# Ollama
Run Producer Pal completely offline with local models.
## What You Need
- [Ollama](https://ollama.com/) installed
- [Ableton Live 12.3+](https://www.ableton.com/live/) with
[Max for Live](https://www.ableton.com/live/max-for-live/). Live 12.4 or later
is recommended — some features don't work on older versions of Live. Use the
version of Max bundled with Live, or make sure your standalone Max is up to
date.
## 1. Install Ollama
Download and install [Ollama](https://ollama.com/) for your operating system.
## 2. Download a Model
Download a model that supports tools. Some good options include:
- `qwen3.8`
- `gemma4`
- `muse-glimmer`
- `mistral-small3.2`
And the others listed in Producer Pal's built-in chat UI settings. If your
computer can handle it, use the larger variants of a model.
Browse [models with tool support](https://ollama.com/search?c=tools) on the
Ollama website.
## 3. Install the Max for Live Device
Download
[Producer_Pal.amxd](https://github.com/adamjmurray/producer-pal/releases/latest/download/Producer_Pal.amxd)
and drag it to a MIDI track in Ableton Live.
It should display "Producer Pal Running":
## 4. Enable Small Model Mode (Optional but Recommended)
In [the Producer Pal "Setup" tab](/guide/device#setup-tab), enable **Small Model
Mode**.
This provides a smaller, simpler interface optimized for small/local language
models.
## 5. Open the Chat UI
In the Producer Pal device's Main tab, click "Open Chat UI". The built-in chat
UI opens in a browser:
## 6. Configure Ollama
In the chat UI settings:
- Provider: **Ollama (local)**
- URL: `http://localhost:11434`
- Use this default URL when everything runs on the same computer. Change
`localhost` to run Ollama on a different computer. Consult
[the user guide](/guide/chat-ui#local-ai-settings-ollama-bionic) for more
info.
- Model: Your model name (e.g., `qwen3.8` or `gemma4`)
Click "Save".
::: tip Ollama Model Aliases
If Producer Pal says a model like `muse-glimmer` is not installed but you
downloaded `muse-glimmer:30b-mlx`, enter the full name with the tag. A bare
`muse-glimmer` is short for `muse-glimmer:latest`, which is a different build —
the shorthand doesn't find the tagged one you have. Run `ollama list` to see the
exact names.
:::
## 7. Connect
Click "Quick Connect" and say "connect to ableton":
::: tip Local Model Limitations
Local models work best for simple tasks. Complex edits may require more capable
cloud models.
:::
## Model Compatibility
If the model responds with garbled text like `<|tool_call_start|>...` or says it
can't connect to Ableton, the model doesn't support tools. Try a different model
from the [tools category](https://ollama.com/search?c=tools).
## Local Model Tips
- **Set the context length high enough.** When you connect, Producer Pal sends a
sizable set of instructions (roughly 9,000 tokens). Ollama's default context
window (`num_ctx`, 4096) is too small and silently truncates them, which
breaks tool use. Use **16k (16000) or higher** — set it per model in a
Modelfile, or globally with the `OLLAMA_CONTEXT_LENGTH=16384` environment
variable. Larger windows use more memory and can run slower, so experiment to
find the balance.
- **Follow-up turns are often much faster.** The local engine automatically
reuses its cached prompt prefix (your instructions and earlier messages)
between turns, so it usually only has to process your newest message. The
first turn after connecting is the slow one; later turns in the same
conversation can be dramatically quicker. There's nothing to configure. The
size of the speedup depends on the model — some keep the prefix stable across
turns and benefit fully, while others reprocess each turn and see little gain.
## Troubleshooting
If it doesn't work, see the [Troubleshooting Guide](/support/troubleshooting).