The docker images are confusing #6479
-
|
What is the difference between There's also Any explanation would be very helpful in me understanding the inconsistencies! Thank you :) EDIT: I realized that the rag-api is in a different repo, which makes sense. Sorry for the confusion there. Along the same lines, what's the difference between the |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
|
for the rag api, lite excludes nvidia dependencies needed for local model embeddings, whereas non-lite doesn't. librechat-api vs librechat are just 2 different strategies for building the container image, the former results in a smaller footprint, it's used in the remote deployment guide, so it's effectively a "lite" version but is more "complex" and requires a different compose file (the deploy-compose.yaml) |
Beta Was this translation helpful? Give feedback.
-
|
Hey, having a similar confusion with using the prebuilt librechat docker image vs building it ourselves. When I deploy the prebuilt one it's ~400mb, but building from source with the Dockerfile only comes to ~30mb and doesn't run properly. Am I missing something/is there documentation on how to build from source? |
Beta Was this translation helpful? Give feedback.
-
|
@djriversq for building the image yourself, the source of truth would be the exact github workflows being used, found in ./.github named "build dev images" and the main Dockerfile. Run these through an LLM and query on your exact needs |
Beta Was this translation helpful? Give feedback.
for the rag api, lite excludes nvidia dependencies needed for local model embeddings, whereas non-lite doesn't.
librechat-api vs librechat are just 2 different strategies for building the container image, the former results in a smaller footprint, it's used in the remote deployment guide, so it's effectively a "lite" version but is more "complex" and requires a different compose file (the deploy-compose.yaml)