telegram-gemini-bot

Telegram Gemini Bot

A telegram bot which answers to messages with Gemini API.

• You can reply to messages for keeping the context of your conversation:

• You can also upload photos with a caption as a prompt:

• Generated text will be received part by part with streaming support:

Prerequisites

• A Telegram Bot Token,
• A Google API key, and
• A machine which can build and run golang applications.
• (Optional) ffmpeg installed for speech generation.

Configurations

Create a configuration file:

$ cp config.json.sample config.json
$ $EDITOR config.json

and set your values:

{
  "google_generative_model": "gemini-3.1-pro-preview",
  "google_generative_model_for_image_generation": "gemini-3-pro-image-preview",
  "google_generative_model_for_video_generation": "veo-3.1-generate-preview",
  "google_generative_model_for_speech_generation": "gemini-2.5-flash-preview-tts",

  "allowed_telegram_users": ["user1", "user2"],
  "db_filepath": null,
  "answer_timeout_seconds": 180,
  "verbose": false,

  "telegram_bot_token": "123456:abcdefghijklmnop-QRSTUVWXYZ7890",
  "google_ai_api_key": "ABCDEFGHIJK1234567890"
}

You can get appropriate model names from here.

If db_filepath is given, all prompts and their responses will be logged to the SQLite3 file.

Using Infisical

You can use Infisical for saving & retrieving your bot token and api key:

{
  "google_generative_model": "gemini-3.1-pro-preview",
  "google_generative_model_for_image_generation": "gemini-3-pro-image-preview",
  "google_generative_model_for_video_generation": "veo-3.1-generate-preview",
  "google_generative_model_for_speech_generation": "gemini-2.5-flash-preview-tts",

  "allowed_telegram_users": ["user1", "user2"],
  "db_filepath": null,
  "answer_timeout_seconds": 180,
  "verbose": false,

  "infisical": {
    "client_id": "012345-abcdefg-987654321",
    "client_secret": "aAbBcCdDeEfFgG0123456789xyzwXYZW",

    "project_id": "012345abcdefg",
    "environment": "dev",
    "secret_type": "shared",

    "telegram_bot_token_key_path": "/path/to/your/KEY_TO_TELEGRAM_BOT_TOKEN",
    "google_ai_api_key_key_path": "/path/to/your/KEY_TO_GOOGLE_AI_API_KEY"
  }
}

Build

$ go build

Run

Run the built binary with the config file's path:

$ ./telegram-gemini-bot path-to/config.json

Run as a systemd service

Create a systemd service file:

[Unit]
Description=Telegram Gemini Bot
After=syslog.target
After=network.target

[Service]
Type=simple
User=ubuntu
Group=ubuntu
WorkingDirectory=/dir/to/telegram-gemini-bot
ExecStart=/dir/to/telegram-gemini-bot/telegram-gemini-bot /path/to/config.json
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target

and systemctl enable|start|restart|stop the service.

Commands

• /image <PROMPT> for image generation.
• /video <PROMPT> for video generation.
• /speech <PROMPT> for speech generation.
• /google <PROMPT> for generation with grounding (Google Search).
• /stats for various statistics of this bot.
• /help for help message.

Todos / Known Issues

• Handle markdown texts gracefully.