I built a Personal AI Assistant for Telegram that deploys on Railway with 3 variables

Description and use

A lightweight Telegram AI assistant deployed on Railway that requires only three environment variables, offering persistent memory, voice transcription, and document analysis.

Problem

Users want a lightweight, self-hosted AI assistant that can be accessed directly within Telegram without additional web interfaces or complex infrastructure.

Outcome

The bot provides a fully functional, pay-as-you-go AI assistant within Telegram, with persistent conversation history, long-term memory, and media processing, all managed through a single Railway deployment.

Workflow

• Deploy the Railway template
• Set TELEGRAM_BOT_TOKEN, OWNER_TELEGRAM_ID, and OPENAI_API_KEY
• Start the bot and chat via Telegram
• Use commands /new, /memory, /remember, /forget
• Send voice messages, documents, or images for AI processing

Evidence

Deploy → enter 3 values → start chatting.

Source

The owner whitelist is checked before OpenAI requests and media processing, so random Telegram users can't burn your API credits.

Source