One-click Ollama setup restored
Settings → Providers → Ollama can install the managed Ollama runtime again. Ollama changed the shape of its GitHub release artifacts, and OpenYak was still requesting the old macOS bare-binary URL, which returned 404 during setup. OpenYak now targets the current release archive names and extracts the runtime before starting it.
Current release assets across platforms
The managed installer now understands the formats Ollama currently publishes: macOS uses ollama-darwin.tgz, Linux amd64 and arm64 use .tar.zst archives, and Windows uses the platform-specific .zip assets. The extraction path validates that the Ollama binary exists, resolves nested Windows binaries, and restores executable permissions on Unix systems before launch.
Real install, chat, and uninstall validation
This release was validated beyond unit tests. We ran the desktop GUI flow against a temporary data directory, installed Ollama v0.22.0, pulled qwen2.5:0.5b, sent a chat through the local model, then removed Ollama and its model directory from the app. Backend, frontend, Ollama-specific, UI regression, bundle, signing, notarization, updater-manifest, and public download checks all passed for the 1.1.8 release.
Upgrading from 1.1.7
Users on 1.1.7 will receive 1.1.8 through the in-app updater, or can download the installer directly from the Download page. No migration is required. The update is especially recommended if you use OpenYak's managed Ollama setup instead of a manually installed local Ollama service.