fix(ai): opendesign quota parseCodingPlanWindows publishes out-of-range remaining raw — "-50 / 100 credits" at 100% used, "150 / 100 credits" at 0% used

Verified 2026-11-06 on origin/main 90453b467f (merge of PR #1874), in a laptop worktree (~/git/probe-2026-11-06-creator-1235, kept). Gates this run: check:workflows, check:gates, check:doc-env-vars GREEN; check:bounded-http 7 / check:numeric-boundaries 3 / check:proto-reads 4 RED = exactly #1724's set — no new drift.

The defect

packages/ai/src/quota/providers/opendesign.ts:78-107 (parseCodingPlanWindows) guards the limit (limit <= 0 entries are dropped, :94 — the #1750 disposition) but never range-checks remainingCredits:

:99 clamps the derived percent into [0, 100]: Math.min(100, Math.max(0, (1 - remaining / limit) * 100))
:104 publishes the raw remaining in the label: `${remaining} / ${limit} credits`

Percent and label therefore derive from different figures and can contradict each other — the exact failure the window contract's rule 5 (whole-window clamp, #1821, implemented for command-code) exists to prevent: "percent and label derive from one clamped used figure, so they can never contradict each other."

Runtime probe (laptop worktree, `probe-quota-caps.ts`, kept in the worktree)

Against the real parseCodingPlanWindows imported from main, one 5h window (durationSeconds: 18000, limitCredits: 100, resetsAt: 1893456000):

remainingCredits: -50 → window published: usedPercent: 100, valueLabel: "-50 / 100 credits" — a negative remaining balance rendered to the user.
remainingCredits: 150 → window published: usedPercent: 0, valueLabel: "150 / 100 credits" — 0% used alongside a label claiming 150 of 100 credits remain.
Control remainingCredits: 25 → usedPercent: 75, valueLabel: "25 / 100 credits" (correct).

Why it matters

The label is the user-facing figure in the opencode quota output. A negative-remaining label is nonsense on its face, and the over-limit case silently hides a real state: command-code's monthly window hit the same over-included-remaining shape and dispositions it explicitly (#1744: publish at 0% used with the real remaining figure annotated, ${used} / ${included} (${remaining} remaining)), because rollover/promo/higher-tier remaining above the nominal limit is information, not noise. OpenDesign currently gets the worst of both: the percent follows the clamp while the label follows the raw wire, with no annotation explaining the discrepancy.

Acceptance criteria

Percent and valueLabel derive from one consistent disposition of an out-of-range remainingCredits — they can never contradict (the fix(ai): command-code quota windows split their out-of-range disposition — percent clamped, label raw; non-positive resetAt published as a 1969/1970 date #1821 rule-5 shape).
Negative remaining never renders as a negative figure in the label (clamp to 0 remaining / 100% used, or an equally explicit disposition).
Over-limit remaining follows an explicit disposition — the fix(ai): Command Code quota monthly window vanishes when remaining credits exceed the hard-coded 10 included — an over-limit payload reads as 'could not be parsed' #1744 annotation shape (0% used with the real remaining named) or a clamped label — chosen in the fix and pinned by a test; not the current silent raw label.
Tests pin all three cases (negative, over-limit, in-range control) for parseCodingPlanWindows; the existing quota suite still passes.

🤖 Agent note · filed

Agent: new-issue-creator job (new-issue-creator-opencode)
Session / ID: cron new-issue-creator-opencode run 2026-11-06T18:35Z
Found via: filed this run
Task: Filed the opendesign parseCodingPlanWindows out-of-range-remaining defect: percent is clamped but the valueLabel publishes remainingCredits raw, so a negative remaining renders "-50 / 100 credits" at 100% used and an over-limit remaining renders "150 / 100 credits" at 0% used. Verified with a laptop probe against the real parser on main 90453b4. Enqueued as a triage packet — the over-limit disposition (fix(ai): Command Code quota monthly window vanishes when remaining credits exceed the hard-coded 10 included — an over-limit payload reads as 'could not be parsed' #1744 annotation vs plain clamp) is the worker's call.
Expectation: effort/s — one consistent disposition for out-of-range remaining, pinned by tests for the negative, over-limit, and in-range cases.
Links: probe probe-quota-caps.ts in laptop worktree ~/git/probe-2026-11-06-creator-1235 (kept)

🤖 Agent note · adopted as design phase 3 by repo design architect (opencode pass 2026-11-07)

Agent: repo-design-architect
Task: This issue is adopted as phase 3 of design design(ai): a figure that cannot exist is not a figure — one figure contract for the quota parsers' amounts, balances, counts and percents #1942 ("a figure that cannot exist is not a figure" — one figure contract for the quota parsers' amounts, balances, counts and percents) and linked as a sub-issue of design(ai): a figure that cannot exist is not a figure — one figure contract for the quota parsers' amounts, balances, counts and percents #1942. Its acceptance already names the contract's disposition; phase 1 (test(ai): quota figure contract — figure inventory + shared figure probe matrix for the 13 quota providers (design #1942, phase 1) #1943) records this site as a known-lossy inventory row until this issue lands and flips it.
Queue: unchanged — this issue keeps its own queue lifecycle; it was NOT re-enqueued by the design pass.
Links: design(ai): a figure that cannot exist is not a figure — one figure contract for the quota parsers' amounts, balances, counts and percents #1942 · test(ai): quota figure contract — figure inventory + shared figure probe matrix for the 13 quota providers (design #1942, phase 1) #1943

🤖 Agent note · picked up

Agent: repo-worker-opencode (OpenChamber scheduled lane)
Session / ID: ses_e5e468b38a6ffezBpdzAu38uDA · lane 3fe37e8a-c61b-419c-9e97-769ecd264d5d
Found via: queue repo-health-opencode-pilot, task 11859
Task: Revalidate issue fix(ai): opendesign quota parseCodingPlanWindows publishes out-of-range remaining raw — "-50 / 100 credits" at 100% used, "150 / 100 credits" at 0% used #1879's remainder against current main. The issue is open; PR docs(ai): update quota inventory to current parser state #1963 is merged documentation/inventory work under docs(ai): quota window-contract inventory verdicts do not reflect the open known-lossy filings — six rows certify or omit parsers with live lossy issues, and turso's row has no verdict at all #1939 and does not fix the out-of-range remaining-credit calculation.
Links: fix(ai): opendesign quota parseCodingPlanWindows publishes out-of-range remaining raw — "-50 / 100 credits" at 100% used, "150 / 100 credits" at 0% used #1879 · docs(ai): update quota inventory to current parser state #1963

🤖 Agent note · triaged

Agent: repo-worker-opencode (OpenChamber scheduled lane)
Session / ID: ses_e5e468b38a6ffezBpdzAu38uDA · lane 3fe37e8a-c61b-419c-9e97-769ecd264d5d
Task: Revalidated on current main 24200c3. The remainder is actionable: parseCodingPlanWindows publishes -50 / 100 credits at 100% used, and 150 / 100 credits at 0% used; the 25 / 100 in-range control is consistent. Current quota.test.ts marks the over-limit label known-lossy. PR docs(ai): update quota inventory to current parser state #1963 is merged inventory/docs work under docs(ai): quota window-contract inventory verdicts do not reflect the open known-lossy filings — six rows certify or omit parsers with live lossy issues, and turso's row has no verdict at all #1939, not the fix(ai): opendesign quota parseCodingPlanWindows publishes out-of-range remaining raw — "-50 / 100 credits" at 100% used, "150 / 100 credits" at 0% used #1879 fix.
Disposition: Created scoped small_fix queue task 12850 (repo-health:v1:fix:developing-today-forks/opencode:issue:1879). The packet selects clamping remainingCredits to [0, limitCredits] once and deriving both usedPercent and valueLabel from it; this matches an explicit disposition allowed by the issue.
Verification: Direct parser probes reproduced negative, over-limit, and in-range cases. The targeted command bun run --cwd packages/ai test quota.test.ts could not load the suite in this fresh worktree because package effect is missing; 0 tests executed.
Links: fix(ai): opendesign quota parseCodingPlanWindows publishes out-of-range remaining raw — "-50 / 100 credits" at 100% used, "150 / 100 credits" at 0% used #1879 · docs(ai): update quota inventory to current parser state #1963

Merge pull request #2101 from developing-today-forks/queue/12850-1879…

Navigation Menu

Repository navigation

fix(ai): opendesign quota parseCodingPlanWindows publishes out-of-range remaining raw — "-50 / 100 credits" at 100% used, "150 / 100 credits" at 0% used #1879

fix(ai): opendesign quota parseCodingPlanWindows publishes out-of-range remaining raw — "-50 / 100 credits" at 100% used, "150 / 100 credits" at 0% used

The defect

Runtime probe (laptop worktree, `probe-quota-caps.ts`, kept in the worktree)

Why it matters

Acceptance criteria

9 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

fix(ai): opendesign quota parseCodingPlanWindows publishes out-of-range remaining raw — "-50 / 100 credits" at 100% used, "150 / 100 credits" at 0% used #1879

Description

fix(ai): opendesign quota parseCodingPlanWindows publishes out-of-range remaining raw — "-50 / 100 credits" at 100% used, "150 / 100 credits" at 0% used

The defect

Runtime probe (laptop worktree, probe-quota-caps.ts, kept in the worktree)

Why it matters

Acceptance criteria

Activity

dezren39 commented on Nov 6, 2026

dezren39 commented on Nov 7, 2026

9 remaining items

developing-today commented on Nov 8, 2026

developing-today commented on Nov 8, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Runtime probe (laptop worktree, `probe-quota-caps.ts`, kept in the worktree)