The Locus team has been hard at work rebuilding the platform behind the Locus extension, and we are happy to announce the availability of Locus 1.3.0 with PDF Support!

While the little search bar you are used to seeing hasn’t changed much, you should definitely see an improvement in how quickly the pages are loaded by Locus so you can begin smart searching almost immediately.

But if you open a PDF document, you’ll see something a little different than the “This document cannot be loaded”. Take a look:

The long requested (by long, I mean couple of months) PDF support is now available as a public beta, so please excuse any rough edges. PDFs are tricky and there may be PDFs out there that our model has trouble handling - please let us know of any examples that you find to be the case. We have found it works best on documents and research paper-type pdfs, and if you do your search as a question.

You’ll notice that if you are on a pdf, the search box will request that the file be opened in the local locus PDF viewer built into the extension, this allows us to show the search highlights that you expect from Locus. The native chrome pdf viewer does not allow customization.

Also, thank you to all our PDF private beta users and their feedback!

This releases backend changes enable us to introduce new features much more quickly in the coming weeks and months (read the P.S. below for a sneak peek)…

You can also post your own requests to our backlog here: New feature requests

Thank you for using our smart search extension, and we hope it is making your lives just a little easier every day!

• The Locus Team

P.S. For the promised sneak peek:

New Sidebar!

History, Folders/Tags

These features will allow you to organize, track and manage your Locus searches.

LLM Integration

This feature will let you use GPT AI models directly in Locus!

Selectable AI Models

This feature will let you select from a curated set of AI models to use for your search use case.

Some models are better suited for questions, others for keywords, and possibly even multilingual searches.

(Disclaimer: the released features may look different as we go through our design and build process)

P.P.S Some nerdy details on some of the backend updates in this release:

We pretty much rebuilt the backend from the foundation up so we could support the new features described above. This includes a rust based engine to generate the index vectors, entirely new services that are faster and lighter, running on an entirely new and modern Kubernetes based platform, everything from the network configuration to the cluster scalability, with better instrumentation and metrics to help monitor and keep the platform performance at its peak as we go into the back to school season.