Openevidence multi env setup

OpenEvidence Multi-Environment Setup

Overview

OpenEvidence clinical AI requires strict environment separation to maintain HIPAA compliance across the data lifecycle. Development uses only synthetic patient data with no PHI access, staging operates on de-identified datasets for clinical validation, and production handles full PHI under BAA-covered infrastructure. Each environment enforces its own audit logging, encryption, and access control policies. Misconfigured environments risk PHI exposure and regulatory violations, making environment validation a hard requirement at startup.

Environment Configuration

const openEvidenceConfig = (env: string) => ({
  development: {
    apiKey: process.env.OPENEVIDENCE_API_KEY_DEV!, baseUrl: "https://api.dev.openevidence.com/v1",
    dataClassification: "synthetic", phiEnabled: false, auditLevel: "basic", encryptionRequired: false,
  },
  staging: {
    apiKey: process.env.OPENEVIDENCE_API_KEY_STG!, baseUrl: "https://api.staging.openevidence.com/v1",
    dataClassification: "de-identified", phiEnabled: false, auditLevel: "full", encryptionRequired: true,
  },
  production: {
    apiKey: process.env.OPENEVIDENCE_API_KEY_PROD!, baseUrl: "https://api.openevidence.com/v1",
    dataClassification: "phi", phiEnabled: true, auditLevel: "full", encryptionRequired: true,
  },
}[env]);

Environment Files

# Per-env files: .env.development, .env.staging, .env.production
OPENEVIDENCE_API_KEY_{DEV|STG|PROD}=<api-key>
OPENEVIDENCE_BASE_URL=https://api.{dev.|staging.|""}openevidence.com/v1
OPENEVIDENCE_DATA_CLASS={synthetic|de-identified|phi}
OPENEVIDENCE_PHI_ENABLED={false|false|true}
OPENEVIDENCE_AUDIT_LEVEL={basic|full|full}
OPENEVIDENCE_BAA_ID=<baa-id>          # production only

Environment Validation

function validateOpenEvidenceEnv(env: string): void {
  const suffix = { development: "_DEV", staging: "_STG", production: "_PROD" }[env];
  const required = [`OPENEVIDENCE_API_KEY${suffix}`, "OPENEVIDENCE_BASE_URL", "OPENEVIDENCE_DATA_CLASS"];
  if (env === "production") required.push("OPENEVIDENCE_BAA_ID");
  if (env !== "development") required.push("OPENEVIDENCE_AUDIT_LEVEL");
  const missing = required.filter((k) => !process.env[k]);
  if (missing.length) throw new Error(`Missing OpenEvidence vars for ${env}: ${missing.join(", ")}`);
  if (env === "production" && process.env.OPENEVIDENCE_PHI_ENABLED !== "true")
    throw new Error("HIPAA violation: PHI must be enabled in production");
}

Promotion Workflow

# 1. Run clinical queries against synthetic data in dev
curl -X POST "$OPENEVIDENCE_BASE_URL/query" \
  -H "Authorization: Bearer $OPENEVIDENCE_API_KEY_DEV" -d @synthetic-query.json

# 2. Validate with de-identified data in staging (audit logs required)
curl -X POST "$OPENEVIDENCE_BASE_URL/query" \
  -H "Authorization: Bearer $OPENEVIDENCE_API_KEY_STG" -d @staging-query.json

# 3. Verify HIPAA audit trail exists for all staging queries
curl "$OPENEVIDENCE_BASE_URL/audit/logs?env=staging" \
  -H "Authorization: Bearer $OPENEVIDENCE_API_KEY_STG" | jq '.totalEntries'

# 4. Deploy to production (requires BAA verification)
OPENEVIDENCE_BAA_ID=baa-2026-001 npm run deploy -- --env production --hipaa-check

Environment Matrix

Setting	Dev	Staging	Prod
Data Type	Synthetic only	De-identified	Full PHI
PHI Access	No	No	Yes (BAA required)
Audit Logging	Basic	Full	Full + HIPAA trail
Encryption at Rest	Optional	Required	Required (AES-256)
Access Control	Developer only	Clinical QA team	Authorized clinicians
BAA Required	No	No	Yes

Error Handling

Issue	Cause	Fix
HIPAA validation failed at startup	PHI_ENABLED not set in production	Set OPENEVIDENCE_PHI_ENABLED=true in prod env file
BAA ID missing	Production deploy without BAA reference	Add OPENEVIDENCE_BAA_ID from compliance team
403 on PHI endpoint	Dev/staging key used against prod API	Use environment-specific API key with correct scope
Audit log gap detected	Staging queries not logged	Verify OPENEVIDENCE_AUDIT_LEVEL=full in staging env

Resources

• OpenEvidence Docs

Next Steps

See openevidence-deploy-integration.