test: Add VLLM provider support to integration tests (#2757)

- Add setup-vllm GitHub action to start VLLM container
- Extend integration test matrix to support both ollama and vllm
providers
- Make test setup conditional based on provider type
- Add provider-specific environment variables and configurations
- vllm tests setup to run weekly or can be triggered manually (only
ollama on PR)

TODO:
investigate failing tests for vllm provider (safety and post_training)
Also need a proper fix for #2713 (tmp fix for this in the first commit
in this PR)
Closes: #1648

---------

Signed-off-by: Derek Higgins <derekh@redhat.com>

.github/actions/setup-vllm/action.yml

@@ -0,0 +1,27 @@
+name: Setup VLLM
+description: Start VLLM
+runs:
+  using: "composite"
+  steps:
+    - name: Start VLLM
+      shell: bash
+      run: |
+        # Start vllm container
+        docker run -d \
+          --name vllm \
+          -p 8000:8000 \
+          --privileged=true \
+          quay.io/higginsd/vllm-cpu:65393ee064 \
+          --host 0.0.0.0 \
+          --port 8000 \
+          --enable-auto-tool-choice \
+          --tool-call-parser llama3_json \
+          --model /root/.cache/Llama-3.2-1B-Instruct \
+          --served-model-name meta-llama/Llama-3.2-1B-Instruct
+
+          # Wait for vllm to be ready
+          echo "Waiting for vllm to be ready..."
+          timeout 900 bash -c 'until curl -f http://localhost:8000/health; do
+            echo "Waiting for vllm..."
+            sleep 5
+          done'

.github/workflows/integration-tests.yml

@@ -14,13 +14,19 @@ on:
       - '.github/workflows/integration-tests.yml' # This workflow
       - '.github/actions/setup-ollama/action.yml'
   schedule:
-    - cron: '0 0 * * *'  # Daily at 12 AM UTC
+    # If changing the cron schedule, update the provider in the test-matrix job
+    - cron: '0 0 * * *'  # (test latest client) Daily at 12 AM UTC
+    - cron: '1 0 * * 0'  # (test vllm) Weekly on Sunday at 1 AM UTC
   workflow_dispatch:
     inputs:
       test-all-client-versions:
         description: 'Test against both the latest and published versions'
         type: boolean
         default: false
+      test-provider:
+        description: 'Test against a specific provider'
+        type: string
+        default: 'ollama'
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.ref }}
@@ -53,8 +59,17 @@ jobs:
       matrix:
         test-type: ${{ fromJson(needs.discover-tests.outputs.test-type) }}
         client-type: [library, server]
+        # Use vllm on weekly schedule, otherwise use test-provider input (defaults to ollama)
+        provider: ${{ (github.event.schedule == '1 0 * * 0') && fromJSON('["vllm"]') || fromJSON(format('["{0}"]', github.event.inputs.test-provider || 'ollama')) }}
         python-version: ["3.12", "3.13"]
-        client-version: ${{ (github.event_name == 'schedule' || github.event.inputs.test-all-client-versions == 'true') && fromJSON('["published", "latest"]') || fromJSON('["latest"]') }}
+        client-version: ${{ (github.event.schedule == '0 0 * * 0' || github.event.inputs.test-all-client-versions == 'true') && fromJSON('["published", "latest"]') || fromJSON('["latest"]') }}
+        exclude: # TODO: look into why these tests are failing and fix them
+          - provider: vllm
+            test-type: safety
+          - provider: vllm
+            test-type: post_training
+          - provider: vllm
+            test-type: tool_runtime
 
     steps:
       - name: Checkout repository
@@ -67,8 +82,13 @@ jobs:
           client-version: ${{ matrix.client-version }}
 
       - name: Setup ollama
+        if: ${{ matrix.provider == 'ollama' }}
         uses: ./.github/actions/setup-ollama
 
+      - name: Setup vllm
+        if: ${{ matrix.provider == 'vllm' }}
+        uses: ./.github/actions/setup-vllm
+
       - name: Build Llama Stack
         run: |
           uv run llama stack build --template ci-tests --image-type venv
@@ -81,10 +101,6 @@ jobs:
 
       - name: Run Integration Tests
         env:
-          OLLAMA_INFERENCE_MODEL: "llama3.2:3b-instruct-fp16" # for server tests
-          ENABLE_OLLAMA: "ollama" # for server tests
-          OLLAMA_URL: "http://0.0.0.0:11434"
-          SAFETY_MODEL: "llama-guard3:1b"
           LLAMA_STACK_CLIENT_TIMEOUT: "300" # Increased timeout for eval operations
         # Use 'shell' to get pipefail behavior
         # https://docs.github.com/en/actions/reference/workflow-syntax-for-github-actions#exit-codes-and-error-action-preference
@@ -96,12 +112,31 @@ jobs:
           else
             stack_config="server:ci-tests"
           fi
+
+          EXCLUDE_TESTS="builtin_tool or safety_with_image or code_interpreter or test_rag"
+          if [ "${{ matrix.provider }}" == "ollama" ]; then
+            export ENABLE_OLLAMA="ollama"
+            export OLLAMA_URL="http://0.0.0.0:11434"
+            export OLLAMA_INFERENCE_MODEL="llama3.2:3b-instruct-fp16"
+            export TEXT_MODEL=ollama/$OLLAMA_INFERENCE_MODEL
+            export SAFETY_MODEL="llama-guard3:1b"
+            EXTRA_PARAMS="--safety-shield=$SAFETY_MODEL"
+          else
+            export ENABLE_VLLM="vllm"
+            export VLLM_URL="http://localhost:8000/v1"
+            export VLLM_INFERENCE_MODEL="meta-llama/Llama-3.2-1B-Instruct"
+            export TEXT_MODEL=vllm/$VLLM_INFERENCE_MODEL
+            # TODO: remove the not(test_inference_store_tool_calls) once we can get the tool called consistently
+            EXTRA_PARAMS=
+            EXCLUDE_TESTS="${EXCLUDE_TESTS} or test_inference_store_tool_calls"
+          fi
+
+
           uv run pytest -s -v tests/integration/${{ matrix.test-type }} --stack-config=${stack_config} \
-            -k "not(builtin_tool or safety_with_image or code_interpreter or test_rag)" \
-            --text-model="ollama/llama3.2:3b-instruct-fp16" \
+            -k "not( ${EXCLUDE_TESTS} )" \
+            --text-model=$TEXT_MODEL \
             --embedding-model=sentence-transformers/all-MiniLM-L6-v2 \
-            --safety-shield=$SAFETY_MODEL \
-            --color=yes \
+            --color=yes ${EXTRA_PARAMS} \
             --capture=tee-sys | tee pytest-${{ matrix.test-type }}.log
 
       - name: Check Storage and Memory Available After Tests
@@ -110,16 +145,17 @@ jobs:
           free -h
           df -h
 
-      - name: Write ollama logs to file
+      - name: Write inference logs to file
         if: ${{ always() }}
         run: |
-          sudo docker logs ollama > ollama.log
+          sudo docker logs ollama > ollama.log || true
+          sudo docker logs vllm > vllm.log || true
 
       - name: Upload all logs to artifacts
         if: ${{ always() }}
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
         with:
-          name: logs-${{ github.run_id }}-${{ github.run_attempt }}-${{ matrix.client-type }}-${{ matrix.test-type }}-${{ matrix.python-version }}-${{ matrix.client-version }}
+          name: logs-${{ github.run_id }}-${{ github.run_attempt }}-${{ matrix.provider }}-${{ matrix.client-type }}-${{ matrix.test-type }}-${{ matrix.python-version }}-${{ matrix.client-version }}
           path: |
             *.log
           retention-days: 1

tests/integration/inference/test_openai_completion.py

@@ -179,9 +179,7 @@ def test_openai_completion_prompt_logprobs(llama_stack_client, client_with_model
         model=text_model_id,
         prompt=prompt,
         stream=False,
-        extra_body={
-            "prompt_logprobs": prompt_logprobs,
-        },
+        prompt_logprobs=prompt_logprobs,
     )
     assert len(response.choices) > 0
     choice = response.choices[0]
@@ -196,9 +194,7 @@ def test_openai_completion_guided_choice(llama_stack_client, client_with_models,
         model=text_model_id,
         prompt=prompt,
         stream=False,
-        extra_body={
-            "guided_choice": ["joy", "sadness"],
-        },
+        guided_choice=["joy", "sadness"],
     )
     assert len(response.choices) > 0
     choice = response.choices[0]