initial commit

.gitignore

@@ -0,0 +1,2 @@
+docker-compose.yml
+venv/

Dockerfile

@@ -0,0 +1,8 @@
+FROM python:3-bullseye
+
+RUN apt update && apt install -y ffmpeg
+WORKDIR /app
+COPY ./main.py /app
+COPY ./requirements.txt /app
+RUN pip install -r requirements.txt
+CMD [ "python3", "/app/main.py" ]

README.md

@@ -0,0 +1,55 @@
+# OpenAI API Telegram Bot
+
+A telegram bot to interact with OpenAI API. You can:
+
+- Generate images with DALL-E: `/imagine`
+- Chat with ChatGPT: Just chat!
+- Transcribe audio to text: Just send a voice message!
+
+Other features include:
+
+- Clear ChatGPT context history (to save tokens).
+- Per-user context and usage metrics.
+- No database, data is saved on-memory.
+
+## Self-hosting
+
+Self hosting this chatbot is pretty easy. You just need to follow this steps:
+
+1. Get your OPENAI Token from [here](https://platform.openai.com/account/api-keys).
+   - You will need to set a payment method.
+
+2. Generate a new bot:
+   1. Start a new chat with [@BotFather](https://t.me/BotFather) bot.
+   2. Type `/newbot` command. Set a name and a username.
+   3. Copy the Token that BotFather will give you.
+   
+3. Get allowed users' IDs:
+   1. Start a new chat with [@RawDataBot](https://t.me/RawDataBot).
+   2. Send a message or forward a message from any desired user.
+   3. Copy the `message.from.id` field value. Should be an ID like: 1234567890
+
+4. Setup the bot:
+   1. Clone this repo.
+   2. Rename the `example.docker-compose.yml` file to `docker-compose.yml`.
+   3. Edit the environment variables:
+      1. Set your OPENAI_TOKEN.
+      2. Set your BOT_TOKEN.
+      3. Set your ALLOWED_USERS (comma separated user ids).
+      4. Set the SYSTEM_PROMPT for ChatGPT. This is always instructed to ChatGPT as the system.
+   4. Build and start the bot: `docker compose up --build -d`.
+   
+5. Enjoy!
+
+## Usage
+
+- By simply sending normal chat messages, you will be interacting with ChatGPT.
+  - The last 5 messages (from user and bot) are sent to ChatGPT API as context.
+  - Use the command `/clear` to clear your context and start over a completely new chat.
+
+- By using the `/imagine <prompt>` command, you will be getting images generated by DALL-E.
+    - Example: `/imagine A cat with a red hat`.
+
+- Sending a voice message to the bot, it will transcribe it to text using Whisper.
+
+- `/info` command allows you to see your usage statistics.

example.docker-compose.yml

@@ -0,0 +1,10 @@
+version: '3.9'
+services:
+  chatbot:
+    build: .
+    environment:
+      - OPENAI_API_KEY=YourApiKey
+      - CHATGPT_SYSTEM_PROMPT=You are a helpful assistant.
+      - CHATGPT_MAX_USER_CONTEXT=5
+      - BOT_TOKEN=XXXXXX:AAAAAAAAAAAA-AAA-AAAAAA
+      - BOT_ALLOWED_USERS=5xxxxxxxx2,6xxxxxxxx1

main.py

@@ -0,0 +1,168 @@
+import os
+import re
+import openai
+import logging
+from pydub import AudioSegment
+from telegram import Update
+from functools import wraps
+from telegram.ext import ApplicationBuilder, CommandHandler, ContextTypes, MessageHandler, filters
+
+logging.basicConfig(
+    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
+    level=logging.INFO
+)
+logger = logging.getLogger(__name__)
+
+openai.api_key = os.getenv("OPENAI_API_KEY")
+users = {
+    "userid": {
+        "context": [],
+        "usage": {
+            "chatgpt": 0,
+            "whisper": 0,
+            "dalle": 0,
+        }
+    },
+}
+ALLOWED_USERS=[]
+SYSTEM_PROMPT=os.getenv("CHATGPT_SYSTEM_PROMPT")
+MAX_USER_CONTEXT=int(os.getenv("CHATGPT_MAX_USER_CONTEXT"))
+
+def restricted(func):
+    @wraps(func)
+    async def wrapped(update, context, *args, **kwargs):
+        user_id = update.effective_user.id
+        if str(user_id) not in ALLOWED_USERS:
+            if "*" != ALLOWED_USERS[0]:
+                print(f"Unauthorized access denied for {user_id}.")
+                return
+        else:
+            if not f"{update.effective_chat.id}" in users:
+                users[f"{update.effective_chat.id}"] = {"context": [], "usage": {"chatgpt": 0,"whisper": 0,"dalle": 0,}}
+        return await func(update, context, *args, **kwargs)
+    return wrapped
+
+@restricted
+async def start(update: Update, context: ContextTypes.DEFAULT_TYPE):
+    if not f"{update.effective_chat.id}" in users:
+        users[f"{update.effective_chat.id}"] = {"context": [], "usage": {"chatgpt": 0,"whisper": 0,"dalle": 0,}}
+    await context.bot.send_message(chat_id=update.effective_chat.id, text="I'm a bot, please talk to me!")
+
+@restricted    
+async def imagine(update: Update, context: ContextTypes.DEFAULT_TYPE):
+    users[f"{update.effective_chat.id}"]["usage"]['dalle'] += 1
+    response = openai.Image.create(
+        prompt=update.message.text,
+        n=1,
+        size="1024x1024"
+    )
+    try:
+        image_url = response['data'][0]['url']
+        await context.bot.send_message(chat_id=update.effective_chat.id, text=image_url)
+    except:
+        await context.bot.send_message(chat_id=update.effective_chat.id, text="Error generating. Your prompt may contain text that is not allowed by OpenAI safety system.")
+    
+@restricted
+async def attachment(update: Update, context: ContextTypes.DEFAULT_TYPE):
+    print(update.message)
+    try:
+        users[f"{update.effective_chat.id}"]["usage"]['whisper'] += update.message.voice.duration
+        file = await context.bot.get_file(update.message.voice.file_id)
+        await file.download_to_drive(f"{update.effective_user.id}.ogg")
+        ogg_audio = AudioSegment.from_file(f"{update.effective_user.id}.ogg", format="ogg")
+        ogg_audio.export(f"{update.effective_user.id}.mp3", format="mp3")
+        os.remove(f"{update.effective_user.id}.ogg")
+        audio_file= open(f"{update.effective_user.id}.mp3", "rb")
+        try:
+            transcript = openai.Audio.transcribe("whisper-1", audio_file)
+        except:
+            await context.bot.send_message(chat_id=update.effective_chat.id, text="Transcript failed.")
+        os.remove(f"{update.effective_user.id}.mp3")
+        if transcript['text'] == "":
+            transcript['text'] = "[Silence]"
+        await context.bot.send_message(chat_id=update.effective_chat.id, text=transcript['text'])
+    except:
+        await context.bot.send_message(chat_id=update.effective_chat.id, text="Can't handle such file. Reason: unkown.")
+
+@restricted
+async def chat(update: Update, context: ContextTypes.DEFAULT_TYPE):
+    if not f"{update.effective_chat.id}" in users:
+        users[f"{update.effective_chat.id}"] = {"context": [], "usage": {"chatgpt": 0,"whisper": 0,"dalle": 0,}}
+    
+    # Save context
+    if len(users[f"{update.effective_chat.id}"]["context"]) <= MAX_USER_CONTEXT:
+        users[f"{update.effective_chat.id}"]["context"].append({"role": "user", "content": f"{update.message.text}"})
+    else:
+        users[f"{update.effective_chat.id}"]["context"].pop(0)
+        users[f"{update.effective_chat.id}"]["context"].append({"role": "user", "content": f"{update.message.text}"})
+    
+    # Interact with ChatGPT api   
+    response = openai.ChatCompletion.create(
+        model="gpt-3.5-turbo",
+        messages=[{"role": "system", "content": SYSTEM_PROMPT}]+users[f"{update.effective_chat.id}"]["context"]
+    )
+    
+    # Save context
+    if len(users[f"{update.effective_chat.id}"]["context"]) <= MAX_USER_CONTEXT:
+        users[f"{update.effective_chat.id}"]["context"].append({"role": "assistant", "content": f"{response['choices'][0]['message']['content']}"})
+    else:
+        users[f"{update.effective_chat.id}"]["context"].pop(0)
+        users[f"{update.effective_chat.id}"]["context"].append({"role": "assistant", "content": f"{response['choices'][0]['message']['content']}"})
+    
+    # Update usage
+    users[f"{update.effective_chat.id}"]["usage"]['chatgpt'] += len(str(users[f"{update.effective_chat.id}"]["context"]))
+    
+    # Send reply
+    await context.bot.send_message(chat_id=update.effective_chat.id, text=response['choices'][0]['message']['content'])
+
+@restricted
+async def clear(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None:
+    if f"{update.effective_chat.id}" in users:
+        users[f"{update.effective_chat.id}"]["context"] = []
+    print(f"Cleared context for {update.effective_chat.id}")
+    await update.message.reply_text(f'Your message context history was cleared.')
+
+@restricted
+async def usage(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None:
+    user_info=users[f"{update.effective_chat.id}"]["usage"]
+    total_spent=0.0
+    total_spent+=(user_info['chatgpt']/750)*0.002
+    total_spent+=float(user_info['dalle'])*0.02
+    total_spent+=(user_info['whisper'])*0.006 
+    info_message=f"""User: {update.effective_user.name}\n- Used {user_info["chatgpt"]} characters with ChatGPT.\n- Generated {user_info["dalle"]} images with DALL-E.\n- Transcribed {user_info["whisper"]}min with Whisper.\n\nTotal spent: ${str(total_spent)}"""
+    await context.bot.send_message(chat_id=update.effective_chat.id, text=info_message)
+    
+@restricted
+async def _help(update: Update, context: ContextTypes.DEFAULT_TYPE) -> None:
+    help_message="""Here's what you can do:\n\n- /imagine <prompt> to generate an image with DALL-E\n- Send a message to chat with ChatGPT\n- Send an audio to transcribe to text with Whisper.\n\n- /usage To get your usage statistics.\n - /clear To clear you chatgpt message context (start a new chat)."""
+    await context.bot.send_message(chat_id=update.effective_chat.id, text=help_message)
+
+if __name__ == '__main__':
+    try:
+        ALLOWED_USERS=os.getenv("BOT_ALLOWED_USERS").split(",")
+    except:
+        ALLOWED_USERS=ALLOWED_USERS
+    print(f"Allowed users: {ALLOWED_USERS}")
+    print(f"System prompt: {SYSTEM_PROMPT}")
+    
+    application = ApplicationBuilder().token(os.getenv("BOT_TOKEN")).build()
+    
+    start_handler = CommandHandler('start', start)
+    application.add_handler(start_handler), 
+    
+    clear_handler = CommandHandler('clear', clear)
+    application.add_handler(clear_handler), 
+    
+    info_handler = CommandHandler('usage', usage)
+    application.add_handler(info_handler), 
+    
+    help_handler = CommandHandler('help', _help)
+    application.add_handler(help_handler), 
+    
+    imagine_handler = CommandHandler('imagine', imagine)
+    application.add_handler(imagine_handler), 
+    
+    application.add_handler(MessageHandler(filters.TEXT & ~filters.COMMAND, chat))
+    application.add_handler(MessageHandler(filters.ATTACHMENT & ~filters.COMMAND, attachment))
+    
+    application.run_polling()

requirements.txt

@@ -0,0 +1,3 @@
+openai
+python-telegram-bot
+pydub