​
Self-host Isofold

Prefer to run Isofold yourself? You can deploy it in your own cloud, edge region, or local environment. This gives you complete control over performance, logging, and compliance boundaries.

​
BigQuery via GCP Marketplace

Isofold offers a GCP Marketplace image for BigQuery workloads. This is the fastest way to run Isofold within your own GCP project.

Launch on GCP

Deploy Isofold directly into your GCP project with IAM and VPC controls.

​
After deployment

Once deployed, copy the external IP or internal load balancer address and point your BigQuery client to it: 
bq query \
  --project_id=myproject \
  --use_legacy_sql=false \
  --api_endpoint=http://<your-isofold-ip> \
  'SELECT * FROM dataset.table'

​
Security tips

  • Use a private VPC connection to BigQuery
  • Restrict access to the proxy via IAM or firewall rules
  • Enable logging or verification mode for auditability

​
Manual deployment (Snowflake, Aurora, or custom)

For Snowflake, Aurora, or other engines, you can deploy Isofold using Docker or Fly.io.

​
Option A — Docker

docker run -p 8080:8080 \
  -e BACKEND_TYPE=snowflake \
  -e BACKEND_HOST=snowflake.mycompany.com \
  -e ISOFOLD_VERIFY_MODE=true \
  isofold/proxy:latest
Route traffic through localhost:8080 or your cloud IP.

​
Option B — Fly.io

  1. Install the Fly CLI
  2. Run fly launch and configure your environment
  3. Deploy to a region close to your warehouse
  4. Use the Fly hostname as your proxy URL
We recommend Fly.io for latency-sensitive, multiregion deployments.

​
Configuration Reference

VariableDescription
BACKEND_TYPEbigquery | snowflake | aurora
BACKEND_HOSTHostname of your target warehouse
ISOFOLD_VERIFY_MODERuns original + optimized queries when true
Need help? Contact support or explore the architecture overview.