Improve URL metrics, resilient hash decoding, and file:// share URL handling

[HKTITAN]

Motivation

• Ensure the status bar, QR, and share/copy flows reflect the actual saved share URL and size rather than stale location.href during edits.
• Make hash decoding more tolerant so legacy/base85-encoded or mis-detected hashes don't break loading.
• Preserve correct share URLs when running from local file:// origins where location.origin === 'null'.

Description

• Added a lightweight cached URL metrics model with contentVersion and urlMetrics plus helpers buildShareUrl, refreshUrlMetrics, syncUrlMetricsToLocation, and scheduleUrlMetricsUpdate to compute and cache the share URL and its byte size.
• Updated UI flows to prefer cached metrics when available by changing updateStatusBar, updateQR, tooltip logic, and isAtInputLimit to use urlMetrics when urlMetrics.version === contentVersion.
• Bumped contentVersion in scheduleSave and other lifecycle points and updated save() to compute and store urlMetrics after saving, so share/QR/clipboard reflect the latest saved URL.
• Made decompressToBlocks resilient by attempting decoding in both base85 and base64 orders (depending on detection), retrying the other format on failure, and preserving encrypted errors (err.type === 'ENCRYPTED') so they propagate unchanged.
• Fixed share URL construction for file:///local origins by preserving the local base path instead of returning an invalid null origin, using normalizeHash and an encoded hash suffix.
• Added debounce utility and wired scheduleUrlMetricsUpdate (350ms) to avoid expensive recomputation during typing, and synced metrics to location changes in theme/version/preview/encryption flows.

Testing

• Attempted an automated end-to-end check by starting a local static server (python -m http.server) and running a Playwright script to type and inspect the UI, but the Playwright run timed out after 30s and did not complete (failure); no further automated tests were present.
• No unit tests were added; existing code paths were exercised via local server runs and static inspection during the change.

---

Codex Task

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
khemessage	Ready	Preview, Comment	Feb 3, 2026 6:57am

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b8fdd3f56a

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Use live content size for input limit enforcement

The new isAtInputLimit checks only urlMetrics (when versions match) or location.href (when they don’t). During active typing/pasting, the URL hash hasn’t been updated yet, so location.href reflects the previous saved state and urlMetrics is stale until the debounced recomputation finishes. This means the limit checks in handleBeforeInput/handlePaste can allow users to type or paste well past the URL threshold before the limit flips, which defeats the size guard that previously used a live estimate from blocks. Consider reverting to (or incorporating) estimateUrlSizeFromBlocks for enforcement so the limit is based on the current content, not the last saved URL.

Useful? React with 👍 / 👎.