V_1.8.0

• FIXED: Gemini deprecated model ID migration (gemini-3-pro-preview → gemini-3.1-pro-preview)
• CHANGED: removed x86/x86_64 ABI filters, reducing APK size (no Android devices use x86 since 2018)

V_1.7.0 Model selection for Claude/Gemini

Added:

• Model selection for Claude/Gemini: now supporting latest models

Updated:

• Engine used for local tensorflow config
• Default engine for Openrouter.ai (previous one has been retired)

V_1.6.1 Fixed missing Openrouter configs from Tasker screens

FIXED: Missing OpenRouter option from MacroDroid/Tasker screens

V_1.6.0 Improved listener features + updated cloud models

• ADDED: Notification listener can now optionally intercept notifications without images. Useful when the app sends a notification without images but the image can be fetched via API.
• ADDED: "Cancel Notification" action: useful to remove false positives' notifications (e.g. after analysing Eufy image preview).
• CHANGED: updated to using Gemini Flash 2.5 instead of 2.0
• CHANGED: updated to using Claude Sonnet 4.5 since 3.5 is being retired

V_1.5.0 added support for Openrouter

• ADDED: support OpenRouter.AI for cloud queries: allowing completely free online image detection!
  This is a very powerful addition since it allows to:
  - use free models
  - use almost any model which is vision-capable (future proof)
  - get higher uptime (Openrouter will use fallback logic on different providers for some models)
• FIXED: issue where NotificationInterceptor will only apply appName filter for the last edited filter!

V_1.4.0 Events when an app creates a notification with an image

ADDED FEATURE: intercept notifications from other apps and Generate a Tasker Event

V_1.3.0 General purpose Image Analysis

• ADDED: General purpose image analysis: this allows to ask any general purpose questions to the cloud LLMs (Gemini/Claude)

With this release you can automate things like:

• check if the lights on the porch are still on and turn them off
• check if you parked the car in the backyard or not
• check and close the garage door after a certain hour (and maybe check that there's no one in between!)

Remember: use caution since LLMs are quite reliable but still have some epic fails. Don't use these AI systems for safety critical tasks!

V_1.2.0 Added Google Gemini

• ADDED: support for Google Gemini
• IMPROVED: filesystem selection will remember last selected image in the test interface

V_1.1.2 - Beta 1 - improved messaging and test UX

• improved test UX: now you'll see the reasoning behind the response
• added: second output variable with explanation

V_1.1.1 Beta 1 - Improved ClaudeAI detection

IMPROVED: Claude.AI detection to lower false positives: it's now safe to trigger alarms (e.g. sirens)
ADDED: now Claude.AI supports both JPEG and PNG images