Dotient

I love your website! Is the expected usage to train a signal before every unique search? If i had a bunch of group photos and I'm looking to search by name, I'd have to create a signal for each person by clicking a bunch of pictures they're in and not in?

local + semantic search is a nice combo, feels like most semantic search tools assume you're fine sending everything to the cloud. how's the search quality holding up running fully local vs what you'd get from a cloud-based embedding model.

As someone who's spent way too much time searching for "that one PDF with the blue chart" or "that screenshot I know I took last month," this feels incredibly relatable. Love the local-first approach as well, privacy shouldn't have to be a trade-off for smarter search.Definitely giving Dotient a spin. Congrats on the launch, and excited to see where you take it next! 👏

Hey, how are we handling privacy in this? Like, what data actually leaves my device or workspace when I run Dotient on something sensitive?

Pinning one well-tuned model is a totally fair call for v1. The one cheap thing I'd still bake in now: stamp every vector row in SQLite with a model id or hash. It costs almost nothing today, and it's what lets you re-embed lazily if you ever do swap models, where a file gets re-embedded the first time it's queried after the upgrade so the cost spreads across normal use instead of one multi-hour background rescan. Retrofitting that stamp after the fact is the part that hurts.

Local embeddings are the right call, but the part that bit me building this kind of thing is model versioning. The day you ship a better embedding model, every vector on disk is from the old one, so you either re-embed the whole drive, which is hours of background CPU, or run mixed old and new vectors where the query model and stored model disagree and recall quietly drops. How are you handling an embedding-model upgrade across an already-indexed machine, re-embed in place or version the index and migrate lazily?

Local-first semantic search is the right call, the data leaving your machine is what kills these for real work. The thing I'd want as a user though: how do I trust it found everything? Keyword search fails loudly (zero results), but semantic search fails quietly, it returns something plausible and you never know what it missed. Do you surface a confidence or a "why this matched" so I can tell a real hit from a near-miss? That's what decides whether I rely on it or still fall back to ctrl-F.

How did this wonderful idea come about?Did it stem from someone's problem?

Local-first + offline visual search is the part I actually care about - most semantic file search tools quietly ship your content to an API, so doing the embeddings on-device is the real differentiator here. Two implementation things: is the index updated incrementally via a file watcher as files change, or is it a manual re-scan, and where does the embedding DB actually live on disk? And for the deep PDF search, are you running OCR on scanned/image-only PDFs, or only indexing PDFs that already have a text layer?

I built this because I was genuinely sick of Windows File Explorer. Finding a file you half-remember is a nightmare, and don't even get me started on hunting through PDFs. Dotient lets you find files based on what they look like, not what you happened to name them two years ago, runs completely offline, and gives you a real bird's eye view of everything on your machine. Hope you find it as useful as I do.