
Building a RAG app without a budget? The free tiers are generous, but they each have limits that quietly break your app.
This 3-page cheat sheet puts them in one place:
The exact free limits for Groq (openai/gpt-oss-20b), Qdrant Cloud and Render, and which one runs out first in a RAG app
Why sentence-transformers usually crashes a 512 MB free host, and the alternative that doesn’t
The LangChain 1.x imports that replaced the 2024 tutorial code
Why tutorials using llama-3.1-8b-instant now fail on Groq’s free tier, and the replacement
A copy-paste RAG chain with rate-limit retries
A grounding prompt that makes the model refuse instead of making things up
A pre-deploy checklist
Free. Pay what you want if it saves you an evening of debugging.