Sassy_AI, the sarcastic assistant that doesn't really like humans

Description

Sassy_AI V2 is a portfolio-ready rewrite of a sarcastic assistant: from CLI project to a web chatbot with static frontend + FastAPI backend.

Core idea: useful, but with attitude.

• backend calls a real LLM via OpenAI Responses API with provider abstraction;
• tone controlled by a versioned persona and sarcasm_level (low/medium/high);
• safety policy handles unsafe requests and fallback behavior.

Goals

1. Build a clear, demo-ready web chatbot.
2. Replace stubs with a robust OpenAI provider integration.
3. Version and enforce a stable persona.
4. Keep scope strict: no auth, no DB, in-memory sessions.
5. Improve UX: typing reveal, skip, quick prompts, copy/reset, safe markdown rendering.

Tech stack

• Backend: Python 3.12, FastAPI, httpx, pytest
• Frontend: HTML/CSS/JS ES modules
• Markdown rendering: marked + DOMPurify
• Workflow: Spec-Kit + GitHub Copilot Agent
• Deployment: Docker + Hugging Face Spaces

Current status

✅ MVP V2 is functional:

• web chat works end-to-end
• persona + sarcasm behavior is consistent
• markdown output is safe and readable
• deployed on Hugging Face

Useful links

• Demo: https://ricolajazz-sassy-ai.hf.space
• GitHub: https://github.com/hericlibong/SassyAI