Telegram Bot with AI Analytics

I need a Telegram bot that greets a user, requests the “share phone number” action and treats that number as the sole identifier—no extra personal data at this stage. Once verified, every message the user sends must be piped to OpenAI’s ChatGPT with a fixed system prompt that frames the model as “Поддержка клиентов”; the reply immediately returns to the user inside the chat.

While the conversation unfolds, the bot should log everything to a database, then launch a background routine that does three things:
• compiles basic stats (message count, response time, keywords, sentiment)
• produces a concise summary of the dialogue plus an insight-oriented profile of the client
• scores the client against my predefined criteria and generates a step-by-step improvement plan

All of this data has to remain editable—both through an admin view in a lightweight web panel and via an endpoint that lets me patch records programmatically.

Keep secrets in environment variables, provide Docker files for one-command setup, and write clear inline comments so I can tweak the prompts later.

Deliverables:
1. Full source code with README and .env template
2. Docker-compose file (or alternative) for local and production launch
3. Simple admin interface for browsing, editing, exporting CSV/JSON
4. Short Loom or text guide that shows the bot passing the phone-number auth, chatting through ChatGPT in the “customer support” role, and generating the analytics output

I will consider the project complete when the bot is live on my server, stores data reliably, and the analytics panel reflects real conversations in near real-time.