Oriloq Documentation

Local multimodal Gemma chat

Oriloq has relied on the Gemma family since 1.0.13; since 1.0.21, it uses Gemma 4 QAT models (replacing MXFP4), with MLX inference and very large contexts. The engine runs directly on your Mac with a reduced memory footprint. It is very likely the best local chat implementation currently available on macOS. Since 1.0.19, native markdown rendering keeps bubble heights stable during generation.

• →Multimodal — understands text, images and sound in the same conversation.
• →Very large context — suitable for long documents and sustained conversations.
• →Local by default — with local engines, no data leaves your machine, ideal for confidential topics.
• →Smart memory management — the model automatically unloads from RAM after 5 minutes of inactivity (1.0.14).

Ollama chat

Since version 1.0.14, Oriloq can connect to your local Ollama instance to chat with the models you have installed there (Llama, Mistral, Qwen, Phi, etc.). Handy if you already manage a fleet of models via Ollama and want to use them directly from Oriloq without switching tools.

Configuration: enter the URL of your Ollama server (default http://localhost:11434) in the chat settings, then select the desired model from the automatically detected list.

Custom providers

Custom providers — more reliable validation (1.0.26); since 1.0.25, local providers available on your Mac (such as Ollama) are detected and added automatically from the model manager.

Unified model selector

Unified model selector (1.0.24) — pin your favorites, check each model's image, audio and reasoning capabilities, and open its advanced settings.

Conversation duplication

Since version 1.0.14, you can duplicate an existing conversation in the Gemma chat to create an independent copy. Useful for exploring multiple discussion branches from the same context without losing the original.

Conversation history management

History management (1.0.24) — import, export, rename, duplicate or delete one or several conversations, including in bulk.

Smart Actions backed by local chat

The local Gemma engine can be used as a backend for Smart Actions. Text transformation then happens without any network call. See Smart Action to configure the execution engine.

Audio workflows

You can send an audio file directly to the Gemma chat for transcription, summarization or analysis. Audio workflows have been significantly improved in 1.0.14 (smoother and more predictable handling of audio files in the chat). Since 1.0.19, the chat's speech synthesis is enhanced with sound effects (GladosKit).