Troubleshooting

"Context directory does not exist". --context points at a folder that is not there. Only ronsel start --context <path> creates one; every other command expects it, so mkdir it first, or cd into the folder you mean and run ronsel with no flag at all.

npm run ronsel says "ronsel: command not found", or "Missing script". The script comes from the package.json that ronsel start wrote, and the command it runs comes from node_modules. Run npm install in the folder if the dependencies are not there — a fresh clone, or a start --no-install — and npx ronsel start in it if there is no script at all. It is safe to run again: it adds what is missing and touches nothing else.

ronsel start says the folder "already holds other files". It is not a context, and it is not empty either, so the tool asks before writing into what may be somebody's project. Answer yes if the folder is the one you mean; the examples still only go into an empty one. A run nobody is watching prints the line and goes ahead.

ronsel start did not add its script to package.json. There was a ronsel script in it already, and the tool never overwrites one — it prints what the script should run so you can decide. Everything else it writes is additive too: your keys, ordering and indentation survive it.

The flow tree is empty, and the examples never appeared. Flows are read from flows/ in the context folder named in the top bar, so check that folder is the one you meant — without --context it is the directory you ran the command from. The examples are only ever written into an empty context: a folder that already holds something is served exactly as it is, so make flows/ yourself there, or start the tool on an empty folder to see them. Use Refresh in the + menu after adding files by hand.

"Application not found" when running a step. The application value must match a folder name under applications/. Open the sidebar and check the exact name, and that the method is exported.

"Missing environment file" when starting a flow. An application the flow uses has no env/<environment>.env. Only the applications of that flow are asked for one, and the message names each missing file. Create them from the Environment variables card on the home page, import a teammate's export there, or pick an environment the flow's applications do have. See Environments.

"View not found" on the CLI. --view takes the name or the slug of a view of views.yaml, and the error lists the ones that exist. The CLI button of a folder view writes the right command.

A browser step fails with "Executable doesn't exist" or "is not installed". Playwright is installed, its browsers are not: run npx playwright install chromium on the machine running the flows, with --with-deps on a bare Linux box.

A step fails only sometimes. Add retry: { times, delay } for genuinely eventual behaviour. If the data is the problem, remember that replacers generate new values on every run: assert with $expr: instead of exact values.

A value from an earlier step arrives empty. Only the steps above have written to the memory, and only parameters are templated. In a test, use $expr: value === memory.key. See Passing data between steps.

The AI section says "Not configured yet". Pick a provider and save a key first, then use Test connection: it does a real round trip, so it also catches a wrong model name or an Ollama that is not running.

No Xray information on a flow. The integration has to be configured and the flow needs xray.testKey in its frontmatter. When Jira cannot be reached, the key is shown as plain text with the error on hover.

Something looks wrong in the UI. Reload the page: unsaved editor content is discarded, which is also the way to throw away an AI edit you do not want.

Nothing else works. Run the flow from the CLI with --debug to see the environment as the tool sees it, and open an issue with the output. See Getting more help.