This release includes breaking changes for platform teams planning a safe upgrade.
✓ No known CVEs patched in this version
Topics
+4 more
Affected surfaces
ReleasePort's take
Light signalReleasePort Layer 1 v2.12.0 adds OpenCode OAuth support for quota tracking and corrects several Copilot/MiniMax quota calculation bugs.
Why it matters: Fixes inverted quota percentage calculations (Copilot, MiniMax) that could mislead usage monitoring; introduces OAuth‑based quota tracking for OpenCode integration, improving secure access control.
Summary
AI summaryUpdates onWatch v2.12.0, Dashboard, and opencode-codex across a mixed release.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Low |
Adds OpenCode OAuth support for ChatGPT/Codex quota tracking. Adds OpenCode OAuth support for ChatGPT/Codex quota tracking. Source: llm_adapter@2026-06-06 Confidence: high |
— |
| Feature | Low |
Makes MiniMax cards full‑width for improved readability of quotas. Makes MiniMax cards full‑width for improved readability of quotas. Source: llm_adapter@2026-06-06 Confidence: high |
— |
| Bugfix | Medium |
Fixes inverted Copilot quota percentage calculation and badge display. Fixes inverted Copilot quota percentage calculation and badge display. Source: llm_adapter@2026-06-06 Confidence: high |
— |
| Bugfix | Medium |
Fixes MiniMax quota parsing to handle percentage-based API fields. Fixes MiniMax quota parsing to handle percentage-based API fields. Source: llm_adapter@2026-06-06 Confidence: high |
— |
| Refactor | Low |
Removes unused Total Delta and Duration columns from MiniMax Cycle Overview. Removes unused Total Delta and Duration columns from MiniMax Cycle Overview. Source: llm_adapter@2026-06-06 Confidence: high |
— |
Full changelog
onWatch v2.12.0
New: OpenCode (opencode-codex) ChatGPT OAuth support (#78)
onWatch now tracks ChatGPT/Codex quota for OpenCode users who log in via OpenAI OAuth - no token copying required.
- Auto-detects
~/.local/share/opencode/auth.json(honorsOPENCODE_HOMEandXDG_DATA_HOME), parsing OpenCode's{"openai":{...}}format including its millisecondexpiresfield. - Feeds the existing Codex/ChatGPT dashboard card - same provider, same usage endpoints.
- Detection priority: Codex (
~/.codex/auth.json/CODEX_HOME) wins when both exist, then OpenCode, thenCODEX_TOKEN. - Configurable at every layer:
onwatch setup,install.sh, theOPENCODE_ENABLED=trueenv flag, and dashboard admin settings. - Refreshed tokens are written back in OpenCode's own format, so one-time-use refresh tokens stay valid.
Fixed: Copilot quota percentages inverted (#75)
Copilot utilization was computed as remaining / total instead of used / total, inverting the reading on free/limited plans and firing false Danger/Critical badges. Percentages and status badges now reflect true consumed usage (e.g. 3987/4000 completions = ~99.7% consumed).
Fixed: MiniMax percentage-based quota
MiniMax migrated its API from request-count quotas to a percentage model (current_interval_remaining_percent / current_weekly_remaining_percent). onWatch now parses the new fields correctly instead of storing zeros, closes cycles for discontinued models, and the Cycle Overview reflects the active quota.
Dashboard
- MiniMax cards are now full-width for better readability of the general/weekly quotas.
- Removed the unused Total Delta and Duration columns from the MiniMax Cycle Overview.
Full Changelog: https://github.com/onllm-dev/onWatch/compare/v2.11.46...v2.12.0
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
Related context
Related tools
Beta — feedback welcome: [email protected]