Guide
Ollama and Recommended Model Installation Guide
Before using the KOKOCAL Google Sheets AI Assistant, install Ollama and the required models on your local computer. This guide covers Windows, macOS, and Linux setup, recommended model commands, and post-install checks.
The local AI assistant connects to Ollama at http://127.0.0.1:11434.
After connecting, the assistant loads installed models and lets you choose models by task.
To use image attachments, choose a Text, Image capable model as the image attachment model.
Before you start
Ollama is an application for running large language models on your local computer. The KOKOCAL Google Sheets AI Assistant connects from browser JavaScript inside the sidebar to Ollama running on your local PC, not from the Google Apps Script server side.
The Ollama connection endpoint is fixed to http://127.0.0.1:11434. If the assistant cannot connect, check whether Ollama is running, whether your browser allows local access, and whether OLLAMA_ORIGINS is configured.
Install Ollama
Open the official Ollama site and use the Download button in the top-right or center of the top page to go to the download page. On the download page, choose the macOS, Linux, or Windows tab for your computer.
Install with the installer
On Windows and macOS, downloading and running the installer from the download page is enough to install Ollama. If you installed with the installer, you do not need to run an additional install command on the same computer.
After installing, continue to recommended model installationInstall with a command
If you are comfortable with terminal operations, or if you use Linux, you can install Ollama with the command shown in the official documentation. Choose either the installer method or the command method; you do not need both.
The OS cards below summarize the commands for command-based installation. Model files can be several GB to more than ten GB, so also check available storage.
Windows
Official stepsUsually, download the installer from Download for Windows on the download page and run it. The command below is an alternative when you prefer installing from PowerShell.
irm https://ollama.com/install.ps1 | iexmacOS
Official stepsUsually, download Ollama.dmg from the download page and place it in Applications. The command below is an alternative when you prefer installing from Terminal.
curl -fsSL https://ollama.com/install.sh | shLinux
Official stepsOn Linux, command-based installation with the official install script is the standard path. In systemd environments, running Ollama as a service is recommended.
curl -fsSL https://ollama.com/install.sh | shInstall recommended models
Install only the models you need, starting with the most important ones. With just the standard model gemma4:12b-it-q4_K_M or the lightweight model qwen3.5:9b, you can try questions, formula creation, and ordinary table processing. Add models for CSV/Text attachments or image attachments when needed.
Standard model set
| Task | Recommended model |
|---|---|
| Normal and integrated processing | gemma4:12b-it-q4_K_M |
| CSV/Text attachment processing | ministral-3:14b-instruct-2512-q4_K_M |
| Image processing | qwen3-vl:8b |
Lightweight model set
| Task | Recommended model |
|---|---|
| Normal and integrated processing | qwen3.5:9b |
| CSV/Text attachment processing | qwen2.5vl:7b |
| Image processing | qwen2.5vl:7b |
Run the model installation commands in PowerShell on Windows, or in Terminal on macOS and Linux. Copy a command, paste it into the terminal window, and press Enter. For PowerShell setup, see How to Install PowerShell 7 on Windows.
Standard normal and integrated processing model
gemma4:12b-it-q4_K_M
The first recommended model for questions, formula creation, and ordinary table processing.
ollama pull gemma4:12b-it-q4_K_MStandard CSV/Text attachment model
ministral-3:14b-instruct-2512-q4_K_M
A candidate for reading structured attachments such as CSV, TSV, text, and Markdown.
ollama pull ministral-3:14b-instruct-2512-q4_K_MStandard image processing model
qwen3-vl:8b
A Vision-capable model for PNG, JPG, WebP, and other image attachments.
ollama pull qwen3-vl:8bLightweight normal and integrated processing model
qwen3.5:9b
A candidate for lighter questions and ordinary processing. Depending on your environment, it may be easier to run than the standard model.
ollama pull qwen3.5:9bLightweight CSV/Text and image processing model
qwen2.5vl:7b
A lightweight candidate for trying CSV/Text attachment processing and image processing.
ollama pull qwen2.5vl:7bImage-capable qwen3-vl models may depend on the Ollama version. Before using image attachments, update Ollama to the latest version and run the connection test.
Post-install checks
After installation, confirm that Ollama and the models are recognized from your terminal. Use PowerShell on Windows, or Terminal on macOS and Linux.
Check the Ollama version
ollama -v
List installed models
ollama list
Check the Ollama API
Open http://127.0.0.1:11434/api/tags in your browser
Run a simple model test
ollama run gemma4:12b-it-q4_K_M
Assistant-side checks
- 1
Open the sidebar
Reload Google Sheets and open the local AI assistant from the KOKOCAL menu.
- 2
Run connection
In Ollama settings, click Connect and confirm that installed models appear in the dropdown menus.
- 3
Save AI model settings
Choose models for normal processing, CSV/Text attachments, image attachments, and integrated processing, then click Save and close.
- 4
Run Ollama response diagnostics
In diagnostic mode, send a minimal prompt and confirm that a response is returned.
- 5
Try question mode
Ask about a function name or syntax, choose an answer style, and run Ask AI.
Official resources
Installation methods and model tags can change. Check the official Ollama pages for the latest information.
Guide
Related Guides
After preparing Ollama and the models, review the Apps Script setup steps and the notes for using multiple scripts together.
Google Sheets Local AI Assistant Guide
Learn how to ask questions, process selected ranges, and create formulas from a sidebar.
Apps Script Basics Guide
Review the basic steps for pasting generated code into Apps Script.
Using Multiple Scripts Together
Check notes for adding multiple generated scripts to the same Apps Script project.