Summary
useCreativeStore.ts:163 advances a timer and displays checkpoint/latent/VAE stages independently of engine events, including cloud runs. engine_manager.py:396 returns placeholder ready/verified log lines when supervisors cannot supply logs. Supervisor stdout/stderr goes to DEVNULL; PID existence can be displayed as running without service health.
Environment and evidence
- Review finding: F33 (2026-10-05 product/technical review).
- Baseline: Windows,
dev at c1c5dbe. The remote dev matched this commit when filing.
- Evidence: Code + Browser.
- Priority recommendation: P2 - material improvement or integration validation. This is review triage, not a production-incident severity declaration.
- No real credentials, paid inference, engine installation, or unrelated process termination were used for review probes. Mocked observations establish the stated code behavior, not live-provider/GPU acceptance.
Reproduction or validation
Inspect the creative progress timer: it invents checkpoint/latent/VAE stages regardless of engine events, including cloud runs. EngineManager.get_logs returns placeholder ready/verified messages when supervisors have no log method, while stdout/stderr is discarded. These code paths were reviewed; fake-ready output is not evidence that any engine was actually healthy.
User impact
A hung or failed engine looks active and support loses the real error. Low-level latent/VAE terminology also violates the intended high-level experience.
Proposed approach
Use observed phases or an honest indeterminate state, retain bounded real logs with redaction, distinguish starting/running/healthy, and show actionable errors.
Acceptance criteria
Related work
Related closed issue: #37. Real execution stages were previously requested. This report concerns simulated stages and placeholder diagnostics rather than observed progress.
Verification scope
Real GPU inference, paid-provider compatibility, and a clean-machine packaged desktop journey remain unverified. Any follow-up implementation should target dev under the repository's contribution/branching rules.
Summary
useCreativeStore.ts:163 advances a timer and displays checkpoint/latent/VAE stages independently of engine events, including cloud runs. engine_manager.py:396 returns placeholder ready/verified log lines when supervisors cannot supply logs. Supervisor stdout/stderr goes to
DEVNULL; PID existence can be displayed as running without service health.Environment and evidence
devat c1c5dbe. The remotedevmatched this commit when filing.Reproduction or validation
Inspect the creative progress timer: it invents checkpoint/latent/VAE stages regardless of engine events, including cloud runs. EngineManager.get_logs returns placeholder ready/verified messages when supervisors have no log method, while stdout/stderr is discarded. These code paths were reviewed; fake-ready output is not evidence that any engine was actually healthy.
User impact
A hung or failed engine looks active and support loses the real error. Low-level latent/VAE terminology also violates the intended high-level experience.
Proposed approach
Use observed phases or an honest indeterminate state, retain bounded real logs with redaction, distinguish starting/running/healthy, and show actionable errors.
Acceptance criteria
Related work
Related closed issue: #37. Real execution stages were previously requested. This report concerns simulated stages and placeholder diagnostics rather than observed progress.
Verification scope
Real GPU inference, paid-provider compatibility, and a clean-machine packaged desktop journey remain unverified. Any follow-up implementation should target
devunder the repository's contribution/branching rules.