Hacktoberfest 2026: as issues que os mantenedores marcaram para outubro, abertas e boas para iniciantes. Ver issues do Hacktoberfest

[Feature]: PAIR feedback: model-level routing toggles, controlled test runs, and concurrency controls

Aberta
  • Algo errado?
#162 0 comentários 0 reações 0 responsáveis Ver no GitHub

Mantenedores costumam responder em até 4 dias

Ninguém assumiu esta issue ainda.

Avaliação

Dificuldade
4/5
Tempo estimado
Mais de uma semana
Facilidade para iniciantes
12/100
Tipo de issue
Funcionalidade
Clareza
Precisa de esclarecimento
Status de atividade
Ativa
Stack de tecnologia
go
Domínio
desktop

Direção de pesquisa

This is a broad feature request covering several separate controls: per-model routing toggles, keep-loaded, concurrency limits, and test-runner scope. No files, tests or entry points are named in the payload, so start by locating the model load/eject actions and the test runner in the desktop UI. Done would mean the request is split into smaller issues with agreed scope.

Escrita pelo modelo de indexação a partir do texto da issue.

Descrição

enhancement
Area

Desktop experience

User problem

Hi NVIDIA PAIR team,

First, thank you for building and releasing Personal AI Router. I’m genuinely excited about the product. I love NVIDIA’s local AI direction, and PAIR is exactly the kind of tool I want to use across my RTX machines.

After setting up PAIR across two Windows systems, I ran into a few workflow issues that I think could be improved with some model-level controls.

Right now, PAIR appears to expose model actions like load and eject, but I could not find a clean way to decide which installed models are eligible for routing. That makes the routing behavior feel too automatic. I would love to see controls such as:

  • Enable/disable routing per model
  • Keep-loaded toggle per model
  • Max concurrent models per node
  • Preferred node per model
  • “Test selected model only”
  • “Test selected node only”
  • A safer “test all” mode that runs one model at a time, or with a visible concurrency limit
  • Time Based Scheduled Loading of Model or Models and Model Grouping

In my setup, the broad test behavior appeared to launch many workload checks across many models at once. Some tests passed, but many failed with messages like “client disconnected before completion” or “context canceled.” When I tested specific models one at a time, the routes worked much more reliably. So the issue seemed less like every model was broken and more like the test runner was overloading the environment or timing out before queued model loads could complete.

I think PAIR would be much easier to trust if installed models, loaded models, and routing-eligible models were treated as separate states in the UI.

Thank you again for the work on PAIR. I’m excited about where this is going, and I hope this feedback is useful.

Best,
Chris

Desired outcome
  • Enable/disable routing per model
  • Keep-loaded toggle per model
  • Max concurrent models per node
  • Preferred node per model
  • “Test selected model only”
  • “Test selected node only”
  • A safer “test all” mode that runs one model at a time, or with a visible concurrency limit
  • Time Based Scheduled Loading of Model or Models and Model Grouping
Alternatives considered

No response

Compatibility and security implications

No response

Validation approach

No response

Confirmations
  • I searched existing issues for duplicates.
  • I agree to follow the Code of Conduct.
Linguagem predominante
Go
Estrelas
1.6k
Forks
266
Merge médio
3d 18h
PRs com merge (30d)
28

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de NVIDIA/Personal-AI-Router

Todas as issues de NVIDIA/Personal-AI-Router

Issues semelhantes

Mais issues de Go

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.