Uses an LLM to identify where tables actually start/end in Excel/CSV files, then extracts clean, typed data.
Handles all the messy formatting (currency, percentages, different number formats) and streams efficiently for large files.
Works with OpenAI, DeepSeek, or any compatible API.