a list of images and a question about the images and get the response.

I read about the assistant new feature and i have develop a simple class that allows me to do the same.
I’m not quite sure in the difference between this to methods to ask questions.

Whats the assistant for?
Maybe in my case there’s no difference at all or I’m missing something.

Could someone explain to me the difference between both methods to get a response.

This is the code that i use to communicate with the assistant method:

def analize_equipment(self,images_path : list[str], questions_list : list[str]) -> dict:
        """ Analize an equipment based on the images an questions list"""

        # Thread contnetn (CHAT GPT MESSAGES)
        thread_content = []

        # Generate a single prompt with all the questions
        entire_prompt = "Answer this questions: \n"
        for i,question in enumerate(questions_list):
            entire_prompt += f"{i}. {question} \n"

        # Add the prompt to the prompt content
        thread_content.append({"type": "text","text": entire_prompt})

        # Uplad and store the files ID to acces them in the prompt and after analsis remove them.
        files_ids = []        
        for path in images_path:

            # Upload file
            file = self.__client.files.create(
            file=open(path, "rb"),
            purpose="vision"
            ) 
            print(f"File {file.id} uploaded")

            # Store file ID
            files_ids.append(file.id)

            # Add the file to the prompt content
            thread_content.append(
            {
                "type": "image_file",
                "image_file": {"file_id": file.id, "detail": "high"}
            })
        

        # Crete a thread to make the questions
        thread = self.__client.beta.threads.create(
        messages=[  
          {
            "role": "user",
            "content": thread_content
          }
        ]
        )

        # Crete the assistant that will analize and send the request
        assistant = self.__create_assistant()
        run = self.__client.beta.threads.runs.create_and_poll(
            thread_id = thread.id,
            assistant_id=assistant.id
        )

        # If the run is cumpleted show the results
        if run.status == 'completed': 
            messages = self.__client.beta.threads.messages.list(
            thread_id=thread.id
            )
            print(messages.data[0].content[0].text.value)
        else:
            print(run.status)
2

Assistants have threads, so it might be more convenient if you want to build an actual chatbot. The assistants thing also has these vector stores and the code interpreter. Conveniences. If you don’t need those things, then assistants might hobble you more than they help, but this is probably a matter of taste :slight_smile:

3

Well, in context of the maintenance use case of @sebasviollaz and extended beyond it, imagine you are a car rental company.

Before a customer checkouts a car , the car rental company takes a set of pictures of the car and detects dents if any.

After the customer returns that car, another set of pictures and detects any dents… compares the two sets of dents in the same thread…if ok, then ok. If not…