Skip to content

LineLoom v0.9.0 — Per-Language Ollama Models Preview

Pre-release
Pre-release

Choose a tag to compare

View all tags
@github-actions github-actions released this 19 Oct 07:05
v0.9.0
18a7de3

LineLoom v0.9.0 — Per-Language Ollama Models Preview

GitHub pre-release · MIT-licensed LineLoom source · local-first VS Code inline autocomplete

Changes since v0.8.1

  • Different models for different programming languages: set a custom installed Ollama model for Python, JavaScript/JSX, TypeScript/TSX, Rust, Kotlin, Go, Java, C or C++. Each language without an override continues to use the global default model.
  • Quick Actions workflow: choose LineLoom: Assign Language Model from the status bar or Command Palette. Pick a language, then an installed model, or choose Use Default Model to clear its override.
  • No automatic model downloads: the picker reads models from the already configured Ollama origin only after user initiation. A remote origin continues to require separately approved HTTPS access.
  • Correct inference routing: model selection affects prompts (including supported FIM routing), Ollama request payloads, exact-result cache hashes and typed-continuation profile hashes. Changes to language-model preferences invalidate stale in-flight suggestions.
  • Safety and tests: unknown keys, invalid model values and control characters in manual settings are ignored. The unit suite covers mapping and fallback; a VS Code extension-host integration test confirms the model field sent to a simulated Ollama server changes for TypeScript and returns to the global default when an override is cleared.
  • Existing features remain: nine supported language IDs, Quick Actions, manual-only mode, conservative exclusion of private files, bounded active-document context, HTTP redirect rejection, per-origin remote HTTPS approval and no telemetry.

Installation and usage

  1. Install VS Code compatible with extension engine ^1.96.0. Start Ollama and install the coding models you want independently.
  2. Download LineLoom-0.9.0.vsix from this GitHub release's assets.
  3. Install through Extensions → … → Install from VSIX, or run code --install-extension LineLoom-0.9.0.vsix.
  4. Open a trusted workspace, click the LineLoom status-bar item and check your Ollama connection.
  5. Choose Select Model for the global default, then Assign Language Model to configure optional per-language overrides. Model names can also be edited under the lineloom.languageModels setting.
  6. Verify the attached SHA-256 checksum with sha256sum -c LineLoom-0.9.0.sha256 on supported systems.

Verification and limitations

Release verification includes locked npm ci, TypeScript, ESLint, Vitest unit tests, extension bundling, a real VS Code extension-host test against mocked Ollama, dependency audit, VSIX ZIP inspection and SHA-256 checksum generation. Repository CI separately checks Windows, macOS and Ubuntu.

Not yet verified: the usefulness or latency of installed real models across languages, comparative local CPU/GPU performance, completely offline end-user setup, and manual VSIX acceptance on each desktop operating system. The language-specific model picker does not rank models or guarantee they are optimized for a language. Go, Java, C and C++ remain experimental.

This is a GitHub pre-release, not stable v1.0.0. No VS Code Marketplace or Open VSX listing is claimed.

License: MIT for authorized project-owned LineLoom material; previous Apache-2.0 release tags retain their own license terms, and dependencies/models are separately licensed. https://github.com/sanskarIN/LineLoom/blob/main/docs/licensing.md

Source: https://github.com/sanskarIN/LineLoom · Issues: https://github.com/sanskarIN/LineLoom/issues · Security: https://github.com/sanskarIN/LineLoom/blob/main/SECURITY.md

Maintainer: https://github.com/sanskarIN · https://x.com/SanskarCodes · https://linkedin.com/in/SanskarCodes

Optional support: https://buymeacoffee.com/SanskarCodes · https://razorpay.me/@SanskarCodes · https://paypal.me/SanskarCodes

Assets 4
Loading