I appreciate uploading research PDFs into GPT4 for summarization; it would be awesome if ChatGPT summaries of technical papers could be provided directly inside zotero, but I think that would also increase the cost exponentially. For government users, being able to store this inside AskSage.ai would also be useful, but the document management is the concern, since zotero seems to handle that a bit better while enabling clear citations.
When I go to https://platform.openai.com/api-keys Underneath the words "secret key" there is sequence below it that contains three periods in a row. I thought that this mean the key was hidden but I now think that it is just that the key contains three periods in a row. I still have not installed the addin.
Thank you for the present that I received by email.
Does anyone have insight into what it costs them to use ARIA or ZoteroGPT for their workflows. I understand that this might vary greatly with how it used, but I'm curious to hear the ranges.
In short, I'm looking at abstracting out the Zotero document loading + metadata part of the Paper-QA package and adding it as a standalone module to LlamaHub for integration into different RAG pipelines. Perhaps with the original author if they are interested. I already have some code I use in a RAG application that I wrote myself but intend to look at Paper-QA and refactor to include the best of both worlds.
I'd be interested in having testers and people who can write up issues. The loader will initially use primarily the Zotero API to make a local db for the purposes of getting the files and adding metadata so it can run on a server/cloud. If someone is interested in a version that works with a local Zotero db that is possible but not my focus now.
Experience with python is a plus.
If this works well, eventually we could do a Typescript version so it could be used as a Zotero add-on.
I don't know if @rabitai is reading this but not everybody wants to feed the big cats like openai. it would have been interesting if you could at least allow someone to include their own model name, API key and base url that are 100% compatible with openai. one such free example is groq API. i was able to paste the key and url but i can't change the model as it is hard code as GPT4 on screen right?
https://platform.openai.com/api-keys
When I go to
https://platform.openai.com/api-keys
Underneath the words "secret key" there is sequence below it that contains three periods in a row. I thought that this mean the key was hidden but I now think that it is just that the key contains three periods in a row.
I still have not installed the addin.
Thank you for the present that I received by email.
In short, I'm looking at abstracting out the Zotero document loading + metadata part of the Paper-QA package and adding it as a standalone module to LlamaHub for integration into different RAG pipelines. Perhaps with the original author if they are interested. I already have some code I use in a RAG application that I wrote myself but intend to look at Paper-QA and refactor to include the best of both worlds.
I'd be interested in having testers and people who can write up issues. The loader will initially use primarily the Zotero API to make a local db for the purposes of getting the files and adding metadata so it can run on a server/cloud. If someone is interested in a version that works with a local Zotero db that is possible but not my focus now.
Experience with python is a plus.
If this works well, eventually we could do a Typescript version so it could be used as a Zotero add-on.
https://github.com/lifan0127/ai-research-assistant
https://www.youtube.com/watch?v=gA3o2MlnPBQ