feat(ai): show run cost and balance in credits in the response bar

[ndemianc]

The response bar read "Opus 4.8 · $0.46 · $12.79 left"; it now reads "Opus 4.8 · 46 credits · 1,279 left". $1 = 100 credits — display only, and it matches the account dashboard exactly because the gateway already sends the same RETAIL micro-$ figures the website renders.

money() is replaced by two helpers off one conversion (MICROS_PER_CREDIT), so the editor and the website can never disagree about the same balance:

• creditBalance() — whole credits, the number people plan around.
• creditCost() — keeps one decimal below 10, because a cheap-model turn costs ~0.4 credits and rounding it to "0" would read as free. The old money() had the same problem and solved it with "<$0.01".

Verified against the real values behind the shipped UI: $12.79 → "1,279", $0.46 → "46", a $0.003577 gpt-oss turn → "0.4", $0.5117 Opus turn → "51", $100 → "10,000", and null/zero → "0". Gate: 22 suites, 0 failures.

[Copilot]

Pull request overview

This PR updates the LevelCode AI chat webview’s response/action bar to display run cost and remaining balance in credits (derived from retail micro-$), aligning the editor UI with the account dashboard’s credit-based presentation.

Changes:

• Replaces the old money(micros) formatter with credit-based conversion/formatting helpers (MICROS_PER_CREDIT, toCredits, creditBalance, creditCost).
• Updates the response/action bar metadata to render X credits for run cost and Y left for remaining balance.

Comments suppressed due to low confidence (1)

extensions/levelcode-ai/media/chat.html:2439

• This comment still references "$cost"/"$left", but the action bar now displays credits. Keeping it in sync helps prevent regressions during future UI changes.

  // Response action bar: (Retry) (Copy) (Helpful) (Unhelpful) … model · $cost · $left.

---

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

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (5)

extensions/levelcode-ai/media/chat.html:2423

• The comment still describes the old $ formatting ("$0.00" / "<$0.01"), but this section now formats values as credits. Updating it will prevent future confusion when editing/validating credit rendering.

  // [LevelCode] Credits are the in-product spending unit: $1 = 100 credits, so 1 credit = 10_000
  // retail micro-$. The gateway sends RETAIL micro-$ (Levelcode.retail_micros) — the same figures the

extensions/levelcode-ai/media/chat.html:2432

• creditBalance() can render a negative value if the gateway ever reports a negative/invalid balance (previously money() clamped <=0 to "$0.00"). Clamping to >=0 keeps the UI stable and avoids showing "-1 left".

    const c = Number(micros) / MICROS_PER_CREDIT;
    return isFinite(c) ? c : 0;
  }
  // A balance reads as a whole credit count. Clamped at zero: an overage can push the remaining
  // balance negative, and "-2 left" is a worse thing to put in front of someone than "0 left" (the

extensions/levelcode-ai/media/chat.html:2443

• This comment still references "$cost"/"$left", but the response bar now displays credits. Please update the comment so it matches the rendered UI.

    const c = toCredits(micros);

extensions/levelcode-ai/media/chat.html:2436

• creditBalance()/creditCost() are pure formatting helpers and are easy to regress (rounding thresholds, "<0.1" behavior, etc.). Since this repo already unit-tests functions extracted from media/chat.html, please add a small test covering the examples listed in the PR description (e.g. $12.79→"1,279", $0.46→"46", $0.003577→"0.4", null/0→"0").

  // dollar formatter this replaced clamped the same way).
  function creditBalance(micros){
    const c = toCredits(micros);
    return (c > 0 ? Math.round(c) : 0).toLocaleString();

extensions/levelcode-ai/media/chat.html:2464

• credits is still retail micro-$ (converted to credits for display), so the variable name here is misleading and can cause confusion when editing the formatter logic. Consider renaming the local parameter in addAgentDone() to something like creditsMicros and using that name throughout the function.

    const meta = document.createElement('div'); meta.className = 'agentbar-meta';

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (1)

extensions/levelcode-ai/media/chat.html:2446

• creditCost() uses toLocaleString() without an explicit locale for costs >= 10 credits, which makes the formatting locale-dependent and can desync from the fixed comma-separated examples/tests. Using an explicit locale keeps the response bar stable across environments.

    const c = toCredits(micros);
    if (c <= 0) { return '0'; }
    if (c >= 10) { return Math.round(c).toLocaleString(); }
    return c < 0.05 ? '<0.1' : c.toFixed(1);

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (1)

extensions/levelcode-ai/media/chat.html:2459

• In addAgentDone(), the parameter named credits actually carries the remaining balance in retail micro-$ (per the comment below and the agentDone message field), not a credit count. Now that the UI renders credits, this name is especially misleading—rename the parameter to reflect its unit (e.g. creditsMicros) to avoid future mixups.

  function addAgentDone(reason, edits, credits, maxSteps, costMicros){
    closeGroup();   // the run is over — finalize any open activity group
    const bar = document.createElement('div'); bar.className = 'agentbar';
    // Stamp the model that produced THIS turn, so a later reaction is attributed to it
    // (not whatever the plan is entitled to at click time — models can change mid-session).