Go Rabby Io Well Known

Rabby Go Well-Known

Boundary

The Rabby Go static site is maintained in the standalone repository:

https://github.com/RabbyHub/go.rabby.io

This rabby-mobile repository should keep only the native app configuration that consumes those links, such as Associated Domains, Android intent filters, and deep-link handling code. Do not reintroduce apps/go.rabby.io or publish Rabby Go .well-known files from rabby-mobile workflows.

The association files are:

.well-known/apple-app-site-association
.well-known/assetlinks.json

Both files must be valid JSON, served directly with HTTP 200, no redirect, and Content-Type: application/json.

Targets

RabbyHub/go.rabby.io maps deploy targets to S3 roots and public hosts:

The current GitHub Actions workflow is .github/workflows/deploy-regression.yml in RabbyHub/go.rabby.io. Its workflow_dispatch input supports regression and production; regression is the default. The deploy script also supports debug, but CI does not expose debug unless a reviewed workflow change adds it.

Edit

1. Work in RabbyHub/go.rabby.io from an up-to-date main branch.

git fetch origin main
git switch -c chore/update-go-rabby-well-known origin/main

2. Edit only the necessary association file(s). Preserve these native contracts unless the app configuration is intentionally changing in the same rollout:

• iOS app IDs use team id ZPNP2SF27Q.
• iOS paths are /mobile-debug/*, /mobile-regression/*, and /mobile/*.
• Android package names are com.debank.rabbymobile.debug, com.debank.rabbymobile.regression, and com.debank.rabbymobile.
• Android fingerprints must come from the actual signing certificate for that package/flavor. Do not copy a fingerprint across flavors by assumption.

3. Validate locally before opening a PR:

python3 -m json.tool .well-known/apple-app-site-association >/dev/null
python3 -m json.tool .well-known/assetlinks.json >/dev/null
corepack yarn install --immutable
corepack yarn check

4. Open a PR to RabbyHub/go.rabby.io:main. Keep PR text limited to the changed app-link contract and validation. Do not include credentials, local-only paths, or secret values.

Deploy

Deploy regression first. Deploy production only after the regression result is verified and the production rollout is approved.

Regression:

gh workflow run deploy-regression.yml \
  -R RabbyHub/go.rabby.io \
  --ref main \
  -f target=regression

Production:

gh workflow run deploy-regression.yml \
  -R RabbyHub/go.rabby.io \
  --ref main \
  -f target=production

The workflow needs these repository secrets:

AWS_ACCESS_KEY_ID
AWS_SECRET_ACCESS_KEY
RABBY_MOBILE_BUILD_BUCKET

The workflow currently uploads to S3 with SKIP_CDN_INVALIDATION=true, so CDN caches may need time to expire. Do not add cloud credentials to code, workflow YAML, PR text, logs, or issue comments.

Verify

After deployment, inspect the workflow run and verify the public files:

gh run list \
  -R RabbyHub/go.rabby.io \
  --workflow deploy-regression.yml \
  --limit 5

curl -fsS -D /tmp/go-rabby-aasa.headers \
  https://go-regression.rabby.io/.well-known/apple-app-site-association \
  -o /tmp/go-rabby-aasa.json
python3 -m json.tool /tmp/go-rabby-aasa.json >/dev/null

curl -fsS -D /tmp/go-rabby-assetlinks.headers \
  https://go-regression.rabby.io/.well-known/assetlinks.json \
  -o /tmp/go-rabby-assetlinks.json
python3 -m json.tool /tmp/go-rabby-assetlinks.json >/dev/null

For production, replace the host with go.rabby.io. Confirm the response is 200, not 3xx or cached 403, and that the content type is JSON.

If the public URL appears stale, distinguish S3 freshness from CDN cache before changing code. With an authorized shell that already has AWS credentials:

aws s3api head-object \
  --bucket "$RABBY_MOBILE_BUILD_BUCKET" \
  --key "rabby-go-regression/.well-known/apple-app-site-association" \
  --query '[LastModified,ContentType,CacheControl,ContentLength]' \
  --output table

aws s3api head-object \
  --bucket "$RABBY_MOBILE_BUILD_BUCKET" \
  --key "rabby-go-regression/.well-known/assetlinks.json" \
  --query '[LastModified,ContentType,CacheControl,ContentLength]' \
  --output table

Use rabby-go instead of rabby-go-regression for production. If S3 is fresh but CDN is stale, wait for cache expiry or perform a deliberate CloudFront invalidation only as an explicit deployment operation.