anomalyco / anomalyco/opencode
go: monthly quota 100% while all local stores total ~$5 of usage
@MrMushrooooom is already working on this.
Since Sep 16, 2026.
- Dominant language
- TypeScript
- Stars
- 209k
- Forks
- 27.5k
- PR merge metrics
- PR metrics pending
Description
Summary
Go monthly quota shows 100% (rate-limited) while the server usage API AND every
local data store on the machine total ~$5 of paid usage in the current window —
a ~$55 discrepancy with no identifiable source.
Environment
- OpenCode version: v2.0.4 (@opencode/cli, active install serving sessions).
A legacy opencode-ai v1.18.30 binary also exists on PATH (second/older
install); v2 is what's actively used. - OS: Windows 11 (10.0.22631, win32 x64) + WSL2 Ubuntu (separate opencode v2
install, same Go key) - Terminal: xterm-256color / truecolor
- Shell: cmd.exe (Windows); bash (WSL)
- Install/channel: latest
- Active plugins: @plannotator/opencode@latest, @upstash/context7-opencode@latest
What Happened
This month I hit the Go monthly limit (100%, rate-limited, reset
2026-09-30T05:23:21Z per GET https://opencode.ai/zen/go/v1/usage) — but every
usage number I can see says I've used only a fraction of the allowance:
- Local Windows opencode.db (Aug 31 05:23 UTC → now): $4.84 total across
opencode-go models (deepseek-flash $1.63, glm-5.3-flash $1.44, omen-alpha
$1.04, mimo-v2.5 $0.62, hy3 $0.01), plus $0.09 on OpenRouter and $0 on free
models. - WSL opencode.db (same key, separate v2 install): $0.00 paid usage in the
window — free models only. - No deepseek-v4-pro at all since 2026-08-12. The $51.61 recorded against
it is entirely May–Aug usage; a date-bucketed query confirms zero rows in
the current monthly window. - Cache reads are large (e.g. 166.4M on deepseek-flash, 35.6M on
glm-5.3-flash) but bill in the $0.50–$1.10 range at published Go rates.
Aggregate server-billed value of everything I can account for is ~$5–8 of the
$60 monthly allowance, yet the server reports 100% and blocking. The missing
~$52 has no source in any store I can inspect: one device, one workspace, one
key across two installs.
Expected Behavior
Monthly percent should roughly match the sum of server-billed usage for the
models I actually ran (~$5–8), or the usage history should expose where the
remaining ~$52 was attributed.
Actual Behavior
Server reports 100% monthly / rate-limited despite the ~$5 of verifiable
usage above.
Additional Context
Same symptom class as #42686, #38936, #42528, #40280, #41976, #41391, #41206,
but this report has the authoritative usage-API output and full local-store
cross-checks (Windows DB + WSL DB). Willing to run any diagnostic or share
redacted account/usage history.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.