The user wants a chat window trained on company-specific data to answer queries. They list various data types like finance, email, calendar, PDFs, customer support, social media, marketing lists, spreadsheets, and chat logs as potential sources for this system.
# TL:DR Rough idea is as follows: - a chat window trained with company specific data. I created this issue to track all he usecases I thought of that you could feed such a system # What sources - Finance data (So you feed invoices sent and received) so you can query when did we last order from shop X - Email data - Calendar data - PDF document with business related logic - Customer support data - Social media posts - Marketing email lists - Spreadsheet / word docs - Chat/Slack data of public channels - code - contracts - employee records - candidates recruitment - Customer reviews - Blog (by URL) - website (by URL) - meeting transcripts - analytics databases - Laws - On daycare: The instructions for kids on daycare. --> So parents write the preference for the kids ^ All of these sources should be normalised and metadata collected (how many emails were sent, how many candidates etc) # Proof of concept code - https://github.com/techleadhd/chatgpt-ret