Configured API7 Gateway to use Kubernetes Service Discovery instead of
static upstream nodes. This enables dynamic discovery of backend Pods
through the Kubernetes API.
Benefits:
- Automatic scaling: New Pods are automatically added to upstream pool
- Health checks: Only healthy Pods receive traffic
- Zero downtime: Automatic updates during deployments and rollouts
- No manual upstream configuration needed
Changes:
- Updated configmap-adc.yaml to use discovery_type: kubernetes
- Service discovery queries Kubernetes API for Pod endpoints
- Falls back to static nodes if serviceDiscovery.enabled is false
- Added documentation in values.yaml explaining the feature
The RBAC permissions (services, endpoints watch) were already configured
in rbac-adc.yaml, so no additional permissions are needed.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Modified Ingress configuration to route all traffic through the API7
Gateway (gateway-0-1759393614-gateway) instead of directly to application
services. This enables API7's advanced routing, rate limiting, CORS, and
other gateway features.
Changes:
- Updated ingress.yaml template to support gateway backend routing
- Modified values.yaml to route traffic to API7 Gateway service
- Disabled web and api services (now optional) as routing is handled by API7
- Removed nginx.ingress.kubernetes.io/rewrite-target annotation
- Maintained backward compatibility with legacy service-based routing
The Ingress now directs traffic to the API7 Gateway which handles all
routing logic defined in the ADC configuration.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Made the global_rules section conditional on logging being enabled.
Previously, when logging was disabled, global_rules was rendered as
an empty object (null), causing a lint error:
"Invalid input: expected record, received null at global_rules"
Now the entire global_rules section is only included when there are
actual rules to add.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Removed the prometheus-metrics global rule configuration which was
causing "custom plugin (prometheus-metrics) not found" error.
API7 Enterprise doesn't support prometheus as a global rule plugin
in this configuration format. Prometheus metrics can be configured
differently if needed.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Changed CORS plugin array values to comma-separated strings:
- allow_origins: from JSON array to comma-separated string
- allow_methods: from JSON array to comma-separated string
- allow_headers: from JSON array to comma-separated string
- expose_headers: from JSON array to comma-separated string
API7 Enterprise expects string values with comma-separated items,
not JSON arrays. This fixes the validation error:
"Invalid type. Expected: string, given: array"
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Changed adminUrl to use HTTPS instead of HTTP:
- Before: http://api7ee3-0-1759339083-dashboard:7080
- After: https://api7ee3-0-1759339083-dashboard:7443
Also enabled tlsSkipVerify for the dashboard's self-signed certificate.
Testing revealed that:
- Port 7080 HTTP doesn't work (pod listens on 7081 localhost only)
- Port 7443 HTTPS is the correct admin API endpoint
- Self-signed certificate requires TLS verification skip
This fixes the ADC sync job 404 errors.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Added conditional --tls-skip-verify flag to ADC sync job arguments.
This flag is controlled by .Values.api7.adc.tlsSkipVerify and allows
ADC to connect to API7 Enterprise dashboard with self-signed certificates.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Changed adminUrl from dp-manager:7900 to dashboard:7080 for API7
Enterprise backend. The dp-manager service is for APISIX, while
API7EE requires the dashboard service which exposes the admin API.
This fixes the 404 errors when ADC tries to fetch configuration:
- GET /api/version
- GET /api/gateway_groups
- GET /api/schema/core
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Changed global_rules from array format (- id: name) to object format
(name:) to match ADC schema requirements. This fixes the lint error:
"Invalid input: expected record, received array at global_rules"
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added OpenAI-compatible LLM endpoints to API backend - Introduced web
frontend with Jinja2 templates and static assets - Implemented API proxy
routes in web service - Added sample db.json data for items, users,
orders, reviews, categories, llm_requests - Updated ADC and Helm configs
for separate AI and standard rate limiting - Upgraded FastAPI, Uvicorn,
and added httpx, Jinja2, python-multipart dependencies - Added API
configuration modal and client-side JS for web app
- Changed chart name from api7ee to api7ee-demo-k8s in Chart.yaml
- Renamed helm/api7ee directory to helm/api7ee-demo-k8s
- Updated all references in build.yml workflow
- Updated all references in helm-release.yml workflow
- Updated main README.md with new chart name
- Updated Helm chart README with new chart name
- Verified all old references have been replaced
- Chart packages correctly as api7ee-demo-k8s-{version}.tgz
