showing zero cost with latest opencode

[tan-wei-xin-alez]

I think it was working last month before opencode was updated

➜  ~ npx copilot-arewecooked --terminal
copilot-arewecooked
Period: all available data

━━ Sources ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
No Copilot usage found.

━━ Included credit comparison (monthly average) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Monthly average: 0 AI credits ($0)
Plan        Included  Remaining     %
----------  --------  ---------  ----
pro            1,000      1,000  100%
pro+           3,900      3,900  100%
business       1,900      1,900  100%
enterprise     3,900      3,900  100%

━━ Model usage and cost (total) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
No priced model usage found.
➜  ~ opencode --version
1.16.0
➜  ~ du -sh .local/share/opencode/*
4.0K    .local/share/opencode/account.json
4.0K    .local/share/opencode/auth.json
18M     .local/share/opencode/log
25M     .local/share/opencode/opencode.db
32K     .local/share/opencode/opencode.db-shm
4.0M    .local/share/opencode/opencode.db-wal
4.0K    .local/share/opencode/repos
788K    .local/share/opencode/snapshot
408K    .local/share/opencode/storage
148K    .local/share/opencode/tool-output

Just to confirm, I am using github copilot in opencode

  ┃                                                                                                                                  Test conversation
  ┃  test
  ┃                                                                                                                                  Context
                                                                                                                                     8,003 tokens
     I’m in Plan Mode (read‑only). I won’t change any files until you confirm and I leave Plan Mode. I don’t yet know your goal      3% used
     — please pick one from the list below or describe a different objective. After you pick, I’ll produce a concrete,               $0.00 spent
     step‑by‑step execution plan and ask any clarifying questions needed.
                                                                                                                                     LSP
     Choose a goal (or reply with your own):                                                                                         LSPs are disabled

     1. Fix a failing test or runtime error.
     2. Implement a new feature (describe feature).
     3. Refactor/rename a public API or function across the repo.
     4. Add or improve unit/integration tests for an area (specify area).
     5. Upgrade dependencies safely.
     6. Add or fix CI configuration (tests, linters, coverage).
     7. Triage and reduce a specific bug (paste error/stacktrace).
     8. General code review and risk report for a pull request/branch.
     9. Other — describe.

     If you pick one, tell me:

     - The repo path/branch to work against (if applicable).
     - Any hard constraints (backwards compatibility, performance, delivery date).
     - How you want verification: run unit tests locally, lint, typecheck, build artifact, or provide patch only.
     - If relevant, paste the failing log/error or link to the issue/PR.

     Example quick plan I’ll produce after you choose “Fix a failing test”:

     1. Read code and test that’s failing; reproduce locally (read-only: inspect files and test outputs you provide).
     2. Grep for related functions/types and call sites to understand scope.
     3. Propose the minimal code change with rationale and edge cases.
     4. Apply change in a single commit, run tests, fix regressions.
     5. Add a focused unit test if the bug was untested.
     6. Push branch/PR and include concise change description and risk notes.

     Tell me which goal and any details, and I’ll produce a tailored, executable plan and list the exact read‑only inspections
     I’ll perform first.

  ┃
  ┃
  ┃
  ┃  Plan · GPT-5 Mini GitHub Copilot                                                                                                ~/opencode
  ╹▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
                                                                                               8.0K (3%) · $0.00  ctrl+p commands    • OpenCode 1.16.0

[PanAchy]

Looks like OpenCode 1.16.0 introduced session-level token tracking for GitHub Copilot (tokens_input/tokens_output/etc. on the session table). The tool was only reading per-message data JSON, so sessions where tokens live only at the session level returned zero. PR #88 is addressing this.

[PanAchy]

@tan-wei-xin-alez can you test branch fix/opencode-session-level-tokens please? If that works, I can merge it in

[tan-wei-xin-alez]

Seems to work, not sure if the numbers are correct though (don't know what that negative percentage means, maybe used more than available credits? But then why is Remaining positive though)

➜  ~ git clone https://github.com/PanAchy/copilot-arewecooked.git
cd copilot-arewecooked
git checkout fix/opencode-session-level-tokens
npm install
Cloning into 'copilot-arewecooked'...
remote: Enumerating objects: 694, done.
remote: Counting objects: 100% (204/204), done.
remote: Compressing objects: 100% (154/154), done.
remote: Total 694 (delta 92), reused 92 (delta 40), pack-reused 490 (from 1)
Receiving objects: 100% (694/694), 5.57 MiB | 6.39 MiB/s, done.
Resolving deltas: 100% (373/373), done.
branch 'fix/opencode-session-level-tokens' set up to track 'origin/fix/opencode-session-level-tokens'.
Switched to a new branch 'fix/opencode-session-level-tokens'
npm warn deprecated prebuild-install@7.1.3: No longer maintained. Please contact the author of the relevant native addon; alternatives are available.

> copilot-arewecooked@0.3.5 prepare
> husky


added 259 packages, and audited 260 packages in 57s

66 packages are looking for funding
  run `npm fund` for details

1 moderate severity vulnerability

To address all issues, run:
  npm audit fix

Run `npm audit` for details.
➜  copilot-arewecooked git:(fix/opencode-session-level-tokens) ✗ npm run build:css
npm run dev -- --terminal

> copilot-arewecooked@0.3.5 build:css
> node scripts/build-report-css.mjs

≈ tailwindcss v4.2.4

Done in 87ms

> copilot-arewecooked@0.3.5 dev
> tsx src/cli.ts --terminal

copilot-arewecooked
Period: all available data

━━ Sources ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Tool      Calls      Tokens  Credits
--------  -----  ----------  -------
OpenCode  1,707  96,289,776  2,111.8

━━ Included credit comparison (monthly average) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Monthly average: 1,055.9 AI credits ($10.559)
Plan        Included  Remaining      %
----------  --------  ---------  -----
pro            1,000       55.9  -5.6%
pro+           3,900    2,844.1  72.9%
business       1,900      844.1  44.4%
enterprise     3,900    2,844.1  72.9%

━━ Model usage and cost (total) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Model       Calls      Input  Cache read  Cache write     Output  Credits  Share
----------  -----  ---------  ----------  -----------  ---------  -------  -----
gpt-5.2       763  3,292,865  38,030,720            0    167,435  1,071.5  50.7%
gpt-5.5       101    493,356  10,662,912            0     42,546    694.7  32.9%
gpt-5-mini    843  5,617,921  37,118,336            0    863,685    345.6  16.4%
Total       1,707  9,404,142  85,811,968            0  1,073,666  2,111.8   100%

[PanAchy]

Thanks for testing! The negative percentage was a display bug — the old % column showed remaining ratio (negative = over budget) while Remaining hid the negative sign with Math.abs. Fixed in the same PR: Remaining now shows negative values directly when over, and the column is renamed to Used % (so 105.6% means you're 5.6% over the pro limit).

You're also welcome to try the visual report just run without --terminal and it'll write an HTML report.

Overall, I'd say numbers look correct: ~1,707 Copilot calls with 2,111.8 credits total vs 1,000 included on pro, so you'd be cooked on that plan but safe on pro+ (27.1% used).

I'll merge the PR now.