sqlite-vss VS ann-benchmarks

I guess this is an answer to the GitHub issue I opened against SQLite-vss a couple of months ago?

https://github.com/asg017/sqlite-vss/issues/124

Perhaps sqlite-vss? It adds vector searches to sqlite.

https://github.com/asg017/sqlite-vss

Utilizing sqlite-vss to store and query vector embeddings managed by a local SQLite database, Semantify conducts fast, precise vector searches within these embeddings to find and recommend relevant content, ensuring readers are presented with articles that truly match their interests.

Whether you’re navigating through well-known options like SQLite, enriched with the sqlite-vss extension, or exploring other avenues like Chroma, an open-source vector database, selecting the right tool is paramount. This article compares these two choices, guiding you through the pros and cons of each, helping you choose the right tool for storing and querying vector embeddings for your project.

Here is a SQLite extension that uses Faiss under the hood.

https://github.com/asg017/sqlite-vss

Not associated with the project, just love SQLite and find it very useful.

Vector Databases: As your data grows, efficiently searching through millions of vectors can become a challenge. Specialized vector databases like FAISS, Annoy, or Elasticsearch's vector search capabilities can be explored to manage and search through large-scale vector data. Your sentence is grammatically correct. In addition, databases like SQLite and PostgreSQL have extensions, such as sqlite-vss and pgvector, that can be used to store and query vector embeddings, respectively.

I had a go at one of those a few months ago: https://datasette.io/plugins/datasette-faiss

Alex Garcia built a better one here as a SQLite Rust extension: https://github.com/asg017/sqlite-vss

Since the problem domain is that of anomaly detection from constructed request feature embeddings, I wonder if an ANN-search methodology using an embedded database (such as https://github.com/asg017/sqlite-vss or similar) was explored.

As I searched for "sqlite vector plugin" I didn't find any results, before a couple of weeks ago. Two weeks ago I found Alex' SQLite VSS plugin for SQLite. The library was an amazing piece of engineering from an "idea perspective". However, as I started playing around with it, I realised it was ipso facto like "Titanic". Beautiful and amazing, but destined to leak water and sink to the bottom of the ocean because of what we software engineers refers to as "memory leaks".

On top of my head, pgvector only supports 2 indexes, those are running in memory only. They don't support GPU indexing, nor Disk based indexing, they also don't have separation of query and insertions.

Also with different people I've talked to, they struggle with scale past 100K-1M vector.

You can also have a look yourself from a performance perspective: https://ann-benchmarks.com/

https://ann-benchmarks.com/ is a good resource covering those libraries and much more.

We utilized the ANN Benchmarks methodology, a standard for benchmarking vector databases. Our tests used the dbpedia dataset of 1,000,000 OpenAI embeddings (1536 dimensions) and inner product distance metric for both Pinecone and pgvector.

Data warehouses are columnar stores. They are very different from row-oriented databases - like Postgres, MySQL. Operations on columns - e.g., aggregations (mean of a column) are very efficient.

Most vector databases use one of a few different vector indexing libraries - FAISS, hnswlib, and scann (google only) are popular. The newer vector dbs, like weaviate, have introduced their own indexes, but i haven't seen any performance difference -

Reference: https://ann-benchmarks.com/

(Blog author here). Thanks for the question. In this case the index for both DiskANN and pgvector HNSW is small enough to fit in memory on the machine (8GB RAM), so there's no need to touch the SSD. We plan to test on a config where the index size is larger than memory (we couldn't this time due to limitations in ANN benchmarks [0], the tool we use).

To your question about RAM usage, we provide a graph of index size. When enabling PQ, our new index is 10x smaller than pgvector HNSW. We don't have numbers for HNSWPQ in FAISS yet.

[0]: https://github.com/erikbern/ann-benchmarks/

In terms of "All You Need" for Vector Search, ANN Benchmarks (https://ann-benchmarks.com/) is a good site to review when deciding what you need. As with anything complex, there often isn't a universal solution.

txtai (https://github.com/neuml/txtai) can build indexes with Faiss, Hnswlib and Annoy. All 3 libraries have been around at least 4 years and are mature. txtai also supports storing metadata in SQLite, DuckDB and the next release will support any JSON-capable database supported by SQLAlchemy (Postgres, MariaDB/MySQL, etc).

pg_vector doesn't perform well compared to other methods, at least according to ANN-Benchmarks (https://ann-benchmarks.com/).

txtai is more than just a vector database. It also has a built-in graph component for topic modeling that utilizes the vector index to autogenerate relationships. It can store metadata in SQLite/DuckDB with support for other databases coming. It has support for running LLM prompts right with the data, similar to a stored procedure, through workflows. And it has built-in support for vectorizing data into vectors.

For vector databases that simply store vectors, I agree that it's nothing more than just a different index type.

The reason https://ann-benchmarks.com is so good, is that we can see a plot of recall vs latency. I can see you have some latency numbers in the leaderboard at the bottom, but it's very difficult to make a decision.

As a practitioner that works with vector databases every day, just latency is meaningless to me, because I need to know if it's fast AND accurate, and what the tradeoff is! You can't have it both ways. So it would be helpful if you showed plots showing this tradeoff, similar to ann-benchmarks.