after long experiments and bug fixing, I finally launched this project. https://enumhq.com
The idea is to provide a simple way to load multiple data sources, extract text and be able to ask questions on them. I started with plain text (copy-paste), PDF files, and websites.
Free indexing up to 200 pages.
Useful for students, explorers, and businesses that would like to provide great support when they are OOO (not implemented yet )
I’m eager to hear your opinion. Do you find it useful? What to improve in the first place?
NB You can use your own OpenAI key and ask unlimited questions. All keys are stored in our database encrypted.
Thanks a lot for your question. Actually, I’m very interested in implementing such functionality. I’d love to look at the example PDF filled with equations and formulas printed in any format to see how I can recognize and index them. If you wouldn’t mind, can you please send me them at firstname.lastname@example.org ?
Can you please also give me an example question you’d ask on such documents?
Hi, I am still waiting for the file examples. Maybe you can point me to where I can find them and how to distinguish when a file contains LaTeX blocks?
Meanwhile, we implemented the rendering normal formulas and equations that can be part of the answer:
We recently finished the integration Enum with Crisp chatbot and really proud of the result (even though it’s just the beginning and not too many settings you can set up but anyway).
You can try the live chat with a chatbot here: https://enumhq.com
Would be happy to hear feedback! Thanks!
2 months later updates:
- we added custom prompts (so you can control your chatbot)
- added multi-language support (currently, our chatbot can speak 107 languages)
- links in the response
- our chatbot is now conversational i.e., it can memorize the context and keep the conversation
- added a lot of automation tools to make the interaction maximally smooth for users and agents
- fixed tons of bugs
Full changelog here: ENUM Changelog
This is what happened while we experimented with custom prompts: