Email AI Assistant: Private inbox prioritization

The problem

Checking an inbox repeatedly is distracting, but simple filters rarely understand why a message matters to a particular person. Email AI Assistant combines local processing with user feedback to make that prioritization more personal.

How it works

• Electron and React provide the cross-platform desktop interface.
• Gmail APIs retrieve messages with the permissions granted by the user.
• Llama runs locally through Ollama, keeping message content on the user's machine during model processing.
• A chat interface records preferences and corrections in a reusable memory.

Engineering focus

The application separates mail access, model processing, preference memory, and notifications. This makes the data boundary visible and allows each part to be tested independently.

The user can explain what matters in natural language instead of maintaining a growing collection of inbox rules. Corrections become part of the preference context used for later messages.

Result

Email AI Assistant demonstrates how a private model can support a useful daily workflow on ordinary desktop hardware. It reduces the need to scan every incoming message while keeping the user in control of both Gmail access and prioritization behavior.