I love this feature. I've long used json_extract to create dynamic columns with txtai sql: https://neuml.github.io/txtai/embeddings/query/#dynamic-colu...

You can do the same with DuckDB and Postgres too.

I've used SQLite as the content storage engine for years with TxtAI. It works great. Also plenty of good add-ons for it such as sqlite-vec for storing vectors. It can take you pretty far and maybe it's all you need in many circumstances.

Looking forward to trying them out. Great to see they are Apache 2.0...always good to have easy-to-understand licensing.

Companies that consider an open source project a critical part of their infrastructure should sponsor or compensate those projects.

Also when someone finds a bug, the maintainers are under no obligation to fix it or fix it with any timeline or even debug what's going on. If someone wants an immediate response they should provide compensation.

A common misconception is that OSS developers do everything for free. They do what THEY want for free. If YOU want to change their priorities, companies need to compensate for that.

Then someone finds out you are rewarding outsiders for something they are doing for free anyway. Such cutting into company profit is inexcusable. You are supposed to ride even your own employees raw to maximize profits, not to splurge money to some weirdos just like that!

Are multiple LLM queries faster than vector search? Even with the example "dog OR canine" that leads to two LLM inference calls vs one. LLM inference is also more expensive than vector search.

In general RAG != Vector Search though. If a SQL query, grep, full text search or other does the job then by all means. But for relevance-based search, vector search shines.

I agree. I did this myself with TxtAI. It can store vectors, data, graphs and keyword indexes all to Postgres. https://medium.com/neuml/postgres-is-all-you-need-for-vector...

Fun project. I'm sure it will get a lot of interest here.

For those into vector storage in general, one thing that has interested me lately is the idea of storing vectors as GGUF files and bring the familiar llama.cpp style quants to it (i.e. Q4_K, MXFP4 etc). An example of this is below.

https://gist.github.com/davidmezzetti/ca31dff155d2450ea1b516...

Glad to see all the interest in the local RAG space, it's been something I've been pushing for a while.

I just put this example together today: https://gist.github.com/davidmezzetti/d2854ed82f2d0665ec7efd...

Sounds about right for someone who is building on their own.

The perception is that it's sloppier than what humans write. I wouldn't waste too much time worrying about that if the code is solving your problem. Sounds like you need to focus on getting a MVP to show to potential customers and if LLMs help you get there faster, so be it.

Thank you! That is a good take on things. If it works, it works. If it fails, then a customer complains and we fix it + make sure other code doesn't suffer the same error.

Good luck!