LineLoom v0.9.1 — Pre-Dispatch Privacy & Model-Picker Safety Preview

LineLoom v0.9.1 — Pre-Dispatch Privacy & Model-Picker Safety Preview

GitHub pre-release · MIT license · local-first VS Code inline autocomplete using independently installed Ollama

Changes since v0.9.0

• Prevent stale requests before sending code: the extension rechecks the current editor, trusted workspace, active model/language settings, custom file exclusions and remote-server approval after the typing delay and private-file scans, immediately before contacting Ollama.
• Stricter cached suggestions: typed continuations and exact-result cache returns are also gated by up-to-date settings, rather than trusting only the configuration captured before debounce.
• Safer model pickers: LineLoom: Select Model and Assign Language Model discard choices retrieved from an endpoint that is no longer configured or approved. Approval dialogs no longer grant consent to an endpoint after the user switches it while the dialog is open.
• Additional configuration guard: changing excluded-path patterns or request timeout during a pending completion invalidates the old request.
• Regression tests: Windows, macOS and Linux CI checks, pure settings/approval tests and actual VS Code extension-host testing with a simulated Ollama server. In the editor tests, changing model or ignore patterns while an automatic completion waits to dispatch produces no Ollama generation request.
• Existing features retained: per-language model assignment, global fallback, nine registered languages (Go, Java, C and C++ experimental), typed continuation reuse, Quick Actions, local-first Ollama inference, conservative file exclusions, no telemetry, explicit HTTPS remote approval and HTTP redirect rejection.

No new runtime dependencies, source indexing, background model downloads, hosted service or account requirement have been introduced.

Install / upgrade

1. Use VS Code compatible with extension engine ^1.96.0, and install/start Ollama plus your chosen coding model yourself.
2. Download LineLoom-0.9.1.vsix from this GitHub pre-release's assets.
3. In VS Code choose Extensions → … → Install from VSIX, or run code --install-extension LineLoom-0.9.1.vsix.
4. Open a trusted source workspace and use LineLoom: Quick Actions to check the current server, select the global model and optionally assign per-language models.
5. Use LineLoom: Trigger Completion to request a suggestion. If settings or consent change during an inference request, request a fresh completion.
6. Optionally verify the LineLoom-0.9.1.sha256 asset with sha256sum -c LineLoom-0.9.1.sha256 on a supported shell.

Automated checks and limitations

The publication workflow requires a locked npm install, TypeScript typechecking, ESLint, Vitest, extension bundle, VS Code extension-host integration with mocked Ollama, npm dependency audit, VSIX ZIP integrity verification and SHA-256 generation. Separate repository CI runs on Windows, macOS and Ubuntu.

Still not verified: real Ollama model quality or latency, physical CPU/GPU performance, fully offline first-run setup, manual install and ghost-text acceptance on all three operating systems. Pre-dispatch revalidation prevents outdated requests from starting; it does not recall code previously sent to an approved endpoint. LineLoom does not scan file contents for secrets and does not perform cross-file context indexing.

This is a GitHub pre-release, not a stable v1.0.0 release, and there is no claimed VS Code Marketplace/Open VSX listing.

License: MIT for authorized project-owned material; historical Apache-2.0 releases keep their original terms. Dependencies and separately installed models are governed by their own licenses. https://github.com/sanskarIN/LineLoom/blob/main/docs/licensing.md

Repository: 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