What it does

Chatd is an Electron desktop app for document Q&A using local AI. It ships with Ollama baked in, so you download, unzip, and run—no separate LLM installation, no cloud data, no terminal gymnastics. It defaults to Mistral-7B and will latch onto an existing Ollama instance if you already have one running.

The interesting bit

The packaging is the product. Most “local RAG” tools assume you’ve already wrestled with model weights and CUDA drivers. Chatd treats Ollama as an embedded dependency, managing its lifecycle automatically. That’s a genuine UX bet, not just a UI skin.

Key highlights

• Zero-dependency executable: Ollama runner bundled for macOS, Windows, and Linux
• Auto-detects existing local Ollama servers; starts its own if absent
• GPU support and custom model selection documented but off the happy path
• Unsigned Windows builds (you’ll get a security warning)
• macOS builds require Apple Developer signing for distribution beyond the build machine

Caveats

• The README doesn’t specify how document ingestion or chunking actually works—whether it’s simple full-text, vector search, or something else is unclear
• No mention of supported file formats (PDF? DOCX? TXT?)
• Windows and macOS packaging friction suggests this is still closer to “advanced user with patience” than “grandparent-friendly”

Verdict

Worth a look if you want local document chat without assembling the plumbing yourself. Skip it if you need fine-grained control over embedding models, chunking strategy, or already have a slick Ollama workflow you’re happy with.