Merge pull request #2053 from containers/rag-separate-container

run RAG operations in a separate container

Author: rhatdan

.tekton/asahi-rag/asahi-rag-pull-request.yaml

@@ -28,12 +28,10 @@ spec:
     value: 5d
   - name: dockerfile
     value: container-images/common/Containerfile.rag
-  - name: parent-image
-    value: quay.io/redhat-user-workloads/ramalama-tenant/cuda:on-pr-{{revision}}
   - name: build-args
     value:
-    - PARENT=quay.io/redhat-user-workloads/ramalama-tenant/cuda:on-pr-{{revision}}
-    - GPU=cuda
+    - PARENT=docker.io/nvidia/cuda:12.8.1-devel-ubi9
+    - TORCH_BACKEND=cu128
   pipelineRef:
     name: pull-request-pipeline
   timeouts:

.tekton/asahi-rag/asahi-rag-push.yaml

@@ -25,12 +25,10 @@ spec:
     value: quay.io/redhat-user-workloads/ramalama-tenant/cuda-rag:{{revision}}
   - name: dockerfile
     value: container-images/common/Containerfile.rag
-  - name: parent-image
-    value: quay.io/redhat-user-workloads/ramalama-tenant/cuda:{{revision}}
   - name: build-args
     value:
-    - PARENT=quay.io/redhat-user-workloads/ramalama-tenant/cuda:{{revision}}
-    - GPU=cuda
+    - PARENT=docker.io/nvidia/cuda:12.8.1-devel-ubi9
+    - TORCH_BACKEND=cu128
   pipelineRef:
     name: push-pipeline
   timeouts:

.tekton/cann-rag/cann-rag-pull-request.yaml

@@ -70,9 +70,12 @@ spec:
     - name: envs
       value:
       - RAMALAMA_DEFAULT_IMAGE="$(tasks.init.results.ramalama-image)"
+      - RAMALAMA_DEFAULT_RAG_IMAGE="$(tasks.init.results.ramalama-rag-image)"
       - RAMALAMA_STACK_IMAGE=$(tasks.init.results.stack-image)
       - >-
         RAMALAMA_IMAGES='{"CUDA_VISIBLE_DEVICES": "$(tasks.init.results.cuda-image)"}'
+      - >-
+        RAMALAMA_RAG_IMAGES='{"CUDA_VISIBLE_DEVICES": "$(tasks.init.results.cuda-rag-image)"}'
     taskRef:
       resolver: git
       params:

.tekton/cann-rag/cann-rag-push.yaml

@@ -23,10 +23,14 @@ spec:
     description: URI of the bats image included in the snapshot
   - name: ramalama-image
     description: URI of the ramalama image included in the snapshot
+  - name: ramalama-rag-image
+    description: URI of the ramalama-rag image included in the snapshot
   - name: stack-image
     description: URI of the llama-stack image included in the snapshot
   - name: cuda-image
     description: URI of the cuda image included in the snapshot
+  - name: cuda-rag-image
+    description: URI of the cuda-rag image included in the snapshot
   - name: TEST_OUTPUT
     description: Test result in json format
   steps:
@@ -59,10 +63,14 @@ spec:
       value: $(results.bats-image.path)
     - name: RESULTS_RAMALAMA_IMAGE_PATH
       value: $(results.ramalama-image.path)
+    - name: RESULTS_RAMALAMA_RAG_IMAGE_PATH
+      value: $(results.ramalama-rag-image.path)
     - name: RESULTS_STACK_IMAGE_PATH
       value: $(results.stack-image.path)
     - name: RESULTS_CUDA_IMAGE_PATH
       value: $(results.cuda-image.path)
+    - name: RESULTS_CUDA_RAG_IMAGE_PATH
+      value: $(results.cuda-rag-image.path)
     - name: RESULTS_TEST_OUTPUT_PATH
       value: $(results.TEST_OUTPUT.path)
     script: |
@@ -87,9 +95,13 @@ spec:
       echo
       component_image ramalama | tee "$RESULTS_RAMALAMA_IMAGE_PATH"
       echo
+      component_image ramalama-rag | tee "$RESULTS_RAMALAMA_RAG_IMAGE_PATH"
+      echo
       component_image llama-stack | tee "$RESULTS_STACK_IMAGE_PATH"
       echo
       component_image cuda | tee "$RESULTS_CUDA_IMAGE_PATH"
       echo
+      component_image cuda-rag | tee "$RESULTS_CUDA_RAG_IMAGE_PATH"
+      echo
       jq -jnc '{result: "SUCCESS", timestamp: now | todateiso8601, failures: 0, successes: 1, warnings: 0}' | tee "$RESULTS_TEST_OUTPUT_PATH"
       echo

.tekton/cuda-rag/cuda-rag-pull-request.yaml

@@ -31,12 +31,9 @@ spec:
     - linux-d160-c8xlarge/amd64
   - name: dockerfile
     value: container-images/common/Containerfile.rag
-  - name: parent-image
-    value: quay.io/redhat-user-workloads/ramalama-tenant/intel-gpu:on-pr-{{revision}}
   - name: build-args
     value:
-    - PARENT=quay.io/redhat-user-workloads/ramalama-tenant/intel-gpu:on-pr-{{revision}}
-    - GPU=cpu
+    - TORCH_BACKEND=xpu
   pipelineRef:
     name: pull-request-pipeline
   timeouts:

.tekton/cuda-rag/cuda-rag-push.yaml

@@ -28,12 +28,9 @@ spec:
     - linux-d160-c8xlarge/amd64
   - name: dockerfile
     value: container-images/common/Containerfile.rag
-  - name: parent-image
-    value: quay.io/redhat-user-workloads/ramalama-tenant/intel-gpu:{{revision}}
   - name: build-args
     value:
-    - PARENT=quay.io/redhat-user-workloads/ramalama-tenant/intel-gpu:{{revision}}
-    - GPU=cpu
+    - TORCH_BACKEND=xpu
   pipelineRef:
     name: push-pipeline
   timeouts: