Skip to content

[Feature]: Unlimited generation timeout #411

Closed
#518
Closed
[Feature]: Unlimited generation timeout#411
#518
Labels
enhancementNew feature or requesttriageAwaiting maintainer review

Description

@thesourcerer8
thesourcerer8
opened on Aug 4, 2026

Problem / pain point

I am using local LLM's for confidentiality reasons. They are too slow for the timeouts offered in the GUI, 2 hours. I would like timeouts like 10 hours, or perhaps even unlimited. I usually run those generation jobs over-night, and I would prefer them to complete correctly than being stopped unnecessarily.

Proposed solution

Either add more timeout options or way to disable the timeout.

Alternatives considered

I could buy faster hardware, but since it will likely get old and too small too fast, I would prefer to use the hardware I have.

Scope

Generation capability

Hard-constraints check

Activity

  1. thesourcerer8
    added
    enhancementNew feature or request
    triageAwaiting maintainer review
    on Aug 4, 2026

  2. github-actions commented on Aug 4, 2026

    @github-actions
    github-actionsBot
    on Aug 4, 2026 – with GitHub Actions
    Contributor

    Thanks for the feature request! Long-running local generation jobs being stopped by a 2-hour ceiling is frustrating, especially for overnight workloads — making the timeout configurable (or optional) is a very reasonable ask.

    I checked the public repository files (CLAUDE.md, AGENTS.md, README, package.json, pnpm-lock.yaml) and couldn't find where the 2-hour timeout is implemented. The relevant code is likely in the app's settings or generation orchestration (apps/desktop or packages/core), but I can't confirm the exact file from this checkout. To help the team triage this efficiently, a few details would be useful:

    • Which version of Open CoDesign are you using (e.g., 0.2.0)?
    • Where exactly do you see the timeout option in the GUI (e.g., a settings screen, a dropdown before generation, etc.)?
    • Which local model backend are you using (Ollama, llama.cpp, LM Studio, etc.)?
    • Would a "No timeout" toggle, or longer presets (e.g., 6h/12h/24h), be preferable for your workflow?

    I've noted the issue as an enhancement for maintainer review. Thanks again for the detailed write-up!

    Open-CoDesign Bot

  3. linarodriguezvargas-dev commented on Aug 4, 2026

    @linarodriguezvargas-dev
    linarodriguezvargas-dev
    on Aug 4, 2026 via email

  4. github-actions commented on Aug 4, 2026

    @github-actions
    github-actionsBot
    on Aug 4, 2026 – with GitHub Actions
    Contributor

    Hola @linarodriguezvargas-dev,

    Entiendo que no deseas recibir estos correos. Solo para aclarar: esta notificación no la envía Open CoDesign, sino GitHub automáticamente porque tu cuenta está suscrita a este hilo (#411). Open CoDesign no guarda tu correo ni te ha registrado en ninguna lista; los avisos llegan a través de GitHub.

    Para dejar de recibirlos:

    No necesitas hacer nada más; si no vuelves a comentar en este hilo, dejarás de recibir estos avisos. ¡Gracias por escribirnos!

    Open-CoDesign Bot

  5. linarodriguezvargas-dev commented on Aug 4, 2026

    @linarodriguezvargas-dev
    linarodriguezvargas-dev
    on Aug 4, 2026 via email

  6. github-actions commented on Aug 4, 2026

    @github-actions
    github-actionsBot
    on Aug 4, 2026 – with GitHub Actions
    Contributor

    Hola @linarodriguezvargas-dev,

    Entiendo la confusión. Tranquila: no tienes una suscripción con Open CoDesign ni una cuenta que debas "eliminar". Open CoDesign es una aplicación de escritorio de código abierto; no guarda tu correo ni te cobra nada.

    Los correos que recibes son notificaciones automáticas de GitHub porque tu cuenta de GitHub está siguiendo este hilo (#411), o el repositorio. Para dejar de recibirlos no necesitas recordar ninguna contraseña ni iniciar sesión en ningún sitio:

    1. En el correo, al final, pulsa el enlace "Unsubscribe" / "Dejar de suscribirse". Ese enlace funciona sin que tengas que iniciar sesión.
    2. También puedes abrir este hilo en GitHub y pulsar Unsubscribe / Dejar de seguir en la parte superior, o Mute / Silenciar si prefieres dejar de recibir avisos sin abandonar el hilo.
    3. Si quieres revisar todas tus notificaciones de GitHub: https://github.com/settings/notifications

    Un consejo: si respondes otra vez a este hilo, GitHub puede volver a notificarte. Después de hacer el paso 1 o 2, no escribas más aquí y los avisos se detendrán.

    ¡Gracias por escribirnos y disculpa las molestias!

    Open-CoDesign Bot

  7. thesourcerer8 commented on Aug 4, 2026

    @thesourcerer8
    thesourcerer8
    on Aug 4, 2026
    Author

    I checked out the Open-CoDesign from GitHub about 2 days ago, so it should be approximately 0.2.0.
    The setting is under: Settings -> Advanced -> Generation Timeout and the maximum value there is "7200 s"
    I am using both Ollama and LM Studio at the moment.
    I think longer timeouts might be better for me, I dont want to have it running forever in case it crashed/looped somehow. But perhaps someone else has usescases for a disabled timeout too?

  8. github-actions commented on Aug 4, 2026

    @github-actions
    github-actionsBot
    on Aug 4, 2026 – with GitHub Actions
    Contributor

    Thanks for the extra details — this confirms the report.

    • Version: confirmed — the workspace root package.json currently says 0.2.0.
    • Settings location: Settings -> Advanced -> Generation Timeout, capped at 7200 s, matches exactly what you described.
    • Backends: Ollama + LM Studio are both local / OpenAI-compatible endpoints, and v0.2 runs generation as a long-lived pi session, so a 2-hour ceiling can absolutely interrupt overnight jobs on slower local hardware.

    I couldn't find the exact file:line for the timeout cap in the public checkout I have access to, so I won't guess at one. The change itself should be small: either add longer presets (e.g., 6/12/24 h) or add an opt-in No timeout setting. Your point about not wanting runaway jobs is fair — a finite default plus an explicit "No timeout" toggle would cover both: safe by default, with an escape hatch for long overnight runs.

    This is already labeled enhancement and is in triage, so maintainers will see it. The discussion about presets vs. no-timeout is useful signal for them.

    Open-CoDesign Bot

  9. Sun-sunshine06
    added 2 commits that reference this issue on Oct 4, 2026

    fix: forward generation timeout to model HTTP requests (#462)

    f35b854

    feat(desktop): notify when a background design run finishes (#472)

    c7b0f23
  10. Sun-sunshine06
    closed this as completedin #518on Oct 23, 2026
  11. Sun-sunshine06
    added a commit that references this issue on Oct 23, 2026

    feat(desktop): allow longer and unlimited generation timeouts (#518)

    75f1e22
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or requesttriageAwaiting maintainer review

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions