Skip to content

Add optional date templating in model system prompt #221

Add optional date templating in model system prompt

Add optional date templating in model system prompt #221

Workflow file for this run

# Based on https://github.com/azimuth-cloud/azimuth/blob/master/.github/workflows/test-pr.yaml
name: Test pull request
on:
pull_request:
types:
- opened
- synchronize
- ready_for_review
- reopened
branches:
- main
# Use the head ref for workflow concurrency, with cancellation
# This should mean that any previous workflows for a PR get cancelled when a new commit is pushed
concurrency:
group: ${{ github.workflow }}-${{ github.head_ref }}
cancel-in-progress: true
jobs:
# This job exists so that PRs from outside the main repo are rejected
fail_on_remote:
runs-on: ubuntu-latest
steps:
- name: PR must be from a branch in the stackhpc/azimuth-llm repo
run: exit ${{ github.event.pull_request.head.repo.full_name == 'stackhpc/azimuth-llm' && '0' || '1' }}
publish_images:
needs: [fail_on_remote]
uses: ./.github/workflows/build-push-images.yml
with:
ref: ${{ github.sha }}
secrets: inherit
publish_charts:
needs: [fail_on_remote]
uses: ./.github/workflows/build-push-charts.yml
with:
ref: ${{ github.sha }}
secrets: inherit
permissions:
# Explicit permissions are required for
# PRs created by dependabot user
contents: write
chart_validation:
needs: [publish_images]
runs-on: ubuntu-latest
env:
CLUSTER_NAME: chart-testing
RELEASE_NAME: ci-test
steps:
- name: Checkout the repo
uses: actions/checkout@v5
with:
ref: ${{ github.sha }}
- name: Reconfigure docker storage
# GitHub runners have an extra 75G disk at /mnt
# so we use this to avoid hitting capacity issues
# with kind cluster + vLLM container images
run: |
sudo bash -c 'cat > /etc/docker/daemon.json' << EOF
{
"data-root": "/mnt/docker-data"
}
EOF
sudo systemctl restart docker
- name: Run Gradio app tests
run: |
echo "Installing ollama..."
curl -fsSL https://ollama.com/install.sh | sh
# Give ollama systemd unit time to start
sleep 5
MODEL=smollm2:135m
echo "Pulling model $MODEL..."
ollama pull $MODEL
echo "Running Gradio test script"
./test-images.sh $(git rev-parse --short ${{ github.sha }})
working-directory: web-apps
- name: Install Helm
uses: azure/setup-helm@v4
- name: Run base chart templating with default values
run: helm template ci-test charts/azimuth-llm
- name: Set up chart testing
uses: helm/chart-testing-action@v2
- name: Run chart linting
run: ct lint --config ct.yaml
- name: Create Kind Cluster
uses: helm/kind-action@v1
with:
cluster_name: ${{ env.CLUSTER_NAME }}
# NOTE(scott): Since the local Chart.yaml uses "appVersion: latest" and this
# only gets overwritten to the correct commit SHA during Helm chart build,
# we need to pull these published images and load them into the kind cluster
# with the tag correct tag.
- name: Load tagged container images into kind cluster
run: ./kind-images.sh $(git rev-parse --short ${{ github.sha }}) ${{ env.CLUSTER_NAME }}
working-directory: web-apps
# https://github.com/helm/charts/blob/master/test/README.md#providing-custom-test-values
# Each chart/ci/*-values.yaml file will be treated as a separate test case with it's
# own helm install/test process.
# NOTE(sd109): There's some kind of incompatibility between the GitHub runner environment
# and vLLM v0.11.0's CPU image. During the testing process the vLLM API pod starts, logs
# some messages about 'Automatically detected platform CPU' and the seemingly gets killed
# and enters a crash loop. Enabling debug logging on vLLM with VLLM_LOGGING_LEVEL=DEBUG
# doesn't provide any useful clues and the same vLLM CPU tests work fine on a standard
# Ubuntu 24.04 VM outside of GitHub actions. Disabling this CI test for now.
# TODO: Try re-enabling this when a newer vLLM version is available.
# - name: Run chart install and test
# run: ct install --config ct.yaml