kopia lustrzana https://codeberg.org/pluja/openai-telegram-bot
77 wiersze
3.2 KiB
Markdown
77 wiersze
3.2 KiB
Markdown
[](https://nogithub.codeberg.page)
|
|
|
|
# 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 and video to text: Just send a voice message or a video file!
|
|
- Voice chat with ChatGPT:
|
|
- Send voice message.
|
|
- Receive voice messages.
|
|
- Use 100% local Text-To-Speech with Language Recognition to give ChatGPT a voice in many languages!
|
|
|
|
Other features include:
|
|
|
|
- Talk to ChatGPT with audio transcriptions (whisper).
|
|
- Clear ChatGPT context history (to save tokens).
|
|
- Reply to any message to use it as context for ChatGPT.
|
|
- Per-user context.
|
|
- See usage metrics and spent $.
|
|
- Persistent data with sqlite3 database.
|
|
- Lightweight: few lines of code.
|
|
|
|
[Jump to selfhosting guide](#self-hosting)
|
|
|
|
|
|
|
|
|
|
## 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.env` file to `.env`.
|
|
3. Edit the environment variables from the `.env` file:
|
|
- Set your OPENAI_TOKEN.
|
|
- Set your BOT_TOKEN.
|
|
- Set your ALLOWED_USERS (comma separated user ids). Set it to `*` to allow all users.
|
|
- Set the SYSTEM_PROMPT for ChatGPT. This is always instructed to ChatGPT as the system.
|
|
- Optional: Edit the MAX_CONTEXT. This variable sets the number of messages that will be sent to ChatGPT API as context for the conversation.
|
|
- WHISPER_TO_GPT allows you to choose wether Whisper transcripts should be instructed to ChatGPT or not.
|
|
- You can also configure this using `/settings` in chat.
|
|
- ENABLE_GOOGLE_TTS the TTS service will be provided by GoogleTTS, producing more natural voices. If disabled, it fallsback to local voice generation using Espeak.
|
|
- VOICE_LANGUAGE country code for the default voice accent.
|
|
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.
|
|
|
|
- Using `/setttings` you can configure a few settings.
|
|
|
|
- `/info` command allows you to see your usage statistics. |