Fix 1% usage reading as 100% across providers - #198
Merged
Conversation
📝 WalkthroughWalkthroughChangesQuota percentage scaling
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant Sakana quota extraction
participant detect_fraction_scale
participant to_percent
participant RateWindow
Sakana quota extraction->>detect_fraction_scale: raw JSON percentage values
detect_fraction_scale-->>Sakana quota extraction: fraction scale decision
Sakana quota extraction->>to_percent: raw value and scale decision
to_percent-->>Sakana quota extraction: clamped percentage
Sakana quota extraction->>RateWindow: normalized usage window
Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
Contributor
Deploying with
|
| Status | Name | Latest Commit | Preview URL | Updated (UTC) |
|---|---|---|---|---|
| ✅ Deployment successful! View logs |
ceiling | 1065631 | Commit Preview URL Branch Preview URL |
Aug 03 2026, 04:04 AM |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR fixes a cross-provider regression where a reported usage value of 1 (meaning 1% on whole-percent APIs) was being interpreted as a fraction and displayed as 100%, by resolving percent scale once per response and normalizing values consistently.
Changes:
- Introduces shared percent-scale detection/conversion helpers (
detect_fraction_scale,to_percent) inrust/src/core. - Updates multiple providers (OpenCode, OpenCode Go, OpenCode scraper, Qoder, Chutes, Sakana) to normalize percent values using response-wide evidence rather than per-window
<= 1.0heuristics. - Adds regression tests for the “1% reads as 100%” case and documents the fixes in the changelog.
Reviewed changes
Copilot reviewed 9 out of 9 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| rust/src/providers/sakana/mod.rs | Distinguishes literal % text from JSON percent keys; applies evidence-based scaling only to ambiguous JSON values. |
| rust/src/providers/qoder/mod.rs | Detects percent scale from payload and applies consistent normalization when building credit windows. |
| rust/src/providers/opencodego/mod.rs | Resolves percent scale across windows before converting rolling/weekly/monthly usage values; adds regression tests. |
| rust/src/providers/opencode/scraper.rs | Normalizes rolling/weekly usage by detecting scale from both windows before parsing. |
| rust/src/providers/opencode/mod.rs | Normalizes rolling/weekly window percents using response-wide scale detection and percent-key provenance. |
| rust/src/providers/chutes/mod.rs | Detects scale from payload percent fields and normalizes quota windows consistently; adds regression test. |
| rust/src/core/percent_scale.rs | Adds shared, evidence-based percent scale detection and conversion helpers with unit tests. |
| rust/src/core/mod.rs | Exposes the new percent_scale module via the core public re-exports. |
| CHANGELOG.md | Documents the regression and the provider-specific fixes. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+269
to
276
| fn find_usage_window(&self, json: &Value, keys: &[&str]) -> Option<(f64, i64, bool)> { | ||
| for key in keys { | ||
| if let Some(obj) = json.get(key) | ||
| && let Some(window) = self.parse_window(obj) | ||
| && let Some((percent, from_percent_key)) = Self::window_percent(obj) | ||
| { | ||
| return Some(window); | ||
| let reset_sec = Self::window_reset_seconds(obj).unwrap_or(0); | ||
| return Some((percent, reset_sec, from_percent_key)); | ||
| } |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@rust/src/providers/sakana/mod.rs`:
- Around line 154-176: Update extract_raw_window so the calculated end index is
advanced to the next valid UTF-8 character boundary after applying the 1400-byte
limit and text length cap, before slicing text[anchor.0..end]. Preserve the
existing anchor and extraction behavior.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: c4e5b3f6-2092-4954-9bd3-4488f968f291
📒 Files selected for processing (9)
CHANGELOG.mdrust/src/core/mod.rsrust/src/core/percent_scale.rsrust/src/providers/chutes/mod.rsrust/src/providers/opencode/mod.rsrust/src/providers/opencode/scraper.rsrust/src/providers/opencodego/mod.rsrust/src/providers/qoder/mod.rsrust/src/providers/sakana/mod.rs
Comment on lines
+154
to
176
| fn extract_raw_window(text: &str, labels: &[&str]) -> Option<RawWindow> { | ||
| let lower = text.to_ascii_lowercase(); | ||
| let anchor = labels | ||
| .iter() | ||
| .find_map(|label| lower.find(label).map(|idx| (idx, *label)))?; | ||
| let end = (anchor.0 + 1400).min(text.len()); | ||
| let segment = &text[anchor.0..end]; | ||
| let percent = extract_percent(segment)?; | ||
| let (percent, from_json) = extract_percent(segment)?; | ||
| let reset = extract_reset(segment); | ||
| let mut window = RateWindow::with_details( | ||
| let reset_description = extract_reset_text(segment); | ||
| let minutes = if anchor.1.contains("week") { | ||
| Some(7 * 24 * 60) | ||
| } else { | ||
| Some(5 * 60) | ||
| }; | ||
| Some(RawWindow { | ||
| percent, | ||
| if anchor.1.contains("week") { | ||
| Some(7 * 24 * 60) | ||
| } else { | ||
| Some(5 * 60) | ||
| }, | ||
| from_json, | ||
| reset, | ||
| None, | ||
| ); | ||
| if let Some(reset_text) = extract_reset_text(segment) { | ||
| window.reset_description = Some(reset_text); | ||
| } | ||
| Some(window) | ||
| reset_description, | ||
| minutes, | ||
| }) | ||
| } |
There was a problem hiding this comment.
🩺 Stability & Availability | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -eu
printf '%s\n' '--- target source ---'
sed -n '1,340p' rust/src/providers/sakana/mod.rs
printf '%s\n' '--- toolchain declarations ---'
fd -a -t f '^(rust-toolchain|rust-toolchain\.toml|Cargo\.toml)$' . -x sh -c 'echo "--- $1"; sed -n "1,120p" "$1"' sh {}
printf '%s\n' '--- relevant usages ---'
rg -n -C 3 'extract_raw_window|snapshot_from_html|is_char_boundary|floor_char_boundary|ceil_char_boundary' rust/src rust/tests 2>/dev/null || true
printf '%s\n' '--- standalone UTF-8 boundary probe ---'
python3 - <<'PY'
text = "reset — usage"
start = text.index("reset")
end = min(start + 8, len(text.encode("utf-8")))
raw = text.encode("utf-8")
print("text:", repr(text))
print("byte length:", len(raw), "computed end:", end)
print("byte at end:", raw[end:end+1].hex())
print("is UTF-8 boundary:", end == 0 or end == len(raw) or (raw[end] & 0xc0) != 0x80)
PYRepository: tsouth89/ceiling
Length of output: 17695
🏁 Script executed:
#!/bin/bash
set -eu
sed -n '145,185p' rust/src/providers/sakana/mod.rs
fd -a -t f '^(rust-toolchain|rust-toolchain\.toml|Cargo\.toml)$' . -x sh -c 'echo "--- $1"; sed -n "1,80p" "$1"' sh {}Repository: tsouth89/ceiling
Length of output: 4523
🏁 Script executed:
#!/bin/bash
set -eu
printf '%s\n' '--- available Rust toolchain ---'
if command -v rustc >/dev/null 2>&1; then
rustc --version
else
echo 'rustc unavailable'
fi
printf '%s\n' '--- standalone Rust slicing probe ---'
if command -v rustc >/dev/null 2>&1; then
tmp="$(mktemp -d)"
trap 'rm -rf "$tmp"' EXIT
cat >"$tmp/probe.rs" <<'RS'
fn main() {
let text = format!("5-hour quota 1% {}", "x".repeat(1370)) + "—";
let anchor = text.find("5-hour").unwrap();
let end = (anchor + 1400).min(text.len());
assert!(!text.is_char_boundary(end), "probe must end inside UTF-8");
let _segment = &text[anchor..end];
}
RS
rustc "$tmp/probe.rs" -o "$tmp/probe"
if "$tmp/probe"; then
echo 'unexpected: probe did not panic'
exit 1
else
echo 'expected: slicing at the computed non-boundary panicked'
fi
fiRepository: tsouth89/ceiling
Length of output: 243
🏁 Script executed:
#!/bin/bash
set -eu
command -v rustc >/dev/null 2>&1 && rustc --version || echo 'rustc unavailable'Repository: tsouth89/ceiling
Length of output: 172
Fix the UTF-8 boundary before slicing.
If end is not a character boundary, advance it to the next boundary before slicing. Otherwise, multi-byte HTML content can cause a runtime panic.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@rust/src/providers/sakana/mod.rs` around lines 154 - 176, Update
extract_raw_window so the calculated end index is advanced to the next valid
UTF-8 character boundary after applying the 1400-byte limit and text length cap,
before slicing text[anchor.0..end]. Preserve the existing anchor and extraction
behavior.
Merged
tsouth89
added a commit
that referenced
this pull request
Aug 3, 2026
## Summary Prepares **Ceiling 1.5.22**, carrying three changes merged since v1.5.21: - **#198** - Fix 1% usage reading as 100% across providers (OpenCode Go, OpenCode, Qoder, Chutes, Sakana) - **#197** - OpenCode Go monthly bar labeled "Monthly" instead of "Extra" - **#196** - Fix Microsoft Store installer parameters (40-char Partner Center limit) Bumps all version sources to 1.5.22 (build 124), moves the Unreleased CHANGELOG section into a versioned 1.5.22 entry (adding the missing #196/#197 notes), and adds `.github/release-notes-1.5.22.md`. ## Changes - `CHANGELOG.md` - retitle Unreleased to `[Ceiling] 1.5.22 - 2026-08-03`, add #196/#197 entries - `version.env` - `MARKETING_VERSION=1.5.22`, `BUILD_NUMBER=124` - `rust/Cargo.toml`, `apps/desktop-tauri/src-tauri/Cargo.toml`, `apps/desktop-tauri/package.json`, `apps/desktop-tauri/src-tauri/tauri.conf.json`, `Cargo.lock` - version 1.5.22 - `.github/release-notes-1.5.22.md` - new release notes ## Validation `powershell.exe -ExecutionPolicy Bypass -NoProfile -File scripts\local-check.ps1 -All -Version 1.5.22` passes, including the release doctor for 1.5.22 (all version sources consistent, CHANGELOG mentions 1.5.22) and the Store submission preparation test. Expected pre-tag warnings only: local tag v1.5.22, local release assets, and GitHub release not found yet. After this merges and checks are green, push annotated tag `v1.5.22` to trigger the `Signed Windows Release` workflow. Co-authored-by: tsouth89 <tsouth89@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Fixes the 1%-reads-as-100% usage bug for providers that report usage as either whole percentages (
23= 23%) or fractions of a limit (0.23= 23%). A lone1is ambiguous in those scales, and the old per-window<= 1.0 -> *100rule turned the first 1% of use into a maxed-out window.This is the same ambiguity already fixed for Claude in #186 and OpenCode Go (this branch's sibling fix). This change resolves the scale once per response from real evidence, and only reads fractions when a window actually holds a fractional value.
Behavior changes
Affected providers, all with the same bug:
usagePercentper window; ausagePercent: 1rolling window rendered as 100% used.usedPercentper window.usage_percentper window."1% used"as 100%. Literal%values are never scaled now; only JSON percent keys use the evidence-based scale.Checked and intentionally left alone: Devin (
< 1.0boundary already treats1as 1%), NanoGPT (API always reports fractions), Codex/LLMProxy/Copilot (read percents raw, no threshold).The rule, shared in
rust/src/core/percent_scale.rs:1.0can only be a percentage -> whole-percent scale;0and1can only be a fraction -> fraction scale;0or1, read as whole percents (so1= 1%, not 100%).Each provider feeds only its ambiguous percent-key values into the scale detection and scales only those;
used/limitratios are never rescaled.Validation
cargo test --manifest-path rust/Cargo.toml: 789 passed. (Note: fourcost_scannertests flake under parallelcargo testdue to a pre-existing env-var race betweencredentials_storetests andscan_claude; they pass in isolation and on cleanmain.)cargo test --manifest-path apps/desktop-tauri/src-tauri/Cargo.toml: 464 passed.cargo fmt --allclean;cargo clippy --all-targets -- -D warningsclean on both manifests.Summary by CodeRabbit
0.5now correctly display as 50%, while values such as1display as 1% when appropriate.1%values.