Ollama setup — popover states

Flow B (model-first / Ollama incidental): choosingModel → needsOllama → waitingForOllama → downloading. Each card shows the toolbar pill and the popover it opens, at native widths (280pt standard, 360pt for the model picker) so wrapping is real.

Honesty rule made visible: the ⏳ hourglass is static (waiting on the human); only the download bar and "Bristlenose-is-working" spinner animate.

STEP 1 · choosingModel · 360pt · daemon down

Set up on-device AI

Choose your on-device model

Tuned to your Mac (16 GB RAM). Bigger models are smarter but need more memory; all are a sizeable download.

Gemma 4 E4B · balanced · recommended
~3 GB download
Llama 3.2 3B · smallest · fastest
~2 GB download
Gemma 4 26B · best quality · needs 24 GB
~16 GB download

These run on Ollama, free software you'll set up next.

STEP 1 · choosingModel · 360pt · low disk space

Set up on-device AI

Choose your on-device model

Tuned to your Mac (16 GB RAM). Bigger models are smarter but need more memory; all are a sizeable download.

Gemma 4 E4B · balanced · recommended
~3 GB download
Llama 3.2 3B · smallest · fastest
~2 GB download
Gemma 4 26B · best quality · needs 24 GB
~16 GB download

Low disk space

These run on Ollama, free software you'll set up next.

STEP 1 · choosingModel · 360pt · daemon already up

Set up on-device AI

Choose your on-device model

Tuned to your Mac (16 GB RAM). One model is already on this Mac.

Already on this Mac
Gemma 4 E4B · ready · no download
3.0 GB on disk
Download
Llama 3.2 3B · smallest · fastest
~2 GB download
Gemma 4 26B · best quality · needs 24 GB
~16 GB download

Already installed — starts right away.

STEP 2 · needsOllama · 280pt · the "ah, but" reveal

Get Ollama

Get Ollama to run Gemma 4 E4B

Bristlenose runs models locally through Ollama — free. Opens ollama.com to download and install it.

STEP 3 · waitingForOllama · 280pt · passive · static ⏳

Waiting for Ollama

Finish installing Ollama

ollama.com is open in your browser.

  1. Download Ollama for macOS
  2. Open the download, install and launch Ollama

Bristlenose will download Gemma 4 E4B automatically once Ollama is running.

Waiting for Ollama install

DOWNLOADING · 280pt · animated bar (BN working)

Downloading Gemma 4 E4B 41%

Downloading Gemma 4 E4B

1.2 GB of 3.0 GB

FINISHING · 280pt · spinner (BN working)

Finishing up

Finishing up

Getting Gemma 4 E4B ready…

FAILED · 280pt · red ✗ · the only red state

Couldn't get model

Couldn't get model

No internet connection. Check your network and try again.

All pill texts, side by side (toolbar context)
Set up on-device AI Get Ollama Waiting for Ollama Downloading Gemma 4 E4B41% Downloading… Finishing up Couldn't get model
Failure detail strings — wrapping at the 248pt text column (title is always "Couldn't get model")

noInternet

No internet connection. Check your network and try again.

timedOut

The download timed out. Check your connection and try again.

cantReach

Lost connection to Ollama. Make sure it's running, then try again.

generic · %@ = system message

The download couldn't finish. The server returned an unexpected response.