Built a small AI learning agent, looking for advice on the architecture
I recently built **Telusuko AI**, a small AI learning assistant for students.
🔗 [https://irfan95sayyad.github.io/Telusuko\\\_AI/\](https://irfan95sayyad.github.io/Telusuko\_AI/)
**Stack:** HTML + Bootstrap + Flowsie + Groq API + GitHub Pages.
The current flow is basically:
`Student → Flowsie Agent → Groq API → Response`
The problem is that Groq's API limit gets exhausted sometimes, so the agent stops responding until the limit resets.
I'm thinking about improving the architecture with things like **multiple LLM providers, fallback models, a backend/API layer, or caching**.
For those who have built LLM/AI agents:
**How would you architect this differently?**
Would you use an LLM gateway, multiple providers, or something else?
I'd really appreciate some practical advice from people who have experience building these systems.