Hacktoberfest 2026: le issue che i maintainer hanno segnato per ottobre, aperte e adatte ai principianti. Sfoglia le issue Hacktoberfest

OTel: parent chat spans keep the last subagent model in  gen_ai.request.model

Aperta
#4,970 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
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
52/100
Tipo di issue
Bug
Chiarezza
Abbastanza chiara
Stato di attività
Attiva
Stack tecnologico
shell
Ambito
cli, observability

Direzione di ricerca

Start by tracing the Copilot CLI's OTel export path and how session.model_change events with source=agent update model state for parent chat spans. Compare where gen_ai.request.model and gen_ai.response.model are populated across parent and subagent calls. Done means later parent spans retain the parent model while genuine dated model snapshots remain valid, with regression coverage for the described sequence.

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

Descrizione

triage
Describe the bug

After a subagent model switch, later parent  chat  spans keep the subagent's model in  gen_ai.request.model .  gen_ai.response.model  and the billed AIU stamp still match the model that actually served the call.

Observed on Copilot CLI 1.0.88. A session was switched to  claude-opus-5.5  with the model picker, then dispatched subagents ( session.model_change  with  source=agent  to  gpt-5.6-sol ,  gpt-6-sol ,  grok-4.7 , and others). Parent calls continued on Opus. Of 226 parent  chat  spans whose  gen_ai.response.model  was  claude-opus-5.5 , only 10 also had  gen_ai.request.model=claude-opus-5.5 . The other 216 were labeled with whichever subagent model had been selected most recently.

The AIU on those spans is correct. One mislabeled span had  gen_ai.request.model=gpt-5.6-sol ,  gen_ai.response.model=claude-opus-5.5 , input 85409, cache read 83052, cache write 2353, output 593, and  copilot_chat.copilot_usage_nano_aiu=4025140000 . That is $0.0402514, which matches the Opus 5.5 rate card exactly, not Sol's. Over that session the mislabeled Opus spans were about $20.43 of a $21.24 Opus total. Across 24 hours this was 521 CLI spans and about $41.37, all  service.name=github-copilot . VS Code  copilot-chat  did not show this shape.

Consumers that attribute cost by  gen_ai.request.model  therefore file Opus spend under the leftover subagent model. The total is not lost. The slice is.

Affected version

GitHub Copilot CLI 1.0.88

Steps to reproduce the behavior
  1. Run Copilot CLI 1.0.88 with OTel export enabled.
  2. Start a session and select a parent model with the model picker, for example  claude-opus-5.5 .
  3. Dispatch one or more subagents that run a different model, so the session log records  session.model_change  with  source=agent .
  4. Let the parent keep making model calls after those switches.
  5. Compare  gen_ai.request.model  and  gen_ai.response.model  on the later parent  chat  spans.

The request label is the last subagent model. The response label, token usage, and  copilot_chat.copilot_usage_nano_aiu  still describe the parent model.

Expected behavior

On a parent  chat  span,  gen_ai.request.model  should be the model that parent call asked for, matching  gen_ai.response.model  except for a genuine dated snapshot of that same model. A nested subagent's model switch should not leak onto later parent spans.

Ignoring  gen_ai.request.model is not a valid workaround. For some models the response label is a dated snapshot of the request, not a different model:  gpt-5.4  is served as  gpt-5.4-2026-03-05 , and  claude-haiku-4.5  as  claude-haiku-4-5-20251001 . Those are aliases. This bug is a different model left in the request label.

Additional context

• Windows, PowerShell, CLI 1.0.88. Spans are  service.name=github-copilot .
• The leak tracks  session.model_change  events with  source=agent . The picker change ( source=model_picker ) sets the parent correctly. Later agent-sourced changes overwrite the request label used on subsequent parent spans.
• Billing attributes are present and consistent with the response model. This is not github/copilot-cli#4224. It is also not github/copilot-cli#4862 (those spans are missing response identity and usage) or github/copilot-cli#4825 (HydraFusion, where  gen_ai.request.model=hydrafusion  is intentional).

Lingua principale
Shell
Stelle
11.2k
Fork
1.9k
Merge medio
17h 6m
PR unite (30g)
5

Preparare l'ambiente

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 github/copilot-cli

Tutte le issue di github/copilot-cli

Issue simili

Altre issue su Shell/Bash

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.