adii0906/supportiq

SupportIQ is a customer support tool that replies to customers in their own language, delivered as a short video of a talking avatar instead of plain text. The idea it is built around: when a customer writes in Spanish, Hindi, French, Arabic, Japanese, or another language and gets a reply in English, they often feel unheard and leave. SupportIQ tries to close that gap by detecting the language a customer wrote in, writing an empathetic reply in that same language, and generating a lip-synced avatar video of that reply. The system runs as a five-step automated pipeline. First, a language model detects the customer's language, emotional tone, and the type of request. Second, a routing step decides whether to answer with a video avatar, used for non-English messages or angry customers, or with a plain text reply for simple English messages. Third, the pipeline writes an empathetic response in the customer's own language. Fourth, the HeyGen platform renders a video of an avatar speaking that response with matching lip movement. Fifth, the finished video streams back to the customer, with the whole process taking about ten seconds. Under the hood, the frontend is built with React and Vite, the backend runs on FastAPI and Python, and LangGraph with LangChain coordinates the five agent steps. The language understanding and reply writing run on Groq's LLaMA 3.3 70B model, while the avatar video itself is generated through the HeyGen API. Running the project locally needs two API keys, one from Groq and one from HeyGen, both of which offer free tiers. The project was built during a 24-hour HeyGen Hackathon in 2026, and the README frames it as a proof of concept aimed at that competition's agent track rather than a finished commercial product.