synochatgpt.py

import shelve
import time
import os
from openai import OpenAI

from flask import Flask, request
from synochat.webhooks import OutgoingWebhook
from synochat.webhooks import IncomingWebhook

app = Flask(__name__)

# In-memory chat history storage
chat_history = {}

# This system prompt sets up the character of the chatbot; change it if you want
chatbot_character = '''You are a AI assistant, a friend of mine, trying to help me and my family as much as possible and in whatever ways you can.
    If the user talks to you in English, you respond in English. If the user talks to you in Chinese, you respond in Chinese.
    Be talkative, personable, friendly, positive, and speak always with love.'''

# Set maximum chat exchanges or idle time gap to start a new conversatoin
max_chat_length = 100 

def process_gpt_response(webhook):
    system_prompt = chatbot_character
    user_id, username = webhook.user_id, webhook.username
    current_timestamp = int(time.time())

    # Open the shelve file to store chat history
    with shelve.open('chat_history', writeback=True) as chat_history:

        # Maintain chat history
        if user_id not in chat_history:
            chat_history[user_id] = {"username": username, "messages": [], "last_timestamp": current_timestamp}
        else:
            chat_history[user_id]["last_timestamp"] = current_timestamp
            # Truncate chat history if it exceeds the maximum length
            if len(chat_history[user_id]["messages"]) > max_chat_length:
                chat_history[user_id]["messages"] = chat_history[user_id]["messages"][-max_chat_length:]

        # Append the user's message
        chat_history[user_id]["messages"].append({"role": "user", "content": webhook.text})

        # Prepare the message list for the API call
        messages = [{"role": "system", "content": system_prompt}]
        for entry in chat_history[user_id]["messages"]:
            messages.append({"role": entry['role'], "content": entry['content']})

        client = OpenAI(
            base_url='http://localhost:11434/v1/',
            api_key='ollama',  # required but ignored
        )

        # Make the API call to the model
        response = client.chat.completions.create(
            messages=messages,
            model='qwen2.5:7b',
        )

        # Process the response and update chat history
        response_role = response.choices[0].message.role
        if response.choices[0].finish_reason == "stop":
            response_text = response.choices[0].message.content
            chat_history[user_id]["messages"].append({"role": response_role, "content": response_text})
        else:
            response_text = f"error: stop reason - {response.choices[0].finish_reason}"
            chat_history[user_id]["messages"].append({"role": response_role, "content": response_text})

    return response_text

@app.route('/echo', methods=['POST'])
def echo():
    token = os.environ.get("SYNOLOGY_TOKEN")
    webhook = OutgoingWebhook(request.form, token, verbose=True)

    if not webhook.authenticate(token):
        return webhook.createResponse('Outgoing Webhook authentication failed: Token mismatch.')

    hostname = os.environ.get("SYNOLOGY_INCOMING_URL")
    bot = IncomingWebhook(hostname, port=5002, token=token, user_ids=[webhook.user_id])
    if app.debug:
        print("\n" + webhook.text + "\n")

    reply = process_gpt_response(webhook)

    if app.debug:
        print("\n" + reply + "\n")

    while (len(reply) > 2000):
        bot.send(reply[:2000])
        reply = reply[2000:]
    bot.send(reply)
    return "echo completed"

if __name__ == '__main__':
   app.run('0.0.0.0', port=5001, debug = True)