feat: list machines for every deployment

[alongubkin]

Background

alien deployments machines is the developer-facing inventory command, but it called a Platform endpoint restricted to platform=machines. That made the command fail for ordinary AWS/GCP/Azure deployments even though their ComputeClusters are managed by Horizon and have machine inventory.

Depends on alienplatform/platform#722.

Root cause

The CLI consumed the old provisioning-specific SDK operation (list_machines_inventory). The API contract now models inventory as a deployment observation and aggregates every Horizon-backed ComputeCluster.

Change

• Regenerate the checked-in Platform OpenAPI contract from platform#722.
• Call GET /v1/deployments/{id}/machines through the generated Rust SDK.
• Show the stack Cluster resource ID in human output, making multi-ComputeCluster deployments unambiguous.
• Sort by network divergence, then cluster resource ID, then machine ID.
• Preserve full clusterResourceId data in --json output.

No new top-level CLI or Horizon/horizond endpoint is introduced. The user command remains:

alien deployments machines <deployment>

Reproduction

Before platform#722 and this PR:

alien deployments machines production/api

against an AWS deployment returns MACHINES_DEPLOYMENT_UNAVAILABLE because the old API checks that the deployment platform is machines.

After both PRs, the same command returns all machines across the deployment's ComputeClusters.

Verification

• cargo build -p alien-platform-api generates DeploymentMachine and list_deployment_machines from the committed OpenAPI.
• cargo check -p alien-cli
• cargo test -p alien-cli machines_inventory_tests --lib

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; the previous Rust SDK generation failure is fixed and no new actionable defects were identified.

Findings

1.  Filtered SDK generation fails ▶

Fix with agent prompt

### Issue 1
crates/alien-cli/src/commands/deployments.rs:undefined-669
The CLI now calls `list_deployment_machines()`, but its filtered Platform SDK configuration still requires the removed `listMachinesInventory` operation. Because `alien-cli` disables the SDK's default features, the build script uses that filter and fails before generating the client, leaving the CLI unable to compile. Update the allowlist to require `listDeploymentMachines`.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Summary

This PR moves machine inventory from the Machines-specific API to a deployment-wide Platform operation and updates the generated clients accordingly.

• Calls GET /v1/deployments/{id}/machines from the CLI and generated SDKs.
• Includes the ComputeCluster resource ID in human and JSON output.
• Sorts divergent machines first, followed by cluster resource ID and machine ID.
• Regenerates the TypeScript client to expose deployments.listMachines and remove the obsolete inventory operation.
• Updates the filtered Rust SDK operation allowlist so the CLI can generate and compile successfully.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    User[alien deployments machines] --> CLI[Alien CLI]
    CLI --> SDK[Filtered Platform Rust SDK]
    SDK --> API[GET /v1/deployments/id/machines]
    API --> Inventory[Deployment-wide machine inventory]
    Inventory --> Sort[Sort by network divergence, cluster, machine]
    Sort --> Human[Human table with Cluster column]
    Sort --> JSON[JSON preserving clusterResourceId]

Loading

_{Reviews (2) · Last reviewed commit: "fix(sdk): regenerate platform client for..."}

[alongubkin]

@greptileai re-comment