Merge pull request #10 from flynnoct/dev

v1.1.0

README.md

@@ -1,81 +1,87 @@
-# chatgpt-telegram-bot
+# ChatGPT Bot for Telegram
 
-Telegram bot implemented by OpenAI ChatGPT API (gpt-3.5-turbo-0301) released on March 1, 2023.
+![](/docs/dialog.png)
+
+## News
+
+- **DALL·E, the OpenAI Image Generation Model**, is now supported! Send a short prompt to the Bot and get your own painting!
+- **Whisper, the OpenAI Intelligent Speech Recognizer**, is now supported! Now chat with the Bot with audio messages!
 
 ## Introduction
 
-This is a Telegram bot implemented by OpenAI ChatGPT API (gpt-3.5-turbo-0301). It is based on [OpenAI ChatGPT API](https://platform.openai.com/docs/guides/chat) and [python-telegram-bot](https://python-telegram-bot.org)
+ChatGPT Bot for Telegram is implemented with [OpenAI ChatGPT API](https://platform.openai.com/docs/guides/chat) released on March 1, 2023. The Telegram integration framework is based on [python-telegram-bot](https://python-telegram-bot.org).
+
+ChatGPT Bot can act as your Telegram contact. You can chat with it either personally or in a group chat. Just like the popular AI on the OpenAI official site, the Bot shares knowledge and inspires exciting new ideas. Many interesting features, such as **DALL·E** and **Whisper** are integrated together to make our Bot smarter and more usable.
 
-## Usage
+We hope you enjoy it!
 
-### Private Chat
+## Features
 
-You can just send a message to the bot in private chat. The bot will reply to you.
+The Telegram Bot features the following functions:
 
-### Group Chat
+- An AI consultant, based on OpenAI ChatGPT, interacts in a conversational way.
+- A flexible speech recognizer which supports audio interaction.
+- A AI painter reponses to user's requirement prompt.
 
-You can add the bot to a group chat. However it will only reply the messages with `@<bot_name>` mentioned.
+Additonal functions are also implemented:
 
-### Commands
+- Set the daily limitation of requirements to **DALL·E**.
+- Grant more resources to _Super Users_.
+
+## Commands
 
 - `/start`: Start the bot.
 - `/clear`: Clear the conversation context.
 - `/getid`: Get your Telegram user ID.
+- `/dalle <prompt>`: Ask DALL·E for a painting based on your prompt.
+
+## Sample Usage
 
-## Installation
+The Bot works in both personal and group chat of Telegram.
+In a personal chat, simply send a message to the Bot and it will reply to you.
+In a group chat, you need to tag the message with `@<bot_name>` to invoke the Bot.
 
-### Prepare
+### Preparation
 
 1. Create a Telegram bot by [@BotFather](https://t.me/BotFather) and get the token.
 2. Create an OpenAI account and get the API key.
-3. A VM or a server with Python 3 is needed to run the bot.
+3. A Linux VM or a server with Python 3 is needed to run the bot.
+4. A practical Internet environment is required.
+5. (Optional) [FFmpeg](https://ffmpeg.org) is required for the Bot to handle voice messages with Whisper. If you are not interested in using voice messages, you don't need to install it and **must set `enable_voice` in the config file to False**.
 
 > **Note**: You should disable the privacy mode of the bot. Otherwise the bot will not receive the messages from the group chat. You can do this by sending `/setprivacy` to [@BotFather](https://t.me/BotFather).
 
-### Deploy
-
-Clone this repository.
-
-```bash
-git clone [email protected]:flynnoct/chatgpt-telegram-bot.git
-```
+### Deployment
 
-Install the dependencies.
+Download the latest release version and install the dependencies.
 
 ```bash
+wget https://github.com/flynnoct/chatgpt-telegram-bot/releases/latest
 pip install -r requirements.txt
 ```
 
-Create config file.
+Then, you need to create a config file to manage the Bot. The config file includes sensitive information, such as telegram_token and openai_api_key, and we only release the corresponding template `config.json.template`. Therefore, you need to create a new `config.json` file and replace the relative fields with your own.
 
 ```bash
 cp config.json.template config.json
 ```
 
-Modify config file. First,
+Follow below procedures to fill you `config.json`:
 
-```bash
-vim config.json
-```
+1. Replace the `telegram_token` and `openai_api_key` with your own.
+2. Add allowed users to the `allowed_users` list. You can get your user id by sending `/start` to [@userinfobot](https://t.me/userinfobot) or send `/getid` to the Bot (after you start it).
 
-then, replace the `telegram_token` and `openai_api_key` with your own.
+> Note: the user ID is a series of numbers, you should add it to the `allowed_users` list as a string (add quotation marks around it).
 
-Add allowed users to the `allowed_users` list. You can get your user id by sending `/start` to [@userinfobot](https://t.me/userinfobot) or send `/getid` to this bot (after you start it).
+Now, you can run the Bot with `start_bot.sh` and try talk to it. Also, you can invite it to group chats and share with your friends!
 
-> Note: the user ID is a series of numbers, you should add it to the `allowed_users` list as a string (add quotation marks around it).
+To clear ChatGPT conversation context and restart the Bot, run shell script `restart_bot.sh`. To shut down the Bot, run `stop_bot.sh`.
 
-Run the bot.
+## Release version and notes
 
-```bash
-nohup python3 telegram_message_parser.py &
-```
-or
-```bash
-pm2 start telegram_message_parser.py
-```
-> Note: A launch script will be added later.
+The latest released version can be found [here](https://github.com/flynnoct/chatgpt-telegram-bot/releases/latest). More interesting new features are comming soon!
 
-Now you can start a private chat to the bot or add the bot to your group chat. Enjoy.
+The release notes are [here](/docs/release_notes.md).
 
 ## License
 
@@ -85,4 +91,4 @@ Now you can start a private chat to the bot or add the bot to your group chat. E
 
 If you like this project, you can buy me a coffee ❤️ or give this repository a free star ⭐️.
 
-Click [Alipay](donate_code/alipay.jpg) to open QR code.
+Click [Alipay](donate_code/alipay.jpg) to open QR code.

config.json.template

@@ -1,9 +1,16 @@
-{
+{ 
     "openai_api_key": "<YOUR_OPENAI_API_KEY_HERE>",
     "telegram_bot_token": "<YOUR_TELEGRAM_BOT_TOKEN_HERE>",
-    "allowed_users": [
-        "<USER_ID_1>",
-        "<USER_ID_2>"
-    ],
-    "wait_time": 300
+    "enable_voice": true,   // When enabled, Bot will accept audio messages with Whisper and reply.
+    "allowed_users": [ 
+        "<USER_ID_1>", 
+        "<USER_ID_2>", 
+    ], 
+    "wait_time": 600,   // The time limit in seconds that the Bot will clear the conversation context.
+    "enable_dalle": true,   // When enabled, Bot will involve DALL·E to handle requests for painting.
+    "super_users": [ 
+        "<SUPER_USER_ID_1>",    // Super users are granted unlimited usage on DALL·E per day.
+        "<SUPER_USER_ID_2>"
+    ], 
+    "image_generation_limit_per_day": 5     // The upper limit that a normal user is allowed to invoke DALL·E per day.
 }

message_manager.py

@@ -1,14 +1,39 @@
 import time
+import datetime
+import os
+import json
 from user_context import UserContext
 from openai_parser import OpenAIParser
 
 class MessageManager:
 
     userDict = {}
     openai_parser = None
+    config_dict = {}
+    user_image_generation_usage_dict = {}
+    user_chat_usage_dict = {}
+    # Fixed by @Flynn
+    usage_dict = {}
 
     def __init__(self):
         self.openai_parser = OpenAIParser()
+        # load config
+        with open("config.json") as f:
+            self.config_dict = json.load(f)
+
+        (image_usage_file_name, now) = self.__get_usage_filename_and_key("image")
+        if not os.path.exists("./usage"):
+            os.makedirs("./usage")
+        if os.path.exists("./usage/" + image_usage_file_name):
+            with open("./usage/" + image_usage_file_name) as f:
+                self.user_image_generation_usage_dict = json.load(f)
+        else:
+            self.user_image_generation_usage_dict = {}
+
+        # Fixed by @Flynn
+        if now not in self.user_image_generation_usage_dict:
+            self.user_image_generation_usage_dict[now] = {}
+
 
     def get_response(self, id, user, message):
 
@@ -31,7 +56,79 @@ def clear_context(self, id):
         except Exception as e:
             print(e)
 
+    def get_generated_image_url(self, user, prompt):
+
+        # Temporary fix by @Flynn, will be fixed in the next version
+        with open("config.json") as f:
+            super_users = json.load(f)["super_users"]
+        if user in super_users:
+            url = self.openai_parser.image_generation(user, prompt)
+            return (url, "Hey boss, it's on your account. 💰")
+        ############################
+
+
+        used_num = self.__check_image_generation_limit(user)
+        if used_num >= self.config_dict["image_generation_limit_per_day"]:
+            return (None, "You have reached the limit.")
+        else:
+            self.__update_usage_info(user, used_num+1, "image")
+            url = self.openai_parser.image_generation(user, prompt)
+            # Temporary fix by @Flynn
+            return (url, "You have used " + str(used_num + 1) + " / " + 
+                    str(self.config_dict["image_generation_limit_per_day"]) + 
+                    " times.")
+
+    def get_transcript(self, user, audio_file):
+        try:
+            return self.openai_parser.speech_to_text(user, audio_file)
+        except Exception as e:
+            print(e)
+            return ""
+
+    def __get_usage_filename_and_key(self, chatORimage):
+        if chatORimage == "chat":
+            filename = "_char_usage.json"
+        elif chatORimage == "image":
+            filename = "_image_generation_usage.json"
+        return (datetime.datetime.now().strftime("%Y%m") + filename, 
+                datetime.datetime.now().strftime("%Y-%m-%d"))
+
     def __sendMessage(self, user, messageList):
         ans = self.openai_parser.get_response(user, messageList)
         return ans
-
+
+    def __check_image_generation_limit(self, user):
+        (_, now) = self.__get_usage_filename_and_key("image")
+        if now not in self.user_image_generation_usage_dict:
+            self.__update_dict("image")
+        if user not in self.user_image_generation_usage_dict[now]:
+            used_num = 0
+        else:
+            used_num = self.user_image_generation_usage_dict[now][user]
+        return used_num
+
+    def __update_dict(self, chatORimage):
+        (filename, now) = self.__get_usage_filename_and_key(chatORimage)
+        if not os.path.exists("./usage/" + filename):
+            if chatORimage == "image":
+                self.user_image_generation_usage_dict = {}
+            elif chatORimage == "chat":
+                self.user_chat_usage_dict = {}
+            return  
+        if chatORimage == "image" and now not in self.user_image_generation_usage_dict:
+            self.user_image_generation_usage_dict[now] = {}
+        elif chatORimage == "chat" and now not in self.user_chat_usage_dict:
+            self.user_chat_usage_dict[now] = {}       
+
+    def __update_usage_info(self, user, used_num, chatORimage):
+        (filename, now) = self.__get_usage_filename_and_key(chatORimage)
+        if now not in self.user_image_generation_usage_dict:
+            self.__update_dict(chatORimage)
+        if chatORimage == "image":
+            self.user_image_generation_usage_dict[now][user] = used_num
+            with open("./usage/" + filename, "w") as f:
+                json.dump(self.user_image_generation_usage_dict, f)
+        elif chatORimage == "chat":
+            self.user_chat_usage_dict[now][user] = used_num
+            # with open("./usage/" + filename, "w") as f:
+            #     json.dump(self.user_chat_usage_dict, f)

openai_parser.py

@@ -44,6 +44,43 @@ def get_response(self, userid, context_messages):
             return response["choices"][0]["message"]["content"]
         except Exception as e:
             return str(e) + "\nSorry, I am not feeling well. Please try again."
+
+    def speech_to_text(self, userid, audio_file):
+        # transcript = openai.Audio.transcribe("whisper-1", audio_file, language="zh")
+        transcript = openai.Audio.transcribe("whisper-1", audio_file)
+        return transcript["text"]
+
+    def image_generation(self, userid, prompt):
+        response = openai.Image.create(prompt = prompt, n=1, size = "512x512", user = userid)
+        image_url = response["data"][0]["url"]
+        # self.update_image_generation_usage(userid)
+        # for debug use
+        # image_url = "https://catdoctorofmonroe.com/wp-content/uploads/2020/09/iconfinder_cat_tied_275717.png"
+        return image_url
+
+    def update_image_generation_usage(self, userid):
+        # get time
+        usage_file_name = datetime.datetime.now().strftime("%Y%m") + "_image_generation_usage.json"
+        now = datetime.datetime.now().strftime("%Y-%m-%d")
+        # create usage folder
+        if not os.path.exists("./usage"):
+            os.makedirs("./usage")
+        # load usage or create new
+        if os.path.exists("./usage/" + usage_file_name):
+            with open("./usage/" + usage_file_name) as f:
+                self.usage_dict = json.load(f)
+        else:
+            self.usage_dict = {}
+        # update usage
+        if now not in self.usage_dict:
+            self.usage_dict[now] = {}
+        if userid not in self.usage_dict[now]:
+            self.usage_dict[now][userid] = {"requests": 0}
+        self.usage_dict[now][userid]["requests"] += 1
+        # save usage
+        with open("./usage/" + usage_file_name, "w") as f:
+            json.dump(self.usage_dict, f)
+
 
     def update_usage(self, total_tokens, userid):
         # get time

restart_bot.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+# This script restarts a Python program named telegram_message_parser.py
+
+echo "Restarting bot..."
+
+pid=`ps -ef | grep telegram_message_parser.py | grep -v grep | awk '{print $2}'`
+
+if [ -z $pid ]; then
+    echo "Bot is not running"
+else
+    echo "Bot has been terminated"
+    kill -9 $pid
+fi
+
+echo "Starting bot..."
+nohup python3 telegram_message_parser.py >/dev/null 2>&1 &
+echo "Bot has been restarted successfully"

start_bot.sh

@@ -1,3 +1,5 @@
 #!/bin/bash
+# This script start a Python program named telegram_message_parser.py and run it in the background
 
-nohup python3 telegram_message_parser.py &
+nohup python3 telegram_message_parser.py >/dev/null 2>&1 &
+echo "Bot has been started successfully"

stop_bot.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+# This script stops a Python program named telegram_message_parser.py
+
+# Find the PID of the process
+pid=$(ps -ef | grep "telegram_message_parser.py" | grep -v grep | awk '{print $2}')
+
+if [ -z "$pid" ]; then
+  echo "Bot is not running"
+else
+  # Terminate the process
+  kill $pid
+  echo "Bot has been terminated"
+fi