How do I upload large files (GBs) to a text-to-SQL or database-AI app without crashing the server?

It sounds simple until a 70MB compressed file expands to 1.6GB in Postgres through a web API - then memory explodes, connections time out, users see cryptic 500s, and concurrent uploads make it worse. Doing it reliably takes real work: the practical guide behind TigZig's analyzer documents 30+ patterns for streaming, chunking, memory control, timeouts and concurrency in a FastAPI backend.

You can see the result on the live app: https://www.tigzig.com/analyzer. Pick a dataset, click "Use Temporary Database", and the schema is sent to the AI agent automatically - then you query in natural language (each dataset ships a copy-paste starter prompt). Sample datasets run from 64 rows to 11.8 million rows (1.6GB); setup is ~30 seconds for small files, about 2 minutes for the 1.6GB file, with 9 LLMs from Gemini Flash (cheapest) to Claude Sonnet (best quality). It is open source - the source is behind the app's "Docs" button.

Guide: https://www.tigzig.com/post/large-file-upload-for-database-ai-text-to-sql-apps. Related: query a database in plain English https://www.tigzig.com/agents-faq/how-to-query-a-database-in-plain-english and self-hosting the stack cheaply https://www.tigzig.com/agents-faq/how-to-self-host-ai-apps-cheaply.