Hi @insiderperks,
I have kept an eye on this since I published the first version of the plugin actually 🙂 This adds an extra level of technical difficulty in the plugin (it does’t mean it’s difficult, but more chances of bugs, issues, or spending a lot of time fixing the implementation as OpenAI is still beta, models are changing and new ones being added, etc).
I also wanted to analyze how to implement it the best way possible so that it’s just an additional useless feature but really something that brings a lot to the users/visitors.
Also wanted to find a way to avoid using pinecone and the extra costs.
I have been playing with it for a little while, and with the new model (which is much faster), I am less reticent to add a few more quick requests to get a perfect response.
If you have experience with it, don’t hesitate to contact me privately, or on my Discord 🙂
I agree this is needed. I love this plugin and really learning how to use it to the max (please create a video channel around this) but embeddings are needed.
I’m currently having to use GPT AI Power to use embeddings with pinecone. Would be great to have all this within your plugin. Or maybe a side plugin, that depends on your plugin as its base, so your code can be separate.
Either way, keep up the good work, thank you.
Hi @sim2k 🙂 Embeddings are already implemented and I have been playing with them for a little while. I just want to make everything perfect; the cost calculation, how it plays with chatbots and the different models, etc. I don’t want to release this half-baked.
Today I was actually playing with different ways to allow the users to enhance their embeddings while performing vector search. It’s super cool, you’ll see 🙂
I would like to make a few updates first without revealing the embeddings to test other parts of the plugin, but by Monday, it will be available (in the Pro, however).
