app.py

import gradio as gr
import openai
from dotenv import load_dotenv
import os
import time
from gradio_client import Client



title = "# Welcome to 🙋🏻‍♂️Tonic's🕵🏻‍♂️Tulu🪴Plant👩🏻‍⚕️Doctor!"
description = """Here you can use Bulbi - an OpenAI agent that helps you save your plants with [Allen-AI](https://huggingface.co/allenai/tulu-2-dpo-70b) [allenai/tulu-2-dpo-13b](https://huggingface.co/allenai/tulu-2-dpo-13b)
Use [Tulu](https://huggingface.co/allenai/tulu-2-dpo-7b) to fix your plants!
### How to use:
- Introduce your🌵plant below. 
- Be as🌿descriptive as possible. 
- **Respond with additional🗣️information when prompted.**
- Save your plants with👨🏻‍⚕️Bulbi Plant Doctor! 
### Join us:
[Join my active builders' server on discord](https://discord.gg/VqTxc76K3u). Let's build together!
Big thanks to 🤗Huggingface Organisation for the🫂Community Grant"""

examples = [
    ["My Eucalyptus tree is struggling outside in the cold weather in Europe",True, None]
]

load_dotenv()
openai.api_key = os.getenv('OPENAI_API_KEY')
assistant_id = os.getenv('ASSISTANT_ID')
client = openai.OpenAI(api_key=openai.api_key)
thread_ids = {}
current_thread_id = None
gradio_client = Client("https://tonic1-tulu.hf.space/--replicas/tjvh5/")

def ask_openai(question, start_new_thread=True, selected_thread_id=None):
    global thread_ids

    try:
        if start_new_thread or selected_thread_id not in thread_ids:
            thread = client.beta.threads.create()
            current_thread_id = thread.id
            thread_ids[current_thread_id] = thread.id
        else:
            current_thread_id = thread_ids[selected_thread_id]

        client.beta.threads.messages.create(
            thread_id=current_thread_id,
            role="user",
            content=question,
        )

        run = client.beta.threads.runs.create(
            thread_id=current_thread_id,
            assistant_id=assistant_id
        )

        response_received = False
        timeout = 150
        start_time = time.time()

        while not response_received and time.time() - start_time < timeout:
            run_status = client.beta.threads.runs.retrieve(
                thread_id=current_thread_id,
                run_id=run.id,
            )
            if run_status.status == 'completed':
                response_received = True
            else:
                time.sleep(4)

        if not response_received:
            return "Response timed out."

        steps = client.beta.threads.runs.steps.list(
            thread_id=current_thread_id,
            run_id=run.id
        )

        if steps.data:
            last_step = steps.data[-1]
            if last_step.type == 'message_creation':
                message_id = last_step.step_details.message_creation.message_id
                message = client.beta.threads.messages.retrieve(
                    thread_id=current_thread_id,
                    message_id=message_id
                )
                if message.content and message.content[0].type == 'text':
                    response_text = message.content[0].text.value
                else:
                    return "No response."
        else:
            return "No response."

        final_result = gradio_client.predict(
            response_text,  
            "I am Tulu, an Expert Plant Doctor, I will exactly summarize the information you provide to me.",
            450, 0.4, 0.9, 0.9, False, fn_index=0
        )

        return final_result

    except Exception as e:
        return f"An error occurred: {str(e)}"

    except Exception as e:
        return f"An error occurred: {str(e)}"

iface = gr.Interface(
    title=title,
    description=description,
    fn=ask_openai, 
    inputs=[
        gr.Textbox(lines=5, placeholder="Hi there, I have a plant that's..."),
        gr.Checkbox(label="Start a new conversation thread"),
        gr.Dropdown(label="Select previous thread", choices=list(thread_ids.keys()))
    ], 
    outputs=gr.Markdown(),
    examples=examples
)

iface.launch()