What it does Wraps OpenAI and Azure OpenAI APIs into a Telegram bot with streaming replies, DALL·E 3 image generation, and 15 preset “bot identities” (system prompts). It tiers users by rate limits, context window size, and max tokens per response — all enforced via MySQL and configured in a single YAML file.

The interesting bit The project treats bot hosting as a multi-tenant service rather than a personal toy. User levels, time-window rate limiting, and per-level token budgets suggest it’s built for small groups or paid communities where “everyone gets GPT-4” would break the bank.

Key highlights

• Supports both native OpenAI and Azure OpenAI endpoints, switchable via config.
• Streaming responses with Telegram’s native button UI; includes a one-click “clear conversation” button.
• 15 preset personalities plus custom identity support.
• Docker and docker-compose deployment; MySQL 8 for persistence.
• Bilingual UI (English / Chinese).

Caveats

• The README’s “More…” under features is doing some heavy lifting — actual capabilities beyond the checklist are unclear.
• Requires manual MySQL setup and a config.yaml with half a dozen required fields; not a one-click deploy.
• The TODO list shows completed items (marked [x]) that read like missing features, which is slightly confusing.

Verdict Worth a look if you need to run a shared ChatGPT bot with basic access control and don’t want to build the Telegram plumbing yourself. Skip it if you just want a personal bot — there are lighter options.