Remote Developer - Claude Code instances in isolated Kubernetes pods
|
Some checks failed
ci/woodpecker/push/woodpecker Pipeline failed
Fixes issues from code review of resilience implementation:
- Wire saga system in main.go (SagaRepository, SagaExecutor, SagaHandler)
- Fix CompletedSteps() to include skipped steps for dependency resolution
- Fix reverse loop bug in saga compensation (use standard swap pattern)
- Add circuit breaker state change callbacks for Prometheus metrics
Phase 1 (Build Resilience):
- Add failure:retry to all component Kaniko build steps
- Add preflight registry health check before builds
- Add services-deployed sync point to decouple docs from critical path
Phase 2 (API Resilience):
- Add pipeline retry endpoint (POST /projects/{id}/pipelines/{number}/retry)
- Wire circuit breakers with metrics callbacks
- Add /health/circuits endpoint for circuit breaker status
Phase 3 (Saga Engine):
- Full domain model (Saga, SagaStep, RetryPolicy, BackoffType)
- PostgreSQL saga repository with CRUD and step management
- Saga executor with retry, compensation, skip step support
- Saga API handlers with CRUD and control operations
Phase 4 (Observability):
- Add saga metrics (total, step_duration, retry, circuit_breaker_state)
- Add logging fields (saga_id, saga_name, step_name)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
|
||
|---|---|---|
| .claude/guides | ||
| ai-lookup | ||
| changelog | ||
| cmd | ||
| cookbooks | ||
| deployments/k8s/base | ||
| docs | ||
| examples/dashboard-app | ||
| history | ||
| ideas/aeres | ||
| internal | ||
| pkg/api | ||
| scripts | ||
| tests/e2e | ||
| .env.local.example | ||
| .gitignore | ||
| .gitkeep-ci | ||
| .golangci.yml | ||
| .woodpecker.yml | ||
| aeries-preparation-thoughts.md | ||
| app-vision-gaps.md | ||
| app-vision-roadmap.md | ||
| app-vision.md | ||
| CHANGELOG.md | ||
| CLAUDE.md | ||
| CODING_GUIDELINES.md | ||
| docker-compose.yaml | ||
| Dockerfile | ||
| Dockerfile.api | ||
| Dockerfile.api.prebuild | ||
| Dockerfile.worker | ||
| go.mod | ||
| go.sum | ||
| IMPLEMENTATION_PLAN_V2.md | ||
| IMPLEMENTATION_PLAN.md | ||
| Makefile | ||
| PLAN.md | ||
| QUICKSTART.md | ||
| README.md | ||
| vision.md | ||
rdev - Remote Developer
Run Claude Code in isolated Kubernetes pods on your k3s cluster.
Quick Start
# 1. Set kubeconfig (REQUIRED - this is k3s, not GKE)
export KUBECONFIG=~/.kube/orchard9-k3sf.yaml
# 2. Authenticate Claude locally (if not already)
claude
# 3. Create credentials secret
./scripts/create-credentials-secret.sh
# 4. Deploy
./scripts/deploy.sh
# 5. Verify
./scripts/verify.sh
Usage
# Check Claude version
kubectl exec -n rdev claudebox-0 -- claude --version
# Interactive Claude session
kubectl exec -it -n rdev claudebox-0 -- claude "what can you help me with?"
# Run in workspace
kubectl exec -it -n rdev claudebox-0 -- bash
cd /workspace
claude "create a hello world go program"
Architecture
k3s cluster
└── rdev namespace
└── claudebox-0 (StatefulSet)
├── Claude Code CLI
├── /workspace (20Gi PVC via Longhorn)
└── /root/.claude (credentials from secret)
Roadmap
- v0.1: Base case - single claudebox pod
- v0.2: Real workspace mounting (pantheon, aeries)
- v0.3: Git integration (push/pull)
- v0.4: Discord bot control
- v0.5: Streaming output
- v0.6: Multi-project routing
Development
# Build image locally
docker build -t rdev-claudebox:dev .
# Build and push to Artifact Registry
./scripts/build-push.sh v0.1.0
Troubleshooting
# Check pod status
kubectl get pods -n rdev
# View pod logs
kubectl logs claudebox-0 -n rdev
# Describe pod for events
kubectl describe pod claudebox-0 -n rdev
# Check credentials mount
kubectl exec -n rdev claudebox-0 -- ls -la /root/.claude/