Temaprint AI

Multi-Model AI Chatbot for Telegram with Streaming Responses

Open-source Telegram bot that connects users to 10+ AI models through OpenRouter. Real-time streaming, multi-chat sessions, bilingual UI, admin panel — ready to deploy in minutes.

Features · Quick Start · Tech Stack · Architecture · Commands

---

Why Temaprint AI?

Most Telegram AI bots lock you into a single model. Temaprint AI gives your users instant access to GPT-4o, Claude, Gemini, Llama, Mistral, and more — with per-chat model switching, streaming responses that edit in real time, and a full admin panel for user management.

✨ Features

Chat

• Streaming AI responses — messages update live in Telegram as the model generates text
• 10+ models via OpenRouter — GPT-4o, Claude, Gemini, Llama, Mistral, Qwen, and more
• Multiple chat sessions — each user can run several independent conversations
• Per-chat model switching — change the AI model without losing context
• Custom system prompts — personalize AI behavior per chat
• Stop generation — cancel a response mid-stream with one tap

Admin & Access

• Telegram-based access control — whitelist users by Telegram ID
• Admin commands — /approve, /revoke, /users to manage access
• Usage statistics — track tokens, messages, and costs per user
• Rate limiting — in-memory sliding window (configurable)

UX

• Bilingual UI — English / Russian with in-app language switcher
• Inline keyboards — model picker, settings, chat list — no typing needed
• Persistent history — all chats stored in PostgreSQL, survives restarts

DevOps

• Docker-ready — multi-stage Dockerfile, docker compose up -d and you're live
• Health check endpoint — /health monitors app + database connectivity
• Environment-based config — all secrets via .env, zero hardcoded credentials

Tech Stack

Layer	Technology
Runtime	Python 3.12, async/await throughout
Bot Framework	aiogram 3.x (async Telegram Bot API)
Web Framework	FastAPI + Uvicorn
AI Gateway	OpenRouter API (unified multi-model access)
Database	PostgreSQL 16 + SQLAlchemy 2.0 async
Containerization	Docker multi-stage build
Config	pydantic-settings from .env

Quick Start

Prerequisites

• Docker & Docker Compose
• Telegram Bot Token (@BotFather)
• OpenRouter API Key (openrouter.ai)

Run

# 1. Clone the repo
git clone https://github.com/YOUR_USERNAME/ai.temaprint.com.git
cd ai.temaprint.com

# 2. Configure environment
cp .env.example .env
# Edit .env — fill in BOT_TOKEN, OPENROUTER_API_KEY, your Telegram ID

# 3. Launch
docker compose up -d

# 4. Open Telegram and send /start to your bot

That's it. Your AI chatbot is live.

Architecture

app/
  main.py              # FastAPI app + bot polling entry point
  config.py            # pydantic-settings (reads .env)
  bot.py               # aiogram Dispatcher & router registration
  database/
    models.py          # SQLAlchemy models — User, Chat, Message, UsageLog
    session.py         # async engine + sessionmaker
  services/
    openrouter.py      # streaming client for OpenRouter API
    chat.py            # chat & message CRUD, usage stats
    auth.py            # user resolution & access control
  handlers/
    start.py           # /start /new /clear /model /help /menu
    message.py         # streaming message handler + FSM for prompts
    callbacks.py       # all inline keyboard callbacks
    admin.py           # /approve /revoke /users /admin
  keyboards/
    inline.py          # inline keyboards + model list builder
  utils/
    i18n.py            # en/ru translations + t() helper
    text.py            # message splitting utilities
    users.py           # shared user resolution
    rate_limit.py      # sliding-window rate limiter

Commands

Command	Description
/start	Welcome screen + authorization check
/new	Create a new chat session
/clear	Clear current chat history
/model	Switch AI model for current chat
/menu	Show main menu
/help	Help & usage instructions
/admin	Admin panel (admins only)
/approve <id>	Grant access to a user
/revoke <id>	Revoke user access
/users	List all registered users

Adding Users

New users are denied by default. The approval flow:

1. User sends /start → bot shows their Telegram ID
2. Admin sends /approve 123456789
3. User sends /start again → access granted

Only users in ALLOWED_TELEGRAM_IDS get instant access on first /start.

Configuration

All settings are managed through .env (see .env.example):

Variable	Required	Description
BOT_TOKEN	Yes	Telegram Bot Token from @BotFather
OPENROUTER_API_KEY	Yes	API key from openrouter.ai
DATABASE_URL	No	PostgreSQL connection string
ALLOWED_TELEGRAM_IDS	No	Comma-separated Telegram IDs with instant access
ADMIN_IDS	No	Comma-separated admin Telegram IDs
DEFAULT_MODEL	No	Default AI model (default: openai/gpt-4o-mini)

Contributing

Contributions are welcome! Feel free to open issues or submit pull requests.

License

Private project. All rights reserved.