Merge pull request #20 from IDinsight/fix-litellm-deployment #6
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
name: Deploy LiteLLM Proxy to GCP | |
on: | |
push: | |
branches: | |
- main | |
- testing | |
- production | |
paths: | |
- "deployment/docker-compose/litellm_proxy_config.yaml" | |
- ".github/workflows/deploy_gcp_litellm_proxy.yaml" | |
release: | |
types: [released] | |
workflow_dispatch: | |
jobs: | |
set-env: | |
runs-on: ubuntu-latest | |
outputs: | |
env_name: ${{ steps.set-env.outputs.env_name }} | |
steps: | |
- name: Resolve deployment environment name | |
id: set-env | |
run: | | |
if [ "${{ github.event_name }}" == "release" ] && [ "${{ github.event.action }}" == "released" ]; then | |
echo "env_name=production" >> "$GITHUB_OUTPUT" | |
elif [ "${{ github.ref_name }}" == "main" ]; then | |
echo "env_name=testing" >> "$GITHUB_OUTPUT" | |
else | |
echo "env_name=${{ github.ref_name }}" >> "$GITHUB_OUTPUT" | |
fi | |
DeployLiteLLMProxyToGCP: | |
needs: [set-env] | |
runs-on: ubuntu-latest | |
permissions: | |
contents: "read" | |
id-token: "write" | |
environment: mc-aaq-${{ needs.set-env.outputs.env_name }} | |
env: | |
RESOURCE_PREFIX: ${{ secrets.PROJECT_NAME }}-${{ needs.set-env.outputs.env_name }} | |
steps: | |
- uses: "actions/checkout@v4" | |
- id: "auth" | |
name: "Authenticate to Google Cloud" | |
uses: "google-github-actions/auth@v2" | |
with: | |
project_id: ${{ secrets.GCP_PROJECT_ID }} | |
workload_identity_provider: projects/${{ secrets.GCP_PROJECT_NUMBER }}/locations/global/workloadIdentityPools/${{ vars.POOL_ID }}/providers/${{ vars.PROVIDER_ID }} | |
service_account: ${{ secrets.GCP_SERVICE_ACCOUNT_EMAIL }} | |
- name: Retrieve secrets from Secret Manager | |
id: "secrets" | |
uses: "google-github-actions/get-secretmanager-secrets@v2" | |
with: | |
secrets: |- | |
domain:${{ secrets.GCP_PROJECT_ID }}/${{ env.RESOURCE_PREFIX }}-domain | |
openai-api-key:${{ secrets.GCP_PROJECT_ID }}/${{ env.RESOURCE_PREFIX }}-openai-api-key | |
gcp-credential-json:${{ secrets.GCP_PROJECT_ID }}/${{ env.RESOURCE_PREFIX }}-gcp-credential-json | |
huggingface-api-key:${{ secrets.GCP_PROJECT_ID }}/${{ env.RESOURCE_PREFIX }}-huggingface-api-key | |
huggingface-endpoint:${{ secrets.GCP_PROJECT_ID }}/${{ env.RESOURCE_PREFIX }}-huggingface-endpoint | |
- name: Copy LiteLLM deployment files | |
working-directory: deployment/docker-compose | |
run: | | |
gcloud compute scp litellm_proxy_config.yaml \ | |
${{ secrets.DEPLOYMENT_INSTANCE_NAME }}:~/litellm_proxy_config.yaml \ | |
--zone ${{ secrets.DEPLOYMENT_ZONE }} | |
echo '${{ steps.secrets.outputs.gcp-credential-json }}' > .gcp_credentials.json | |
gcloud compute scp .gcp_credentials.json \ | |
${{ secrets.DEPLOYMENT_INSTANCE_NAME }}:~/.gcp_credentials.json \ | |
--zone ${{ secrets.DEPLOYMENT_ZONE }} | |
- name: Deploy LiteLLM Proxy container | |
id: "compute-ssh" | |
uses: "google-github-actions/ssh-compute@v1" | |
with: | |
instance_name: "${{ secrets.DEPLOYMENT_INSTANCE_NAME }}" | |
zone: "${{ secrets.DEPLOYMENT_ZONE }}" | |
ssh_private_key: "${{ secrets.GCP_SSH_PRIVATE_KEY }}" | |
command: | | |
docker stop litellm_proxy | |
docker rm litellm_proxy | |
docker run -d \ | |
-v ~/litellm_proxy_config.yaml:/app/config.yaml \ | |
-v ~/.gcp_credentials.json:/app/credentials.json \ | |
-e OPENAI_API_KEY="${{ steps.secrets.outputs.openai-api-key }}" \ | |
-e VERTEXAI_PROJECT=${{ secrets.GCP_PROJECT_ID }} \ | |
-e VERTEXAI_LOCATION=${{ vars.VERTEX_AI_LOCATION }} \ | |
-e VERTEXAI_ENDPOINT=https://${{ vars.VERTEX_AI_LOCATION }}-aiplatform.googleapis.com/v1 \ | |
-e GOOGLE_APPLICATION_CREDENTIALS=/app/credentials.json \ | |
-e HUGGINGFACE_EMBEDDINGS_API_KEY="${{ steps.secrets.outputs.huggingface-api-key}}" \ | |
-e HUGGINGFACE_EMBEDDINGS_ENDPOINT="${{ steps.secrets.outputs.huggingface-endpoint }}" \ | |
--log-driver=gcplogs \ | |
--restart always \ | |
--network aaq-network \ | |
--name litellm_proxy \ | |
ghcr.io/berriai/litellm:main-v1.40.10 --config /app/config.yaml --num_workers 4 | |
docker system prune -f || true | |
- name: Show deployment command output | |
run: |- | |
echo '${{ steps.compute-ssh.outputs.stdout }}' | |
echo '${{ steps.compute-ssh.outputs.stderr }}' | |
- name: Wait for Application to start | |
id: wait-for-app | |
run: sleep 1m | |
shell: bash | |
- name: Check if deployment was successful | |
id: check-deployment | |
run: | | |
curl -f -X 'GET' \ | |
'https://${{ steps.secrets.outputs.domain }}/api/healthcheck' \ | |
-H 'accept: application/json' |