Privacy and limits
What leaves the network, and the limits per user
What leaves the network
For the forms that take part, the following is sent to the provider:
- the form names and descriptions, the field labels, help texts, choices and the values being filled in;
- the conversation.
The following is never sent to the provider, the page or the log:
- passwords (the user enters those in the browser), stored credentials and the API key;
- a job’s output, and its stored extravars (the assistant reads a job’s status only, and only when Let the assistant follow its jobs is on in the settings).
Anything under a key that looks like a secret (password, secret, token, api_key, …) is masked before the model or the page sees it.
To keep form data in-house, use your own tenant (Azure OpenAI) or a local model (Ollama), not a public API.
Limits
The following limits are set on the settings page:
- Messages per conversation (20) - after which the user must start a new conversation.
- Tool rounds per message (6) - how many times the model may call a tool before it answers.
- Timeout (60 s) - per call to the provider.
The following limits are fixed: one message at a time per conversation, at most 4000 characters per message, 60 messages per user per hour, and 5 conversations per user. A conversation is discarded after 2 hours without a message.
A summary card with a Launch button can be used once, and expires after 15 minutes: Launch then answers that the summary expired, and the user asks the assistant for a fresh one. Conversations and summaries are kept in memory, so after a restart the user must start a new conversation.