[Bug]: t3_thread_read drops the typed failure (usage_limit class and resetAt) on error items

Aperta Adatta ai principianti
#35,270 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
2/5
Tempo stimato
1-3 ore
Idoneità per principianti
75/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
typescript

Direzione di ricerca

Inizia dai contratti MCP in packages/contracts/src/orchestratorMcp.ts e da timelineItem() e dal generatore dei dettagli del thread in apps/server/src/mcp/OrchestratorMcpService.ts. Segui come l’elemento di errore e i campi dello shell del thread vengono mappati nella risposta, poi controlla i test esistenti dei contratti e del servizio MCP. Il lavoro è completato quando t3_thread_read espone la classe di errore tipizzata e l’ora di ripristino per gli elementi di errore e i dettagli del thread.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

Before submitting
  • I searched existing issues and did not find a duplicate. (#31057 is about limited children resuming on the wrong account; it is not this.)
  • I included enough detail to reproduce or investigate the problem.
Area

apps/server: the orchestrator MCP read (t3_thread_read)

Steps to reproduce
  1. Start a thread on a provider that is out of usage. Seen with Codex, which fails the turn with usageLimitExceeded.
  2. The turn fails. T3's own store holds the typed failure on the error item: in orchestration_v2_projection_turn_items, failure = { class: "usage_limit", code: "usageLimitExceeded", resetAt: "<ISO>" }. The thread shell's lastErrorClass and usageLimitResetAt are set from it (threadErrorSummary).
  3. Read the thread through MCP: t3_thread_read with view: "activity", and the default view.
Expected behavior

An MCP client can tell a usage-limit failure, and its reset time, from any other failure using typed fields, without reading the human-readable message. One of these:

  • OrchestratorMcpThreadTimelineItem for type: "error" carries failure: { class, code, resetAt }, or at least failureClass and resetAt;
  • the thread object returned by t3_thread_read carries lastErrorClass and usageLimitResetAt, as the shell projection already does.
Actual behavior

The error item comes back with type, status, title, text and ids only. failure reaches the client only as text, because turnItemText returns item.failure.message for case "error" (apps/server/src/mcp/OrchestratorMcpService.ts, timelineItem(), at ac781dcb). The thread object has no lastErrorClass or usageLimitResetAt.

Claude limits are visible, because they also produce a system_notice whose item id ends usage-limit:…:<window>:<reset epoch seconds>. Codex limits produce only the error item, so through MCP a Codex usage limit looks like any other failed turn.

Impact

An MCP client that orchestrates agents cannot treat a usage limit as a wait without parsing English error text. It sees a plain failure, and may fail or relaunch work that T3's own UsageLimitRecoveryWorker would resume at resetAt. The fix is small: add the typed fields to the MCP timeline item and thread contracts in packages/contracts/src/orchestratorMcp.ts, then fill them in timelineItem() and the thread detail builder.

Version or commit

main at 59118847; OrchestratorMcpService.ts last changed in ac781dcb.

Environment

T3 server on Linux, threads started over MCP.

Logs or stack traces

The error item's text on the Codex thread: "You've hit your usage limit. Visit https://chatgpt.com/settings/usage to purchase more credits or try again at Nov 17th, 2026 7:32 PM."

— Model: claude-opus-5-5 · Harness: Claude Code (T3 Code)

Lingua principale
TypeScript
Stelle
24.8k
Fork
6.4k
Merge medio
4h 16m
PR unite (30g)
352

Preparare l'ambiente

Apri in Codespaces

Avvia il container di sviluppo del progetto nel browser, con il tuo account GitHub.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di pingdotgg/t3code

Tutte le issue di pingdotgg/t3code

Issue simili

Altre issue su TypeScript

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.