[KAJdev]

Brings decorator-native GPU compute into the core runpod package, with queue, API, and task resources sharing one app lifecycle and a live development workflow.

Worker execution is versioned independently through runpod-sdk-runtime, allowing runtime images and SDK clients to evolve on separate release cycles. Flash lifecycle commands live under rp flash, while account, secret, registry, pod, and SSH commands stay at the root.

Completes CON-605
Completes CON-601
Completes CON-750

[KAJdev]

Should be ready for a re-review. Took a look at all these comments, main things were addressed.

• rebase with the new REST wrapper API client for all supported operations
• resource validation and GPU/datacenter/volume placement, including multi-GPU and multi-volume configurations
• endpoint and pod cleanup across partial startup, failures, timeouts, and cancellation plus retry policies that avoid duplicating mutations
• nested sync/async worker calls, concurrent source synchronization, result/error handling, and dependency packaging
• independently versioned runtime images, custom-image bootstrapping, and runtime publication.
• .remote() calls on queue based endpoints now use /runsync first, then switch to polling during longer requests

[deanq]

Re-review after the REST-wrapper rebase — the earlier blockers all look resolved (placement datacenter enum, dev teardown orphaning, the targets.py pod leaks, etc.). A few new issues in the new/rebased code, inline below.

The tasks.py gpu="any" one is the same enum→device-name class as the placement.py fix from last round and would block task deploys; the rp_info.py get_pod change is a retry regression on the rp pod connect path. The rest are robustness/correctness.

[deanq]

hey @KAJdev — poking this one. the 7 threads from my last pass are still open and the code hasn't moved. couple of them are actual bugs, not nits:

• apps/tasks.py — gpu="any" / GpuGroup.ANY collapses to a literal gpuTypeIdList=['any'], so that path just can't deploy
• apps/api.py — current.get('type') on a maybe-None 204 body, plus the instanceIds fallback ending up [None] and deploying with instanceId=None
• api/rest.py — session._retry_connection = False does nothing (aiohttp has no such attr), so PUT/DELETE replay isn't actually prevented
• cli/utils/rp_info.py — the raise ValueError makes it bail on the first null response instead of retrying till timeout

the gpu.py docstring mismatch and the serialization.py nested-def thing are the minor two. ping me once these have a look and i'll re-review

[Copilot]

Copilot review overview

🟡 Changes recommended

Authentication, live refresh, SSE parsing, and task secret-validation bugs remain unresolved.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 5 · 1

Open (6)

• GraphQL requests send masked value instead of API key · New
• Async transport sends masked instead of real authorization key · New
• Log requests use masked authorization instead of API key · New
• Invocation requests use masked authorization instead of API key · New
• Pod runner uses masked token for task authentication · New
• Rescan reuses cached module and stale app registrations · New

Resolved since last review (12)

• When no RUNPOD_PACKAGE_SPEC is set, this shim never installs the client SDK. That breaks the…
• _execute() retries every transport failure, but it is shared by non-idempotent mutations such as…
• session.start() runs before the try/finally that calls session.stop(). If provisioning a…
• If polling times out or unwrap_task_response raises for a failed task, _done is still false, so…
• The spawn path has no cleanup if pod startup, readiness, or /submit fails after the pod is…
• The stock lookup key omits spec.gpu_count, and _fetch_gpu therefore queries…
• The endpoint is removed from _endpoints before deletion succeeds. If delete_endpoint fails, the…
• dev_endpoint_name() appends a digest after the resource name, and resource names may contain…
• volume_list and attach_endpoint_volumes support multiple endpoint volumes, but this public…
• datacenter=DataCenter.EU_RO_1 is a valid public enum value, but this helper stores the enum…
• The or True makes this assertion unconditional, so the test never verifies that…
• The exception guidance contains the malformed fragment str(...) it, which is confusing at the…

[promptless]

Promptless documentation updates

• docs: Use rp login in SkyPilot setup instead of removed runpod config updates the SkyPilot setup guide to authenticate with rp login --api-key instead of the removed runpod config. Docs PR