@calltelemetry/ct-lab-mcp 0.8.149 → 0.8.151

package/README.md

@@ -99,6 +99,7 @@ than use the legacy Node VMID/name slot state when the bridge is unavailable.
   drift, or clustered Proxmox. Force host
   power-off is not implemented.
 - **Claimed-Guest Appliance Update**: `appliance_update` is a thin canonical bridge for `lab-inventory vm update`: dry-run by default, apply requires `confirmAssetId` equal to `assetId`, an active lease, a `calltelemetry-lab-claim/v1` claim file authorizing `appliance-update`, and an exact release version. The older in-guest `appliance_exec` shared-login flow is retired; use `appliance_update` for updates, `uat_access` for governed web access, and `lab-inventory vm ssh` (run from an operator machine or a lab-LAN host; the `ct-uat-proxmox` runner is retired as a load driver) for operator shell access.
+- **Candidate Reconciliation**: `appliance_candidate_update` accepts `reconcile: true` for one existing operation ID, its exact v1 manifest digest, and the current Ready ApplianceClaim UID. It reads the original local and coordinator evidence plus fixed pinned-SSH image and container observations. `deployment_outcome` is separate from `service_health`; JTAPI health is not queried and qualification remains `not_assessed`. `completed` requires the original terminal receipt, matching coordinator completion, and exact live candidate images; `failed` requires recorded rollback and matching previous image IDs; `not_applied` requires an explicit durable `candidate_mutation_dispatch_state: not_started` fact and exact live image observations, while quarantine remains held. Missing, old, or ambiguous evidence stays `unknown`. It never reapplies candidate transport or calls retired legacy diagnostics.
 - **Leased Remote Development**: `lab_dev` is a canonical-only façade for the `ct-local-dev-v1` profile. It supports strict `doctor`, `plan`, `prepare`, `start`, `status`, `sync`, `stop`, and `cleanup` actions bound to one asset, lease, Linear issue, operation ID, and reviewed receipt chain. It accepts no guest commands, remote paths, or mock fallback.
 - **Claim custody annotation** (REL-1746; slots and reservation Leases are retired): Proxmox claim creation requires the actual canonical node, explicit VMID and VM name in `ct-lab`, with a finite TTL. The original-owner record lives on the claim itself as annotation `calltelemetry.com/claim-custody` (`claim-custody/v1`: guest tuple, phase, operation nonce, creation-request hash, context, provider binding), written in the same single create call as the claim. No Lease is ever created, renewed or required; per-guest serialization is in-process keyed by node and VMID, and the existing claim-occupancy scan plus the claim UID/resourceVersion protect against other actors. Whole effective immutable CRD fields and creation labels are rehashed on retry and deletion; only the declared lifecycle fields and stale label can change. A matching name alone grants no custody. Claims created before the retirement still carry a `ct-physical-slot-reservation.*` Lease: the MCP verifies them read-only (reporting `custody.source: "legacy-lease"`, `migration: "operator-pending"`) and the appliance operator converts them into the annotation; the MCP never writes a Lease.
 - **One configured Kubernetes claim backend**: ApplianceClaim CRUD and read-only legacy reservation Lease reads share one backend resolver. Local stdio deployments must set an absolute `LAB_CLAIM_KUBECONFIG`, the exact `LAB_CLAIM_K8S_CONTEXT`, and an HTTPS-origin `LAB_CLAIM_K8S_SERVER`; the selected kubeconfig is validated against that server before a request. In-cluster deployments must set the context/server and provide the service-account token and CA for that same cluster. Cross-cluster deployments use the explicit kubeconfig path. The process never falls back to `KUBECONFIG`, `~/.kube/config`, or an embedded server, and a missing or mismatched binding is refused before any claim effect. The container image is published to Docker Hub by the Docker Build and Publish workflow, which only builds and pushes; it does not edit any cluster manifest. ct-lab-mcp is not Flux-deployed, and the retired DOKS cluster is no longer a target.

package/dist/src/generated-version.d.ts

@@ -2,5 +2,5 @@
  * Auto-synced package version fallback for non-filesystem runtimes (Cloudflare Workers).
  * Updated automatically during release publication in .github/workflows/publish.yml.
  */
-export declare const PACKAGE_VERSION = "0.8.149";
+export declare const PACKAGE_VERSION = "0.8.151";
 //# sourceMappingURL=generated-version.d.ts.map

package/dist/src/generated-version.js

@@ -2,5 +2,5 @@
  * Auto-synced package version fallback for non-filesystem runtimes (Cloudflare Workers).
  * Updated automatically during release publication in .github/workflows/publish.yml.
  */
-export const PACKAGE_VERSION = "0.8.149";
+export const PACKAGE_VERSION = "0.8.151";
 //# sourceMappingURL=generated-version.js.map