Show HN: Tezcat – local-first AI recall in Obsidian via a remembrance agent
github.comTezcat is a POC implementation of a remembrance agent (https://www.bradleyrhodes.com/Papers/remembrance.html) in Obsidian. It uses text embeddings and vector similarity search (or hybrid search) to do recall of fragments of notes that you've written in the past based on what you're writing right now.
I don't like AI chat interfaces in knowledge management tools and I have been looking to implement something that would integrate better with the flow of actually writing. This operates best with Ollama on your machine, for a local-first experience, but can work with OpenAI embeddings.
Try it out and let me know what you think! This is sick! Got it up and running, I don't have a huge vault but I love the idea of having relevant things surfaced as I write. Is there any way to just get the links of relevant notes / do additional filtering? I'd love to have the ability to fine tune how things get surfaced You can insert a link to the surfaced content using the button on the search results - unfortunately it won't directly link to the block the content comes from. If the search result is a full "Note", clicking the result object will copy a link to it to your clipboard (so you can put it wherever), while if it's a "Block" (section in Obsidian parlance), it copies the content. Reminded me of https://news.ycombinator.com/item?id=39107426 (2024) https://github.com/nichwch/nomad-hypertext https://nomadhypertext.nicholaschen.io/ Oh neat, I haven't seen this before, though I imagine that the source of inspiration is similar. I implemented mine on Obsidian as it's a little easier than starting with a new tool wholesale. Have you used Nomad before? hi