Skip to content

Deploy with Cloudflare ​

The current supported app is Node.js plus PostgreSQL. You can use Cloudflare for Tunnel ingress, DNS, security controls, private R2 object storage, and Pages documentation while running the application on your Docker host. Follow the Docker guide for executable setup.

Can everything run on Cloudflare? ​

There is currently no maintained, tested deployment of this release entirely on Cloudflare. The repo retains a former Workers/D1/Durable Objects/Queues/KV/R2 implementation for compatibility and historical recovery. Its wrangler.toml, resource IDs, migration history and legacy:cloudflare:* scripts are not a fresh-account installer. Do not deploy those files against the maintainer's resources or run PostgreSQL migrations against D1.

This follows the repository's portable-runtime architecture decision. Restoring a Cloudflare-only release is a separate project; the self-hosting announcement must not describe it as available.

Use private R2 object storage ​

Create a private R2 bucket in your own account. Create S3 API credentials with Object Read & Write permissions scoped to that bucket. Keep public bucket access disabled: transcripts and evidence are served through the application's authorization and signed-link paths.

Set these in the app's private environment:

dotenv
S3_BUCKET_NAME=your-private-haf-bucket
S3_REGION=auto
S3_ENDPOINT=https://YOUR_ACCOUNT_ID.r2.cloudflarestorage.com
S3_FORCE_PATH_STYLE=true
AWS_ACCESS_KEY_ID=replace-with-your-r2-access-key-id
AWS_SECRET_ACCESS_KEY=replace-with-your-r2-secret-access-key

Use R2's S3 access key pair, not a general Cloudflare API bearer token. Configure only one object provider; remove Azure Blob settings when using S3/R2. TICKET_TRANSCRIPTS_BUCKET is a legacy Worker binding and is not how the Docker app selects a bucket. Test uploading, reading and deleting an authorized test object, and back up the object data as well as PostgreSQL.

These settings use Cloudflare's documented S3 interface and R2 region/endpoint convention.

Publish documentation on Pages ​

The documentation is a static VitePress site. From a source checkout, run pnpm install --frozen-lockfile, then pnpm docs:build. The output is docs-site/.vitepress/dist. Publish that directory to your own Pages project with the Pages direct-upload flow.

Update the hostname, canonical metadata, navigation and hosted-service links in docs-site/.vitepress/config.mts and relevant pages for your fork. The repository's docs:deploy script names the maintainer's Pages project; use your own project instead. Pages hosts the documentation, not the Node application or Discord Gateway.

What a Cloudflare Containers deployment would need ​

Cloudflare Containers can run an OCI image, but this repository does not supply the Worker/Durable Object controller and routing configuration required to operate it there. Cloudflare documents the container lifecycle and the Durable Object Container API.

A future adapter must be implemented and tested for:

  • Worker-to-container routing, public origin handling, required secrets and port readiness;
  • continuous Gateway, jobs and scheduler operation even when no public HTTP requests arrive, including restart, inactivity, rollout, reconnect and lease-renewal behavior;
  • durable PostgreSQL outside the container's ephemeral filesystem, with verified TLS, migrations, database backups and recovery; Durable Object SQLite and D1 do not satisfy the current PostgreSQL store contracts;
  • private R2 through the existing S3 interface, with credentials and object recovery;
  • singleton Gateway ownership and scheduler fencing across restarts and deployments;
  • bounded resource use, current platform pricing, monitoring and an independently tested rollback.

Cloudflare documents that container disks are ephemeral by default. Do not treat a PostgreSQL container filesystem or an R2 FUSE mount as a substitute for the production database's durability and recovery requirements. A Containers adapter would still require a PostgreSQL service, so it would not by itself provide a deployment using only Cloudflare products.

What reviving Workers/D1 would need ​

A separate supported release would need fresh-account resource provisioning, sanitized example Wrangler configuration, explicit secret setup, current D1 migration checks, Gateway/queue/retry tests, scoped deployment credentials, budget controls and recovery evidence. PostgreSQL state and SQL are not interchangeable with D1; Cloudflare explicitly documents the PostgreSQL import limitation.

Until that work is complete, use Docker with Tunnel and optional R2. This gives you Cloudflare ingress and storage while retaining the verified Node/PostgreSQL deployment shape.