A desktop switchboard for the LLM-agnostic

What it does ChatWise is a native desktop chat client for macOS and Windows that talks to all the usual LLM suspects—OpenAI, Anthropic, Google AI, Groq, and Ollama—without opening a browser. It ingests text, images, audio, and PDFs, then reads replies aloud via OpenAI or ElevenLabs text-to-speech. The README is sparse on UI details, but the pitch is clear: one window, many backends.

The interesting bit Rather than pretending every model speaks the same dialect, ChatWise acts as a pragmatic compatibility layer. When a backend doesn’t natively support PDFs or audio, the app extracts the text and injects it as a prompt. It is a low-friction admission that “multi-modal” is still a patchwork of vendor-specific capabilities.

Key highlights

• Aggregates major commercial and local LLM providers behind a single desktop interface
• File support degrades gracefully: PDFs and audio fall back to text extraction for models that lack native vision or ears
• Text-to-speech powered by OpenAI and ElevenLabs
• Runs offline when paired with local models like Ollama; cloud APIs still need a connection, obviously
• Mobile version is listed as “soon(ish)”

Caveats

• The README never explains the repo’s “fastest” claim, nor offers benchmarks or architecture notes to back it up.
• Native multi-modal support is fragmented: PDFs and audio are only native to specific models (Claude 3.5 Sonnet, Gemini, GPT-4o); everything else gets a plain-text translation.
• A mobile app is explicitly not available yet.

Verdict Grab it if you want one desktop app for both cloud and local LLMs and can tolerate silent text extraction for some file types. If you live inside one provider’s ecosystem and need its full native feature set, stay in the browser.