Configure oh-my-openagent (omo) plugin for multi-agent orchestration
using ollama-cloud and local llama-swap providers. Primary model is
ollama-cloud/glm-5.1 with fallback chains. Add runtime fallback,
background task concurrency limits, and disable incompatible agents
(hephaestus, multimodal-looker).
- Add local-inference NixOS role using llama-swap (from nixpkgs-unstable)
with llama.cpp (CUDA-enabled, from nixpkgs-unstable)
- Serves Qwen3.6-35B-A3B via HuggingFace auto-download with --cpu-moe
- Add nixosSpecialArgs for nixpkgs-unstable module access
- Configure opencode with llama-local provider pointing to zix790prors:8080
- Update gptel from Ollama backend to OpenAI-compatible llama-swap backend
- Remove ollama service from zix790prors
- Add packages/opencode with pre-built binaries from GitHub releases (v1.4.0)
- Support all 4 platforms (aarch64-darwin, x86_64-darwin, x86_64-linux, aarch64-linux)
- Add update-opencode flake app for automated version bumps
- Install opencode via home.roles.base so it's available on all machines
- Reformat flake.nix and packages with nixfmt
- Update system attribute to stdenv.hostPlatform.system in overlays
- Migrate git config to new settings structure (user.name, user.email)
- Move ssh.addKeysToAgent to matchBlocks configuration
- Disable ssh default config to prevent future deprecation warnings
Configure nix.gc to automatically clean up old home-manager generations
weekly, which complements the existing system-level gc. Uses
--delete-older-than 10d on Linux to maintain a rollback window, but
overrides to -d on Darwin to work around a launchd bug where multi-part
options aren't properly split into separate arguments.
