When a promise failed, the whole queue was ruined

[jpggvilaca]

📝 Description

Issues addressed:

• Browser (Chrome): intermittent JSEP Resize kernel failures → WASM heap corruption → every subsequent ortSession.run() returns garbage.
• Tauri (WebView2 / WKWebView): ortSession.run() occasionally hangs forever (WebGPU / memory related). The serial runQueue chained every later call onto the never-settling promise, blocking the session indefinitely.
• Random issue we had before on chrome that the worker was forever loading (dont await for cache.put)
• Cleanup/simplification

Manually tested Geti Tune browser and tauri app (by updating the hash of the packages that 'tiged' uses)

✨ Type of Change

Select the type of change your PR introduces:

• 🐞 Bug fix – Non-breaking change which fixes an issue
• 🚀 New feature – Non-breaking change which adds functionality
• 🔨 Refactor – Non-breaking change which refactors the code base
• 💥 Breaking change – Changes that break existing functionality
• 📚 Documentation update
• 🔒 Security update
• 🧪 Tests

🧪 Testing Scenarios

Describe how the changes were tested and how reviewers can test them too:

• ✅ Tested manually
• 🤖 Run automated end-to-end tests

✅ Checklist

Before submitting the PR, ensure the following:

• 🔍 PR title is clear and meaningful
• ✍️ PR description clearly explains the changes and their reason
• 📝 I have linked the PR to the corresponding GitHub Issues, if any
• 💬 I have commented my code, especially in hard-to-understand areas
• 📄 I have made corresponding changes to the documentation
• ✅ I have added tests that prove my fix is effective or my feature works

[Copilot]

Pull request overview

This PR hardens the smart-tools Web inference pipeline against browser/WebView pathologies where onnxruntime-web runs can hang indefinitely or leave the WASM heap corrupted, and where cache writes can stall model loading.

Changes:

• Make model caching best-effort (fire-and-forget cache.put) so model loads aren’t blocked by slow/blocked Cache API writes.
• Add session “poisoning” semantics, optional per-run timeouts, and a reset() path to recreate the underlying InferenceSession from cached model bytes.
• Refactor the serial run queue implementation to avoid propagating prior failures through the queue chain.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
web_ui/packages/smart-tools/src/utils/tool-utils.ts	Makes caching non-blocking by not awaiting cache.put() and swallowing cache errors.
web_ui/packages/smart-tools/src/segment-anything/session.ts	Adds timeout/poison/reset mechanics and refactors the serialized run() queue implementation.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[dwesolow]

Tested those changes in tauri, works fine.
Thanks!