Scalable Python Telegram Bot Creation

I want to build a highly scalable Telegram bot using Python with unlimited users. The user limit will depend on the VPS configuration, but the code should be clean, efficient, and production-ready. The bot should be multi-model, integrating models from different providers: Vertex AI, Gemini, and Azure AI Foundry.

Models

Vertex AI: Veo 3 Fast (supports text-to-video and image-to-video)

Gemini API: Nano Banana

Azure AI Foundry:

Sora (text-to-video and image-to-video)

Flux Kontext Pro (text-to-image and image-to-image)

API Usage and Rate Limits

Use Google Cloud Service Accounts with Vertex AI permissions.

Rate limits:

Veo 3 Fast: 10 requests/min per service account

Gemini API: 500 requests/min per API key, 2000/day

Azure AI Foundry: 60k tokens/min, 60 requests/min per endpoint

Flux Kontext: 300 requests/min

Support up to 20 service accounts for each provider.

Automatically use available service accounts. If limits are reached, requests should wait until the next available time slot.

Requests must not exceed rate limits; otherwise, an error will be returned.

Nano Banana: allow up to 20 API keys

Azure AI Foundry: support up to 20 endpoints (default 3 for Sora, 1 for Flux Kontext), using only available endpoints.

Backend & Admin Panel

Endpoints and service accounts will be hosted in a PHP backend.

Provide a basic admin panel to add, edit, or manage keys for each model.

Bot Plans & First-Time User Flow

Users receive 1 free request per model when they first start the bot.

After the free request, users are prompted to upgrade to a paid plan using an admin-generated code.

Provide a command to upgrade immediately, without waiting to use free