Zipstack / unstract

No-code LLM Platform to launch APIs and ETL Pipelines to structure unstructured documents
https://unstract.com
GNU Affero General Public License v3.0
348 stars 28 forks source link

privacy when self-hosting #427

Closed legaltextai closed 10 hours ago

legaltextai commented 2 days ago

I 've been playing with your hosted version. It's impressive. I 've tried all known parsers, including llamaparse and unstructured. But your parsing of bank/credict card statements is second to none. Great job! If I self - host, does it mean LLMWhisperer is run locally too? My biggest concern is the privacy, I 'd like to ensure everything stays local. I plan to use ollama models for both embeddings and LLM. Which of the open source models worked best in your experience? Thank you

shuveb commented 2 days ago

@legaltextai yes. LLMWhisperer can be privately hosted. Your data won't leave your network that way.

shuveb commented 1 day ago

@legaltextai missed your query about open source models. Llama3 is probably the best, at least for now.

legaltextai commented 10 hours ago

thank you very much @shuveb