OpenEvidence Deploy Integration

Table of Contents

• Overview
• Prerequisites
• Instructions
• Output
• Error Handling
• Examples
• Resources

Overview

Deploy OpenEvidence clinical AI integrations to production healthcare environments with canary releases, Kubernetes manifests, health checks, and rollback procedures.

Prerequisites

• CI pipeline configured (see openevidence-ci-integration)
• Production credentials from OpenEvidence
• Cloud platform configured (GCP, AWS, Azure)
• Signed BAA for production environment

Instructions

Step 1: Multi-Environment Configuration

Define environment configs (dev/staging/prod) with baseUrl, timeout, retries, secret paths, and feature flags.

Step 2: Create Deployment Workflow

Build GitHub Actions workflow with test -> build -> deploy-staging -> deploy-production stages. Use canary releases for production (no-traffic deploy, 10% canary, monitor, full rollout).

Step 3: Configure Kubernetes

Create Deployment (3 replicas), Service, Ingress, and HPA manifests. Use K8s secrets for API credentials, liveness/readiness probes against /health.

Step 4: Implement Health Checks

Create /health (basic) and /health/openevidence (deep) endpoints that verify API connectivity and report latency.

Step 5: Prepare Rollback

Create rollback script that lists recent revisions, identifies previous, and shifts traffic back with health verification.

Output

• Multi-environment deployment configuration
• CI/CD pipeline with canary releases
• Kubernetes deployment manifests (Deployment, Service, Ingress, HPA)
• Health check endpoints (basic + deep)
• Rollback procedures and scripts

Error Handling

Examples

Deploy to Staging

gcloud run deploy clinical-evidence-api-staging \
  --image $IMAGE --region us-central1 \
  --set-env-vars NODE_ENV=staging \
  --set-secrets OPENEVIDENCE_API_KEY=openevidence-staging-api-key:latest

Rollback Production

gcloud run services update-traffic clinical-evidence-api \
  --region us-central1 --to-revisions PREVIOUS_REVISION=100

See detailed implementation for advanced patterns.

Resources

• Cloud Run Documentation
• Kubernetes Deployments