Smartphones

While there isn't a single "plug-and-play" app that mimics the Google Pixel's AI call screening exactly on Linux mobile (Phosh), you can achieve a similar workflow by layering a few tools. Since you're already familiar with Phosh Anti-Spam, you know it works by intercepting the DBus signal to gnome-calls.
To go beyond simple hanging up and move into "AI screening," here is the current landscape for 2026:

1. The "Contact App" Approach: GNOME Calls + Plugins​

The primary call handler on Phosh is GNOME Calls. It uses a plugin-based architecture (libpeas).

• The Status: There is no official "AI Screen" button yet, but the Ubuntu Touch/Lomiri ecosystem and some PostmarketOS community members have been experimenting with Whisper.cpp (for STT) and local LLMs to process incoming call audio.
• The Hack: Some users are routing calls through a SIP/VoIP provider (like Twilio or JMP.chat) that supports programmable voice. You can set up a "vocal menu" or an AI agent on the provider side that screens the call before it even hits your Linux phone's Phosh Anti-Spam filter.

2. Emerging AI Apps for Linux Mobile​

If you want something that runs locally on your device:

• Kai 9000: This is a newer open-source AI assistant (available on GitHub) that has a Linux version. It is designed to execute "Tools." While it doesn't hook into the modem directly yet, developers are working on using it as a "Heartbeat" service that can monitor notifications and potentially interface with the audio stream of an active call to transcribe and suggest replies.
  
  
  
• Phosh Anti-Spam (Advanced Config): You can technically modify the Phosh Anti-Spam script. Instead of just hangup, you can point it to a script that answers the call and plays an audio file (using pacmd or pw-play), then records the response for you to review via a notification.

3. DIY "AI Screener" Workflow​

If you're comfortable with some "mechanical guy" tinkering, this is the most effective setup right now:

1. Phosh Anti-Spam: Keeps your "Spam" list and "Non-Contacts" blocked or silenced.
   
   
   
2. JMP.chat / Cheogram: Use a SIP number for public use. These services allow you to bridge your calls to an XMPP bot.
3. The Bot: You can point that bot to a local or API-based LLM. When a stranger calls, the bot answers, asks "Who is this?", transcribes the answer, and sends you an XMPP message (which shows up as a standard Phosh notification). You then decide whether to "patch through" the call to your actual dialer.

Summary Table​

Pro-Tip: If you're printing 3D brackets for your tech gear, you might consider a small "desk dock" setup for your phone. Some users are running more powerful AI screening scripts by keeping the phone docked and using the desktop-class CPU power of the PinePhone Pro or Librem 5 to handle the STT/LLM processing without killing the battery.