Docker image is very large > 18GB?

[ianpurton]

Would it be possible to reduce the size of the image?

This gives us several issues

• We can't use unstructured-api in a Gitgub actions as the virtual machine fills up.
• The attack surface for this image is very large.
• Deployments are slow

The main issue seems to be that the image includes the whole of libraoffice.

Here are the results from dive https://github.com/wagoodman/dive

[awalker4]

Hi there! Apologies for the delay - this repo isn't super active, seems this may have slipped under the radar. Normally I'd expect an unstructured-api image to be 8GB tops. I tried analyzing the image from your screenshot and it comes in at 6.9. It does look like my base hash is different, so I'm wondering if we had uploaded a larger image by accident and replaced it. At one point we were preloading some of the ML models, but this shouldn't be the case in our public images. Can you delete the local image and try pulling again?

dive downloads.unstructured.io/unstructured-io/unstructured-api:4ffd8bc

[lambda-science]

Hi there! Apologies for the delay - this repo isn't super active, seems this may have slipped under the radar. Normally I'd expect an unstructured-api image to be 8GB tops. I tried analyzing the image from your screenshot and it comes in at 6.9. It does look like my base hash is different, so I'm wondering if we had uploaded a larger image by accident and replaced it. At one point we were preloading some of the ML models, but this shouldn't be the case in our public images. Can you delete the local image and try pulling again?

dive downloads.unstructured.io/unstructured-io/unstructured-api:4ffd8bc

Same issue, this image is 17.89go