- This topic has 1 reply, 1 voice, and was last updated 7 months, 2 weeks ago by
.
-
Posts
-
If we treat the index of anytxt as a personal KB (knowledge database), will it be possible to make it readable by AI services like chatGPT, so that when I talk to my chatbot, its knowledge is based on my document index?
I’m fascinated by this approach because in many cases, I want the chatbot to learn from my Word/PPT/PDF instead of its own (often too generalized) training data, to produce output answer more tailored to my company’s shared knowledge.
Similar projects/products:
1. PDFgear: it can read your pdf and call chatGPT 3.5 to generate answer. Page references are also generated. So if it’s a long PDF you don’t have time to read, PDFgear can read it for you and find/generate answers. But it’s not stable to say the least: 50% of the time, you get a network error or something.
2. A project called PrivateGPT on github: it uses local LLM to crunch your local PDFs and build its database, then answer from that point. But since it requires locally running an LLM, the hardware requirement is high.
