AI DIAL Generic RAG backend 0.3.2 with group-0 permissions for OpenShift (arbitrary UID)
32
AI DIAL Generic RAG backend rebuilt so it starts under OpenShift's default restricted-v2 SCC, which runs containers with an arbitrary UID and GID 0.
Unofficial build. This repository is not affiliated with or endorsed by EPAM. The application code and the bundled embedding model come unchanged from the upstream image
epam/ai-dial-generic-rag-backend.
On OpenShift the upstream image fails at startup with:
FileNotFoundError: Path /home/appuser/bge-small-en not found
The model directory is owned by appuser and is not reachable by the random UID that OpenShift assigns. This image only changes file permissions:
/home/appuser is made group-0 owned, with group permissions equal to the owner's, and readable/traversable by othersMPLCONFIGDIR=/tmp/matplotlib, so matplotlib does not try to write to /.configNothing else is modified, and the original USER of the base image is kept.
| Tag | Base image |
|---|---|
0.3.2-openshift | epam/ai-dial-generic-rag-backend:0.3.2 |
Settings are read from environment variables. Main ones (see the upstream README for the full list):
| Variable | Required | Description |
|---|---|---|
DIAL_URL | yes | URL of DIAL Core |
DIAL_API_KEY | no | API key for background jobs such as indexing |
DB_HOST | yes | PostgreSQL host (with the pgvector extension) |
DB_PORT | no | PostgreSQL port, default 5432 |
DB_NAME | yes | Database name |
DB_USERNAME | yes | Database user |
DB_PASSWORD | no | Database password |
DB_MSI_ENABLED | no | Use MSI authentication, default false |
IN_MEMORY_CACHE_ENABLED | no | In-memory file cache, default true |
IN_MEMORY_CACHE_CAPACITY | no | Cache size, for example 128MiB |
The container listens on port 5000, and /health is the health check endpoint.
docker run --rm -p 5000:5000 \
-e DIAL_URL=http://dial-core:8080 \
-e DB_HOST=postgres -e DB_NAME=statgpt \
-e DB_USERNAME=statgpt -e DB_PASSWORD=<password> \
fireport/ai-dial-generic-rag-backend:0.3.2-openshift
generic-rag:
image:
registry: docker.io
repository: fireport/ai-dial-generic-rag-backend
tag: 0.3.2-openshift
docker run --rm --user 1000970000:0 --entrypoint sh \
fireport/ai-dial-generic-rag-backend:0.3.2-openshift \
-c 'ls /home/appuser/bge-small-en'
docker build --platform linux/amd64 \
--build-arg BASE_TAG=0.3.2 --build-arg APP_USER=appuser \
-t fireport/ai-dial-generic-rag-backend:0.3.2-openshift .
Content type
Image
Digest
sha256:0559faa69…
Size
1.1 GB
Last updated
3 days ago
docker pull fireport/ai-dial-generic-rag-backend:0.3.2-openshift