SpecLens v1.1.0
Local AI - on-device summaries, fully private. Nothing runs or downloads until you enable it and fetch a model; nothing ever leaves your machine.
Highlights
- AI summaries everywhere - a golden ✨ button on the project Overview and on every document (proposals, tasks, design docs, specs, ADRs, playbooks) streams a summary into a resizable right-side panel. Capability names link straight to their specs.
- Reviewer-oriented document summaries - what the document is, its key points, and a "worth a reviewer's attention" list of risks and open questions.
- Background generation - keep reading while it works; a notification tells you when the summary is ready. Summaries cache per document and regenerate only when content changes.
- Five curated models - Gemma 4 E2B (default) and E4B, Qwen3.5 4B, Phi-4 Mini, SmolLM3 3B - downloaded on demand with checksum verification, all managed from Settings.
- Bring your own model - import any llama.cpp-compatible GGUF file, or drop it into the models folder.
- Ollama support - models from a local Ollama server appear automatically and generate through it.
- Settings redesign - a scannable model list with per-row download/delete, storage totals, and a simpler two-tab layout.
Known limitations
- AI inference is Metal-accelerated on Apple Silicon and CPU-based elsewhere; small models can be slow on older machines
- Windows installers remain unsigned - SmartScreen will warn on first run