A Telegram bot that formats text for Twitter, Bluesky, and Instagram. Detects names with spaCy, looks up handles via Bluesky API and Google (Serper), and splits long posts correctly per platform.

Talk to @BotFather, create a bot, copy the token.

Talk to @userinfobot to get your numeric user ID.

Sign up at serper.dev and copy your API key.

Required only if you enable auto_emoji. Sign up at console.anthropic.com.

cp .env.example .env
# fill in TELEGRAM_BOT_TOKEN, ADMIN_USER_ID, SERPER_API_KEY, and optionally ANTHROPIC_API_KEY

uv sync
uv run python -m spacy download en_core_web_md
uv run --env-file .env python bot.py

config.json will be created automatically in the project root on first run.

1. Push this repo to GitHub and create a new Railway project from the repo
2. Add a Volume in Railway, mounted at /data
3. Set these service variables in Railway:
   • TELEGRAM_BOT_TOKEN
   • ADMIN_USER_ID
   • SERPER_API_KEY
   • ANTHROPIC_API_KEY (optional, for auto-emoji)
   • DATA_DIR → /data
4. Deploy.

Railway will build the Docker image, install dependencies, and download the spaCy model at build time. config.json will be created at /data/config.json on first run and persist across redeploys.

• Send any text → bot detects names, lets you select which to look up, finds handles, asks you to confirm, then sends formatted output
• /start → run the setup wizard (platforms, prefixes, suffixes, ignored names)
• /config → same as /start, for updating settings
• /users → (admin only) manage which Telegram users can access the bot
• /cancel → abort current operation

Editable via /start or /config in Telegram.

• enabled: toggle a platform on/off
• allowed_users: Telegram user IDs permitted to use the bot (admin access only)
• auto_paragraph: reflow text into paragraphs before formatting
• auto_emoji: prepend an apt emoji to each paragraph using Claude (requires ANTHROPIC_API_KEY)

Splits happen after handle substitutions so character counts are accurate.