This plugin does something similar to smart-connections, but imho with a number of big benefits.
Automatically start from the context of the current note.
Can correct previous assistant/bot responses - sometimes, a bot get things wrong, and I need it to get it right in order to move forward. In the ChatGPT UI / smart connections, I have to correct it by moving forward in the chat. Here, I can correct the assistant history and just move forward without additional correction messages.
Better UI - obsidian native instead of a side chat thing.
Stores chat's right next to my other notes instead of as a separate history
All in all, it's the perfect writers block defeater or brainstorm pair partner - I can get my ideas out there by chatting with a bot and then trim down my conversation to the final output without a bunch of copy / paste.
It's really great - the only thing it's missing is being able to incorporate context from other notes in my vault!
If you'd be interested, I'd be interested in adding embedding search and context support.
My sense for how it could work:
When you initiate the chat, the plugin does cosine similarity on embeddings similar to smart-connections
For any page matches worth sending along as context, it adds an obsidian link to the match on the current page at the end of the user prompt section.
When it comes time to send the data to the GPT API, it sends text of pages directly linked alongside.
Does sound like that right implementation you'd want for this functionality?
This plugin does something similar to smart-connections, but imho with a number of big benefits.
All in all, it's the perfect writers block defeater or brainstorm pair partner - I can get my ideas out there by chatting with a bot and then trim down my conversation to the final output without a bunch of copy / paste.
It's really great - the only thing it's missing is being able to incorporate context from other notes in my vault!
If you'd be interested, I'd be interested in adding embedding search and context support.
My sense for how it could work:
When you initiate the chat, the plugin does cosine similarity on embeddings similar to smart-connections
For any page matches worth sending along as context, it adds an obsidian link to the match on the current page at the end of the user prompt section.
When it comes time to send the data to the GPT API, it sends text of pages directly linked alongside.
Does sound like that right implementation you'd want for this functionality?