If you (like me) look at this and were like ???, a more compelling summary is perhaps:

> Trieve sets itself apart by being BSL-licensed self-hostable and built in public at github.com/devflowinc/trieve whereas Algolia is fully closed … At Trieve, we are hard at work building these features to reach full parity with Algolia (0)

Or this long winded post about using it to implement search (1)…

Long story short, like Algolia.

The rustiness of the project, I’m not sure I see the significance of; it has a bunch of very heavy non rust dependencies.

[0] - https://trieve.ai/trieve-vs-algolia [1] - https://trieve.ai/building-search-for-yc-company-directory

Yeah.... :exhale. Our docs and examples are pretty out of date and generally lacking. Should have more cool stuff coming out soon.

Just in case anyone else is thinking "WTF is RAG?" (I wish GitHub had explanation popups for its tags, like StackOverflow):

https://aws.amazon.com/what-is/retrieval-augmented-generatio...

yeah what's up with that? The entire github readme does not explain what RAG is, and it's not a common acronym at all

> and it's not a common acronym at all

It's everywhere in LLM circles these days, and has been for months. I would agree that they should define it - I think any professional project should be defining acronyms on first use on their sales pitch - but it's not exactly an esoteric acronym, either.

What's the advantage of this over just qdrant-client or maybe something like usearch? Or a crate with pgvector support (I assume the popular postgres crate includes it?)

Merchandising dashboard, multi-tenant, can start free on the cloud version, built-in RAG endpoints, chunk_group abstraction, relevancy weighting, date-recency biasing, maximal marginal relevance....

Probably a few more

It might be the first time that I see a phone number on a Github Repo, I wonder how many phone calls they will actually receive. It's probably the last thing I'd like to do when looking at an open source repository.

I hope a few! We have gotten some so far. In general, I just want people who are interested in using the software to be able to get a human on the phone immediately.

I find it difficult locate the core search features ... please correct me if I'm mistaken, but it appears that the search feature is a wrapper built on top of Qdrant ?

I think "wrapper" is a bit of a reductive term, but we do use Qdrant for indexing and search. There are a lot features we implement client-side though like the ingestion-microservice, relevancy weighting, hybrid with cross-encoder, date-recency biasing, etc.

I think calling it "Open-source" is misleading considering its license.

Agree, and it's explicitly declared as not open source in its own terms, so I presume a mistake in the posting here. Title here should be changed I think.

For anybody interested a summary of "Business Source" terms here:

https://fossa.com/blog/business-source-license-requirements-...

Its open-source as in OSINT

This is almost certainly best described as "source-available"

Eh, we could get into how to best explain the Business Source License. Not saying source available is wrong, but I'd say the term is incomplete.

Perhaps "source-curious" should suffice.

Agree, we call it "source-available" everywhere. If we missed a spot, I will change it

The license isn't open source compatible though.

It is open-source as OSINT.

OSINT is not open-source though

One of the first heavy uses of Rust in AI that I have seen

There are much better known examples, such as https://qdrant.tech/ and https://github.com/lancedb/lancedb

We use Qdrant under the hood!

What about candle?

Exactly

Oh thanks for all these projects! I guess I haven't been paying attention to Rust x AI

What is your relationship with this repository and the group behind it?

I am one of the founders, not sure who the OP is

Random person on the Interwebs ;) Just found your project and thought it was cool