Troubleshooting

"Ollama is not running" / no models in Setup

Start Ollama (ollama serve, or open the Ollama app), then ollama pull gemma4:e2b, and press Check the model again.

A run is very slow

A local model on a CPU takes minutes per step; that's normal. A smaller model (gemma4:e2b) or a hosted model with your key is faster.

"tool … is not available" (Spreadsheet analyst)

tableski isn't connected or isn't running. Connect it in Setup, or start your local tableski; kowalski keeps retrying and picks it up.

The Workbooks card says the tableski runs locally

A local tableski has no upload: start it with your file (tableski --file your.xlsx).

A run says Needs you

A step wants to run a command or change files. Open the run and press Approve and continue, or cancel it. Always allow stops it asking for that step of that horde; revoke it in Setup → Always allowed.

A run failed

Open it from Runs: the failed step and the reason are at the top. Common causes: the model isn't available, a page couldn't be fetched (the run notes which), or a step timed out.

Web search fails

DuckDuckGo limits automated searches. Add a Brave, Staan or Tavily key in Setup.

Save in Setup fails

The message is next to the Save button. A notes vault must be an existing folder; the files folder is created for you.

Where are my files?

Everything is beside ~/.config/kowalski/config.toml: the built-in hordes (builtin-hordes/), your hordes (hordes/) and the database (db/). The Mac app logs to ~/Library/Logs/Kowalski/kowalski.log.