Thread details shows the local Claude login's usage for a Claude instance routed through a custom ANTHROPIC_BASE_URL
Mantenedores costumam responder em até 1 dia
Avaliação
- Dificuldade
- 2/5
- Tempo estimado
- 1-3 horas
- Facilidade para iniciantes
- 35/100
- Tipo de issue
- Bug
- Clareza
- Claramente especificada
- Status de atividade
- Estagnada
- Stack de tecnologia
- typescript
- Domínio
- backend-api-design
Direção de pesquisa
Read apps/server/src/provider/ClaudeProvider.ts where customApiEndpoint is computed and usageLimits is built from capabilities.usage; compare how the custom endpoint is represented with the proposed unavailable reason. Check apps/web/src/components/chat/ThreadDetailsUsage.tsx to confirm the displayed fallback behavior. Done means custom endpoint threads no longer show the local OAuth account’s limits; an open pull request is already linked.
Escrita pelo modelo de indexação a partir do texto da issue.
Descrição
What happened
Thread details → Usage limits always shows Weekly 100% and Weekly · Fable 100% (full bars, % left) and no Session row, in every Claude thread. It never changes as I use Claude. The main limits view shows real, non-zero usage for the accounts that actually serve my requests.
Diagnosis
My Claude requests go through a CLIProxyAPI hub: ~/.claude/settings.json sets env.ANTHROPIC_BASE_URL to the hub. The claude CLI is also signed in with an OAuth Claude subscription that the hub doesn't use for these requests.
ClaudeProvider.ts sees the non-Anthropic base URL and marks the instance as customEndpoint ("First-party account fields cannot verify custom endpoint credentials"). It still publishes capabilities.usage (the CLI's get_usage result) as the instance's usageLimits. get_usage reports the local OAuth login, not the account the custom endpoint routes to. That login is unused, so the instance gets valid-looking limits at 0% used:
"five_hour": null,
"seven_day": { "utilization": 0, "resets_at": "2026-11-26T06:00:00Z" },
"model_scoped": [{ "display_name": "Fable", "utilization": 0 }]
(claude 2.1.329, get_usage control request, captured while the hub accounts were in active use.)
ThreadDetailsUsage renders only the instance's own snapshot, so it shows 100% left for Weekly and Weekly · Fable and drops Session (five_hour: null). The streamed rate_limit_events through the proxy carry no numbers ({"status":"allowed","isUsingOverage":false}), so nothing corrects it during a turn.
This differs from #30990. There the CLI has no OAuth login, so the probe fails and the card says Unavailable. Here the CLI returns a different account's numbers, which look valid. The fallback proposed in #27740 runs only when the instance's limits carry a notice, so it would not trigger here.
Suggested fix: when the Claude instance resolves to customEndpoint, don't attribute the CLI's get_usage result to it. Publish unavailable limits with a reason, so the card reads as unavailable or falls back to the usage hub as in #27740, instead of showing another account's numbers.
Steps to reproduce
- Sign the
claudeCLI in to Claude subscription account A (claude auth login). - In
~/.claude/settings.json, setenv.ANTHROPIC_BASE_URLto a CLIProxyAPI hub that serves Claude through other accounts. (My config also sets_CLAUDE_CODE_ASSUME_FIRST_PARTY_BASE_URL=1; I haven't tested without it.) - Use the default
claudeAgentinstance, with no T3-level env overrides. Add the hub as acliproxyusage limit source. - Use Claude in a few threads, then open Thread details with Usage limits enabled.
Expected: the limits of the account serving the requests, or Unavailable.
Actual: account A's limits (Weekly 100%, Weekly · Fable 100%, no Session). They never move.
Version
0.1.11-nightly.20261120.4979 (desktop app and local server). Source read at main @ df09ced7.
Environment
macOS (Darwin 27.0), T3 Code (Nightly) desktop app against its local server, Claude Code CLI 2.1.329, CLIProxyAPI hub on the LAN configured as a cliproxy usage limit source.
Evidence
- Thread details card:
Weekly ▬▬▬ 100%,Weekly · Fable ▬▬▬ 100%, no Session row. get_usageoutput above (trimmed).apps/server/src/provider/ClaudeProvider.ts:customApiEndpointis computed, thenusageLimitsis still built fromcapabilities.usage.apps/web/src/components/chat/ThreadDetailsUsage.tsx: uses the account matchinginstanceId, elseprovider.usageLimits.
Related issues
#30990 (same setup, but the card shows Unavailable), #27740 (hub fallback, which only runs when the limits carry a notice), #36297.
Workaround
None in T3. Use Usage → Limits for the hub accounts' real numbers.
Produced by Opus 5.5 (Claude Code) following the T3 Code triage playbook.
- Linguagem predominante
- TypeScript
- Estrelas
- 24.8k
- Forks
- 6.4k
- Merge médio
- 4h 29min
- PRs com merge (30d)
- 488
Preparar o ambiente
Inicia o contêiner de desenvolvimento do projeto no navegador, com a sua própria conta do GitHub.
- Sem Dockerfile nem arquivo Docker Compose
- Tem um modelo de pull request
- Ler o guia de contribuição
Primeiros passos
- Leia a issue inteira e depois o guia de contribuição do projeto.
- Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
- Faça um fork do repositório e trabalhe em uma branch.
- Abra um pull request que referencie o número da issue.
Mais de pingdotgg/t3code
-
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 75/100
Mantenedores costumam responder em até 1 dia
-
needs more info
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 72/100
pingdotgg/t3code#39400 · 1 comentário ·
Mantenedores costumam responder em até 1 dia
-
[Bug]: Wrapped PR titles lose bottom padding in the thread details panelTalvez já em andamento @iamdatvt94 assumiu há 1 dia. Abertabug needs-triage
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 72/100
Mantenedores costumam responder em até 1 dia
-
bug needs-triage
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 66/100
Mantenedores costumam responder em até 1 dia
-
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 73/100
Mantenedores costumam responder em até 1 dia
Todas as issues de pingdotgg/t3code
Issues semelhantes
-
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 84/100
twilio/twilio-node#1241 ·
Mantenedores costumam responder em até 1 dia
-
documentation
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 85/100
hugomrdias/playwright-test#786 · 1 comentário ·
Mantenedores costumam responder em até 1 dia
-
area:stdlib effort:S impact:med platform:linux type:bug
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 84/100
Mantenedores costumam responder em até 1 dia
-
Dificuldade 2/5 1-3 horas Facilidade para iniciantes 82/100
WenyuChiou/agent-board#1745 ·
Mantenedores costumam responder em até 1 dia
-
Dificuldade 1/5 1-3 horas Facilidade para iniciantes 88/100
Mantenedores costumam responder em até 1 dia