Hacktoberfest 2026: los issues que los mantenedores marcaron para octubre, abiertos y aptos para principiantes. Explorar issues de Hacktoberfest

Cheaper subagents: run task and oracle on a mid-size model

Abierto
  • ¿Algo no está bien?
#78 2 comentarios 0 reacciones 0 asignados Ver en GitHub

Los mantenedores suelen responder en 1 día

Nadie ha tomado este issue todavía.

Evaluación

Dificultad
4/5
Tiempo estimado
3-5 días
Aptitud para principiantes
35/100
Tipo de issue
Nueva funcionalidad
Claridad
Bastante claro
Estado de actividad
Activo
Stack tecnológico
typescript
Área
ai-infra-agents, desktop

Línea de trabajo

Start by reading subagentConversation in chat_manager.ts and the model defaults in src/shared/models.ts; then review the agent benchmark in issue #49. Compare task quality, cost, and time on Opus and Sonnet before proposing a default, and check how a Settings option can preserve the chat's provider. Done means the setting and model behavior match the chosen policy without moving subagents across providers.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

enhancement performance priority: medium severity: low

Today

  • finder already runs on the provider's small model (SMALL_MODELS).
  • task (read-only research) and oracle (hard problems) run on the chat's own model (subagentConversation in chat_manager.ts). With the default Claude Opus 5.5, every task run is billed at Opus prices.

In a recent session, one task run took 93 s just to map how commands are run (globs, greps and file reads), work a mid-size model does about as well.

Proposal

  • task: run on a mid-size model by default (Claude Sonnet 5.5 at half Opus' price; the matching OpenAI model for OpenAI chats). Keep oracle on the chat's model, since its whole point is the strongest reasoning.
  • Setting: add a "Subagent model" option (Same as chat / mid-size / small) in Settings, with the default stored in src/shared/models.ts.
  • Effort: run subagents at a lower effort than the chat by default (see the speed issue).
  • Measure: cost, time and answer quality for task on Opus vs Sonnet with the agent benchmark (#49) before changing the default.

Keep the chat's provider: a subagent must not move an Anthropic chat to OpenAI or the other way round.

Lenguaje dominante
TypeScript
Estrellas
2
Forks
2
Merge medio
8 h 28 min
PR fusionados (30 d)
26

Preparar el entorno

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Más de PierrunoYT/patch

Todos los issues de PierrunoYT/patch

Issues similares

Más issues de TypeScript

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.