CircleCI-Public / cimg-postgres

MIT License
8 stars 25 forks source link

Feature Request: Add `pgvector` extension #120

Open pencil opened 9 months ago

pencil commented 9 months ago

Describe the Feature Request Add the pgvector extension to CircleCI's cimg-postgres Docker image to support efficient storage and search of vector data from machine learning models, aligning with industry trends towards ML integration. E.g. Amazon RDS installs the extension in their Postgres datbases.

Is your feature request related to a particular problem?

Yes, the absence of pgvector necessitates custom Docker images, slowing down CI/CD pipelines and introducing potential inconsistencies.

How will this feature request benefit CircleCI jobs using this image?

Incorporating pgvector will streamline workflows for ML projects, reduce setup times, and offer a consistent environment, making CircleCI more attractive to developers working with vector embeddings.

Describe the solution you would like to see

The pgvector extension installed by default in the cimg-postgres Docker image, compatible with its PostgreSQL versions and easily enabled via CREATE EXTENSION vector;

Describe alternatives you have considered

Manually installing pgvector or creating a custom Docker image. The former isn't possible and the latter is cumbersome and increases complexity.

kendagriff commented 8 months ago

@pencil: I submitted a PR for this very feature, but the maintainers have been silent.

jbauerrfid commented 1 week ago

Adding pgvector for AI applications is a great idea. But pgvector is a huge extension, and it would be preferable to publish this under a separate docker image tag, like "cimg/postgres:16.5-pgvector".

Most CI users use the postgres image to run baseline integration tests, and don't need pgvector for this.

rmonteiro89 commented 11 hours ago

Just in case somebody else gets stuck trying to use an image of Postgres with pgvector. I used the pgvector postgres image instead of the circleci image.

- image: pgvector/pgvector:pg14