This roadmap breaks down the strategic analysis into granular, actionable engineering tasks. The goal is to transition Lumo from a CLI tool into a monetizable Enterprise SRE Platform.
Goal: Expand the remediation catalog so Lumo can actually fix common problems, not just kill processes. This is critical for the "Auto-Remediation" value proposition.
- Implement
CleanTmpFilesAction- Logic: Identify files in
/tmpolder than X days. - Safety: Verify not currently open (lsof).
- Risk: Moderate.
- Logic: Identify files in
- Implement
RotateLogsAction- Logic: Compress/truncate logs in
/var/loglarger than X MB. - Safety: Ensure logrotate isn't already running.
- Risk: Low.
- Logic: Compress/truncate logs in
- Implement
CheckDiskSpaceAction(Diagnostic + Minor Fix)- Logic: Clear package manager caches (
apt clean,yum clean). - Risk: Safe.
- Logic: Clear package manager caches (
- Implement
RestartSystemdServiceAction- Logic:
systemctl restart <service>. - Validation: Check
systemctl statuspre/post. - Risk: Moderate.
- Logic:
- Implement
StopSystemdServiceAction- Logic:
systemctl stop <service>. - Risk: High (requires approval).
- Logic:
- Implement
RestartDockerContainerAction- Logic:
docker restart <container_id>. - Risk: Moderate.
- Logic:
- Implement
K8sRolloutRestartAction- Logic:
kubectl rollout restart deployment/<name>. - Use case: Stuck pods, config updates.
- Risk: Moderate.
- Logic:
- Implement
K8sScaleDeploymentAction- Logic: Scale replicas up/down.
- Use case: Handling traffic spikes.
- Risk: Moderate.
- Implement
K8sDeletePodAction- Logic: Delete pod to force recreation.
- Use case: Stuck in
TerminatingorError. - Risk: Low/Moderate (if part of ReplicaSet).
Goal: Build the "Manager" that enterprises pay for. This separates the open-source CLI (single-player) from the commercial platform (multi-player).
- Agent Management Endpoints
GET /api/v1/agents: List all connected agents with status.GET /api/v1/agents/:id/health: Detailed health history.
- Issue Tracking Endpoints
GET /api/v1/issues: Global feed of detected problems.POST /api/v1/issues/:id/resolve: Mark as resolved manually.
- Approval Workflow Endpoints
GET /api/v1/approvals: List pending remediation actions.POST /api/v1/approvals/:id/approve: Trigger the action on the agent.POST /api/v1/approvals/:id/reject: Cancel the action.
- Scaffold
lumo-dashboard(Next.js + Tailwind + ShadcnUI). - Dashboard Views:
- Overview: High-level stats (Healthy vs. Unhealthy hosts).
- Agent List: Filterable table of all infrastructure.
- Live Feed: Real-time stream of incoming diagnostics.
- Action Center: Dedicated UI for approving/rejecting AI recommendations.
Goal: Remove barriers to adoption by ensuring safety and compliance.
- Implement
AuditService- Record: Who (User/AI), What (Action), Where (Host), When, Result.
- Storage: Append-only table in PostgreSQL.
- UI View:
- Read-only view of the Audit Log in the dashboard.
- Command Injection Audit:
- Review all
exec.Commandcalls. - Ensure arguments are never passed via shell string concatenation.
- Review all
- Dry-Run Implementation:
- Ensure every
Actionhas aDryRun()mode that returns exactly what command would run without running it.
- Ensure every
- Agent Auth: mTLS or Token-based auth for Agent <-> API communication.
- User RBAC:
Admin: Can approve Critical actions.Operator: Can approve Moderate actions.Viewer: Read-only.
Goal: Reduce friction for new users.
- One-Line Installer:
curl | bashscript that detects OS, installs binary, and registers agent. - Docker Compose: Simple quickstart for self-hosting the API/Dashboard.
- Live Demo: A read-only instance of the Dashboard connected to dummy agents simulating issues.