Complete checklist for deploying Exa search integrations to production. Covers API key management, error handling verification, performance baselines, monitoring, and rollback procedures.
.env files in .gitignorerequestId captured from error responsesmoderation: true) for user-facing searchfast/auto/neural)numResults minimized per use case (3-5 for most)maxCharacters set on text and highlightsset -euo pipefail
echo "=== Exa Pre-Flight ==="
# 1. Verify production API key works
HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" \
-X POST https://api.exa.ai/search \
-H "x-api-key: $EXA_API_KEY_PROD" \
-H "Content-Type: application/json" \
-d '{"query":"pre-flight check","numResults":1}')
echo "API Status: $HTTP_CODE"
[ "$HTTP_CODE" = "200" ] || { echo "FAIL: API key invalid"; exit 1; }
# 2. Verify tests pass
npm test || { echo "FAIL: Tests failing"; exit 1; }
echo "Pre-flight PASSED"
import Exa from "exa-js";
const exa = new Exa(process.env.EXA_API_KEY);
app.get("/health/exa", async (_req, res) => {
const start = performance.now();
try {
const result = await exa.search("health check", { numResults: 1 });
const latencyMs = Math.round(performance.now() - start);
res.json({
status: "healthy",
latencyMs,
resultCount: result.results.length,
timestamp: new Date().toISOString(),
});
} catch (err: any) {
res.status(503).json({
status: "unhealthy",
error: err.message,
errorCode: err.status,
latencyMs: Math.round(performance.now() - start),
});
}
});
set -euo pipefail
# Deploy canary (10% traffic)
kubectl apply -f k8s/production.yaml
kubectl rollout pause deployment/exa-service
echo "Canary deployed. Monitor for 10 minutes..."
echo "Check: /health/exa endpoint, error rates, latency"
# After monitoring, resume to full rollout
# kubectl rollout resume deployment/exa-service
set -euo pipefail
# Verify production endpoint
curl -sf https://your-app.com/health/exa | python3 -m json.tool
# Check error rates (if Prometheus available)
curl -s "localhost:9090/api/v1/query?query=rate(exa_search_error[5m])" 2>/dev/null
set -euo pipefail
# Immediate rollback
kubectl rollout undo deployment/exa-service
kubectl rollout status deployment/exa-service
echo "Rollback complete. Verify /health/exa endpoint."
| Alert | Condition | Severity |
|---|---|---|
| API Down | 5xx errors > 10/min | P1 |
| Auth Failure | 401/403 errors > 0 | P1 |
| Rate Limited | 429 errors > 5/min | P2 |
| High Latency | P95 > 5000ms | P2 |
| Budget Warning | Daily searches > 80% of limit | P3 |
| Issue | Cause | Solution |
|---|---|---|
| Health check fails | API key not set in prod | Verify secret injection |
| Latency spike after deploy | Missing cache warm-up | Pre-populate cache |
| Rate limit on launch | Traffic spike | Enable request queue |
| Rollback needed | Error rate spike | kubectl rollout undo |
For version upgrades, see exa-upgrade-migration. For incident response, see exa-incident-runbook.