v1.2.2

Clippy Vision 1.2.2

v1.2.0

macOS desktop support, delete conversations, markdown rendering in chat, in-app update notifications

Clippy Vision v1.0.1

Clippy Vision v1.0.1

Fix: keep text + vision models loaded together

Ollama defaults to loading one model at a time. That meant qwen3:8b (chat) and qwen3-vl:4b (vision) kept unloading each other — noticeable lag when capture and chat ran on the same machine.

This release:

• Setup wizard sets OLLAMA_MAX_LOADED_MODELS=2 and OLLAMA_NUM_PARALLEL=1 (persisted via Windows user env)
• Restarts Ollama during setup so the setting applies immediately
• Preflight on every launch ensures the same config for existing installs
• Warms both qwen3:8b and qwen3-vl:4b at the end of setup

Install

Download ClippyVision-Setup-1.0.1.exe below.

If Windows shows SmartScreen: More info → Run anyway (unsigned).

Existing v1.0.0 users: install over the previous version, or just launch — preflight will apply the Ollama setting and restart the service once.

Clippy Vision v1.0.0

Clippy Vision v1.0.0

A fully local AI assistant that watches your work to eliminate the context problem. 100% private — no cloud, no data leakage.

What's in this release

This is the first packaged release of Clippy Vision. Everything runs on your machine — no API keys, no subscriptions, no data leaving your device.

Installer features:

• Built-in setup wizard that installs Python, Ollama, and all required AI models automatically
• No terminal required — works for any Windows user
• Preflight health checks on every launch; redirects to setup wizard if any dependency is missing

Core features:

• Passive screen awareness — captures active windows, clipboard, typing patterns, and screenshots in the background
• Three-tier event classification (rule-based → feature-based → LLM fallback)
• Vision OCR and activity inference via qwen3-vl:4b
• Hierarchical memory: events → session summaries → distilled long-term facts
• Fine-tuned MiniLM query router for fast context prefetch before the LLM is called
• ReAct agent with SQL generation, memory recall, and fact saving
• Privacy-first: Clippy Vision's own window is blacked out in every screenshot before the AI sees it
• Per-app redaction rules (WhatsApp, Telegram, incognito windows, etc.)
• Capture toggle from tray icon or in-app button, with desktop notification on change

Models used (downloaded automatically on first launch):

Model	Size	Purpose
qwen3:8b	~4.7 GB	Reasoning, summarization, SQL, QA
qwen3-vl:4b	~2.9 GB	Vision / OCR
nomic-embed-text	~274 MB	Embeddings

Installation

1. Download ClippyVision-Setup-1.0.0.exe below
2. Run it — if Windows shows a "protected your PC" warning, click More info → Run anyway (the app is unsigned; see note below)
3. Follow the setup wizard

Requirements: Windows 10/11 (64-bit), 8 GB RAM min (16 GB recommended), ~10 GB free disk, internet on first run.

Note on code signing

This release is unsigned. The "Windows protected your PC" SmartScreen prompt is expected and safe to bypass. Code signing will be added in a future release.

Source

Full source code is available at https://github.com/protocorn/clippy-vision (AGPL-3.0).