Build improvements: cache policy, stock CUDA Ollama, and Nova concurrency #83
No reviewers
Labels
No labels
host:electra
host:fleet
host:lyra
host:vega
host:vega
phase:cutover
phase:deploy
phase:mcp
phase:module
phase:packaging
phase:prep
phase:validation
project:attic-postgres-lyra-rollout
project:auto-update-reliability
project:declarative-purity-cleanup
project:external-review
project:host-facts-refactor
project:lyra-nixos-deploy
project:lyra-service-stack-migration
project:nebula-mesh-network
project:security-hardening
project:service-stack-migration
project:vega-sillytavern-cutover
service:auto-update
service:mem0
service:sillytavern
No milestone
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
nimmo/nixos-config!83
Loading…
Reference in a new issue
No description provided.
Delete branch "build-improvements"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
Validation
nix flake checkbash -n scripts/attic-cache-build.sh./tests/auto-update-regressions.shnix build .#nixosConfigurations.lyra.pkgs.ollama-cuda --no-link --print-build-logsAutomated code review
Reviewed commit:
2060523f5bef408d539d49bee00e938d33fe5684Verdict: Requires further work
Resolve the blocking or important findings and investigate failed deterministic checks before merging.
Overall assessment
The pull request removes the Mouzi flake input and package, replacing it with a declarative systemd user timer/oneshot service for download organization. It updates the Attic cache workflow to check upstream substituters before pushing (removing --ignore-upstream-cache-filter) and rebalances Nova's build concurrency to 8 jobs × 4 cores while maintaining the 32-thread budget. The Ollama CUDA package override is removed, relying on an upstream nixpkgs fix.
Blocking findings
None.
Important findings
Removed nvidiaPackage override that patched CUDA_PATH and CUDAToolkit_ROOT; now uses pkgs.ollama-cuda directly.
Impact: If the pinned nixpkgs revision has not yet merged the upstream fix for CUDAToolkit_ROOT, Ollama CUDA builds will fail during configuration.
Suggested fix: Verify that the current nixpkgs pin includes the upstream correction. If not, retain a minimal override or update the nixpkgs input to a revision containing the fix.
Suggestions
None.
Tests and validation
Questions
Review limitations
Diff coverage
.forgejo/workflows/attic-cache.yml: reviewed — included in a context-limited batchREADME.md: reviewed — included in a context-limited batchflake.lock: reviewed — included in a context-limited batchflake.nix: reviewed — included in a context-limited batchhome/nimmo.nix: reviewed — included in a context-limited batchlib/auto-update-policy.nix: reviewed — included in a context-limited batchmodules/desktop/apps.nix: reviewed — included in a context-limited batchmodules/profiles/ollama.nix: reviewed — included in a context-limited batchpackages/mouzi.nix: reviewed — included in a context-limited batchscripts/attic-cache-build.sh: reviewed — included in a context-limited batchReview metadata
Review follow-up: verified the pinned Nixpkgs CUDA Ollama recipe. It now constructs its CUDA toolkit with
cuda_nvccand addscuda_nvcctonativeBuildInputs(the previous local workaround predated this). The exact stock derivation also evaluates and substitutes successfully withnix build .#nixosConfigurations.lyra.pkgs.ollama-cuda --no-link --print-build-logs.No compatibility override has been restored: doing so would create a distinct local derivation and forfeit the verified upstream substitute.