NVIDIA PAIR 1.0.0

This release brings llama.cpp and RTX Spark support to PAIR, and establishes
its public development and contribution workflow on GitHub.

What's New

• llama.cpp support. Install and manage llama.cpp directly from PAIR,
  download compatible GGUF models from Hugging Face, and route inference
  requests across paired devices. Refer to
  Managing Engines.
• Engine launch options. Customize the launch arguments and environment
  variables of PAIR-managed engines, on this device or a paired one. Tune
  settings such as context size where the engine supports them. Refer to
  Engine Settings.
• RTX Spark support. PAIR's Windows on ARM build has been validated with
  llama.cpp and CUDA on RTX Spark N1X hardware.
• Improved terminal experience. The terminal interface has a redesigned
  layout organized around each device, so you can manage engines, download
  models, and pair devices from one place. A new Jobs tab follows inference
  work across paired devices and runs the Inference Demo, giving headless
  machines the same core tools as the desktop app. Refer to
  Terminal Interface.

Compatibility Notes

• Browser clients. PAIR's local endpoints now follow each engine's CORS
  policy instead of allowing every browser origin. A browser-based client may
  need the engine configured to allow it — for Ollama, set OLLAMA_ORIGINS in
  the engine's launch options.
• The macOS app is now NVIDIA PAIR. The new NVIDIA PAIR.app installs
  beside the old PAIR.app, so remove the old one to keep just one copy. Your
  settings, cluster membership, and downloaded models carry over.
• LM Studio installed by an earlier PAIR. It keeps working as before. PAIR
  now removes only engines it has a record of installing, and earlier versions
  kept no record, so uninstalling or resetting PAIR leaves that LM Studio in
  place. You can remove it from ~/.lmstudio yourself if you no longer want it.

The Upgrade Guide walks through each of these.

Targeted Bug Fixes

• Uninstalling LM Studio from PAIR, or uninstalling PAIR with its data, no
  longer deletes the models you downloaded. They stay in ~/.lmstudio/models,
  just as Ollama's stay in ~/.ollama and llama.cpp's in ~/.llamacpp.

Community Contributions

• Added Anthropic Messages API routing through Ollama and LM Studio, so
  compatible clients such as Claude Code can use PAIR's model-aware routing and
  failover. Thanks @Ratherrayees!
  (#27)
• Made engine monitoring more efficient by reusing HTTP connections for health
  checks. Thanks @Psych0h3ad!
  (#37)
• Fixed unnecessary cluster trust refreshes: notifications are now sent only
  after new peer endorsements are saved. Thanks @RecursiveIntell!
  (#62)
• Fixed workload synchronization between nodes: broadcasts preserve event
  order, and a failed broker notification stays eligible for retry. Thanks
  @mkalkere! (#80)
• Fixed scheduler priority updates being marked as delivered after a failed
  notification, so they are retried and the status stays accurate. Thanks
  @sylvesterkaczmarek!
  (#122)
• Isolated the service integration and proxy failover tests from local PAIR
  settings, so tests no longer read or overwrite a developer's configuration.
  Thanks @somtri! (#95)
• Corrected the release-intent examples to use the required services version
  key. Thanks @somtri!
  (#120)
• Updated the contributor instructions to branch from and rebase onto
  develop. Thanks @somtri!
  (#126)
• Fixed node cards showing only the first GPU on multi-GPU systems; they now
  show every GPU. Thanks @CryptoJones!
  (#34)

Special Thanks

Thank you to these community members for sharing ideas, submitting proposals
and fixes, investigating issues, and testing PAIR:

@XxSilentStormIxX, @agenticocho, @Mayimbe07, @cruzanstx, @pettersandvand,
@cguldogan, @mkalkere, @CryptoJones, @woodsonl, @stevbev, @habakan, @arcaven,
@somtri, @imluffy000, @ndizazzo, and @fpvo.

We are grateful for every contribution, and for the warm welcome from the local
AI community.