---

Web chatbots are great — but they're often laggy, rate-limited, and locked behind a browser tab. This project puts the best models from OpenAI and Anthropic right inside Telegram, with your own API keys and no daily limits.

🚀 Try the live bot: @jadvebot · 🌐 Web: jadve.com — or deploy your own in two commands (Quick start ↓).

✨ Features

• ⚡ Low latency — replies usually take 3–5 seconds.
• 🔁 No request limits — you pay your API provider directly, nothing in between.
• 📝 Message streaming — answers stream into Telegram word-by-word.
• 🧠 Frontier models — GPT-5.5 and Anthropic Claude (Opus 4.8 / Sonnet / Haiku) via OpenRouter, plus fast & cheap defaults out of the box.
• 👁️ Vision — send an image and any vision-capable model (GPT-5.5, Claude, …) will read it.
• 🎨 Image generation — create images with OpenAI gpt-image-1 (switch to 👩‍🎨 Artist mode).
• 🎤 Voice messages — record a voice note and Whisper transcribes it for you.
• 🎭 15 chat modes — Assistant, Code Assistant, Psychologist, Elon Musk and more. Add your own in config/chat_modes.yml.
• 👥 Group chat support — run /help_group_chat for setup instructions.
• 💻 Code highlighting — formatted, readable code blocks.
• 🔒 Access control — restrict the bot to a list of allowed Telegram users.
• 💰 Balance tracking — see exactly how much you've spent on the API with /balance.

🧠 Supported models

Models are config-driven — add or remove any in config/models.yml with no code changes.

Plus gpt-image-1 for image generation and Whisper for voice transcription.

💡 OpenAI models use the native API. Claude and GPT-5.5 are routed through OpenRouter — just set openrouter_api_key and pick the model in /settings. Any other OpenRouter-routed model works too: declare provider: openrouter in config/models.yml.

🎭 Chat modes

Create your own by editing config/chat_modes.yml.

🚀 Quick start

1. Get an OpenAI API key.

2. (optional) Get an OpenRouter API key to use Claude and GPT-5.5.

3. Get a Telegram bot token from @BotFather.

4. Fill in your tokens and rename the config files:

mv config/config.example.yml config/config.yml
mv config/config.example.env config/config.env
# then edit config/config.yml — set telegram_token, openai_api_key (and openrouter_api_key for Claude/GPT-5.5)

5. 🔥 Run it:

docker-compose --env-file config/config.env up --build

That's it — message your bot on Telegram.

⚙️ Configuration

Main options in config/config.yml:

Per-model pricing and capabilities live in config/models.yml.

💬 Bot commands

🗂️ Project structure

bot/
  bot.py           # Telegram handlers, streaming, commands
  openai_utils.py  # model dispatch (OpenAI + OpenRouter), token counting, vision
  config.py        # loads config.yml / models.yml / chat_modes.yml
  database.py      # MongoDB storage for users & dialogs
config/
  config.yml       # your tokens & settings
  models.yml       # model catalog, pricing, capabilities
  chat_modes.yml   # chat-mode prompts

🛠️ Tech stack

python-telegram-bot 20.x · OpenAI Python SDK 1.x · MongoDB · Docker · Python 3.12

📰 Changelog highlights

• Jun 2026 — Switched image generation to gpt-image-1, refreshed the model menu and pricing.
• Mar 2026 — Added OpenRouter support → Claude (Opus / Sonnet / Haiku) and GPT-5.5.
• Feb 2026 — Migrated to the OpenAI Python 1.x SDK; Docker now runs on Python 3.12.
• 2025 — Config-driven model catalog: add or swap models without touching code.

See the full history in CHANGELOG.md.

❤️ Top donations

You can be on this list:

1. LilRocco — $11000 (!!!)
2. Mr V — $250
3. unexpectedsunday — $150

📄 License

MIT — do whatever you want, just keep the notice.