fix: bound Similarity Explorer resource use

[frathe]

Explorer now exposes persisted limits under Settings -> Limits -> Similarity Explorer: 512 MB of serialized map data per update and 10,000 items by default. Exceeding either limit shows an error toast with that settings path. The toast wraps in narrow windows, and a limit failure preserves installed-model readiness for retry.

The item setting replaces the fixed 5,000-image ceiling and can be raised for the previously qualified 50,655-image library. Client and worker capture and enforce the same configured item limit. Map snapshots travel as metadata, individual items, hierarchy merges, and a final marker, so a collection no longer depends on fitting its entire map into a single 128 MiB event. Both peers enforce the configured aggregate serialized-data budget, and the UI receives only complete snapshots. The memory setting bounds map data, not total process RSS or native inference memory.

Validation:

• Race tests passed for preferences, similarity transport, Settings, and the Explorer feature.
• Focused viewer race regressions passed for Explorer, preference wiring, toasts, and overlay ordering.
• Transport tests cover configured 50,655-item subprocess delivery, exact aggregate boundaries, truncated/malformed frames, and preserving limit errors after worker cleanup. The historical native inference/cache trial was not rerun.
• Settings persistence, editable defaults, invalid-input handling, stale-error suppression, and readable narrow-window toasts are covered. Actual Settings and toast renderings were inspected.
• make verify-build, translation/manual guards, Linux test-shard inventory, and GoLand inspections passed.
• Linux/amd64 make golden exposed two outdated invalid-drop toast baselines. Both were visually inspected and refreshed; the complete E2E render suite then passed.
• Complete local Docker verification requires a native Linux/amd64 daemon; this machine reports Linux/ARM64. GitHub CI provides the complete native platform checks.

[github-actions]

Qodana for Go

It seems all right 👌

No new problems were found according to the checks applied

💡 Qodana analysis was run in the pull request mode: only the changed files were checked
☁️ View the detailed Qodana report

Contact Qodana team

Contact us at qodana-support@jetbrains.com

• Or via our issue tracker: https://jb.gg/qodana-issue
• Or share your feedback: https://jb.gg/qodana-discussions

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-16T18:51:34.505926Z	27b4b6c	Draft marked ready
🔒 Security Review	✅ Completed 2026-09-16T18:52:30.391520Z	27b4b6c	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.