Oyma v0.7.0: Local Ollama AI & BYOK Encryption

Release Highlights: Private AI On Your Own Terms

Version 0.7.0 brings comprehensive artificial intelligence capabilities to Oyma while maintaining our strict commitment to data privacy.

Architectural Overview

100% Offline Ollama Inference

For users requiring complete confidentiality, Oyma connects directly to your local Ollama daemon running on localhost:11434. You can summarize complex 90-minute design reviews using Llama 3.1 8B or Mistral without sending a single prompt token over the internet.

Secure Bring-Your-Own-Key (BYOK)

If you prefer frontier cloud models (Claude 3.5 Sonnet or GPT-4o), you can enter your personal API keys. Keys are stored encrypted inside the macOS Keychain. Calls travel directly from your Mac to the provider’s API with zero intermediary proxies.

Keychain Security & Prompt Sandboxing

We engineered the AI subsystem with strict hardware security standards:

• API credentials use macOS SecItemAdd and SecItemCopyMatching routines with accessibility attributes restricted to the running binary.
• Prompts and generated text are never cached to temporary disk files or transmitted to telemetry endpoints.
• When local-only mode is active, all outbound network sockets are blocked at the app sandbox boundary.

Explore our Ollama AI guide, test model configurations with our Ollama model picker, read the BYOK documentation, and inspect our privacy architecture.