AI Assistant & Images

What the assistant does

Notara’s AI assistant is designed as a workspace helper rather than a separate chat app. The current implementation supports:

• drafting and rewriting help
• note summaries
• idea generation
• image generation from prompts
• session continuity while you move around the app
• chat archive saving
• saving conversations into markdown notes
• saving generated images into vision boards or local media storage

Pollinations configuration

The app stores a small Pollinations config locally with:

• API key
• text model
• image model

The default values in code are:

Setting	Default
Text model	gemini-fast
Image model	flux
API key	empty

How requests are routed

Browser and hosted web runtime

Web builds send AI requests through proxy endpoints:

• /api/pollinations/text
• /api/pollinations/image

That design avoids browser CORS issues and allows an optional server-side token to be injected.

Tauri desktop runtime

Desktop builds do not rely on those browser routes. Instead, they use Tauri’s native HTTP plugin to talk directly to https://gen.pollinations.ai/ endpoints.

That split matters because it keeps the desktop app functional even outside a hosted proxy setup.

Saving AI output

Generated and imported images can be stored in the local workspace under data/media/. Saved AI conversation archives are written to data/ai/conversations.json when filesystem access is available.

Token behavior

If you provide a Pollinations token, Notara normalizes it into a Bearer authorization header. If you do not, the app still works for flows that do not require an authenticated token, and users can also enter their own key in Settings.

Optional auth is separate from AI

Pollinations setup is independent from Supabase auth. The codebase keeps legacy authentication flows available, but AI features do not require you to enable account sign-in.