Chats

A chat in Pallm lives on your Mac. Your phone, iPad and other Macs show the same chats, and a reply started on one streams on all of them.

Pick a model

The bar under the message box shows three things: the chat’s model, its thinking level and how full its context is. Tap the model to switch.

• Automatic is the default for new chats. It uses the model already loaded on your Mac, so a new chat doesn’t wait while another one loads. If none is loaded, it uses the one you used last.
• A memory-chip icon marks the model that’s loaded right now.
• Favourite models… in Settings lets you star models. Favourites come first in every picker, on every device.
• A chat keeps the model it started on. Changing the default later doesn’t change old chats.

To choose a default for one device only, use Settings › Models › Just this device.

Thinking

Reasoning models can think before they answer. The thinking level is Auto, Off, Low, Medium or High. Higher levels take longer and often answer harder questions better. Models that don’t reason ignore it.

Context and compacting

The percentage in the bar is how much of the model’s context window the chat uses. Near the limit, tap it and choose Compact conversation: the model writes a summary of the chat so far, and from then on sees that summary in place of the full history. A divider in the transcript shows where it happened. Pallm also compacts by itself when a chat would otherwise overflow.

Attachments

Use the attach button, paste, or drag files onto the message box:

• Photos and screenshots, for models that can see.
• PDFs, read as text. A scanned PDF is sent as page images.
• Text and code files, sent as text.
• Audio and video, for models that take them. Video is sent as a few frames and its sound.

Files are prepared on your device before they’re sent. If the chat’s model can’t take a file, the message box says so before you send it.

From any other app on iPhone or iPad, use the share sheet: choose Pallm, add a line if you like, and pick a chat or start a new one.

Artifacts

When you ask for something you’d want to look at or use, such as a web page, a chart, a diagram or a small React app, the model can make it as an artifact. Artifacts run right in the chat, with no internet needed: Pallm carries React, Tailwind, Recharts, D3, Three.js, Mermaid and others.

• Tap an artifact to open it. Preview and Source switch between the result and its code.
• When the model changes an artifact, it keeps the old versions, so you can step back through them.
• The chat’s side panel (a sheet on iPhone) lists every artifact under Artifacts, next to Files and Web.

Voice

• Dictation. Tap the microphone in the message box, speak, then Insert or Insert & Send. Speech is turned into text on your device, never on a server.
• Read aloud. Every reply has a button to read it out with your device’s voice. It can start while the reply is still being written.

When you’re offline

Send a message with no connection and Pallm keeps it. It goes as soon as your device can reach your Mac again. A reply that was streaming when the connection dropped picks up exactly where it stopped, with nothing missed and nothing repeated.

Notifications

When a reply finishes and you’re not looking at it, Pallm tells you:

• Looking at that chat on any device: nothing.
• Using Pallm on another device: the news appears in the bar at the top of that device.
• Not using Pallm anywhere: a notification.

A chat with a dot has something you haven’t seen. Opening it anywhere clears it everywhere.