If you build AI into your own applications, you need to be able to explain to your customers what happens to their data. Here you can see at a glance what FlowModels stores for each request, for how long, and what you can delete yourself.
Overview
| What | How long | Who deletes it |
|---|---|---|
| Log of each call (time, model, tokens, runtime, status, identifier) | 400 days | automatically |
| Text of your request and the model's response | 30 days | automatically |
Stored responses from /responses with store: true |
24 hours after completion | automatically |
Responses to an Idempotency-Key |
24 hours by default | automatically |
| Transcription results | 24 hours after completion | automatically, or you via DELETE |
| Uploaded audio files for transcriptions | until completion | automatically |
Files via POST /files |
until you delete them | you |
| Brain data: conversations, memories, end-user files | until you delete them or the end user | you |
| Billing and usage data | permanently for retention reasons | can't be deleted |
The log and the content
For each call, we store a log with time, model, token counts, runtime, status and flowai_request_id. It serves as proof of usage and billing and is kept for 400 days.
We also keep the submitted request content and the response text for 30 days, so support can trace a reported call. After that, we delete both automatically.
You won't see this content anywhere in the interface. The request log shows only the technical details, and GET /usage/requests doesn't return any text either.
Turn off content storage
If you don't want any content stored with us, we can completely turn off the storage of request and response text for your workspace. Contact support for this. After that, the logs contain only the technical details. Keep in mind, though: without content, support can't trace a faulty call later on.
What you control yourself
Without the Brain, we don't keep any conversation history or memories for you. You hold the history yourself in your application.
With the Brain, so with external_identifier, we store conversations, memories and files per end user. This data stays until you delete it:
- individual memories via
DELETE /memories/{id}, see Create, search and change an end user's memories via API - individual files via
DELETE /end_users/{id}/files/{file}, see Upload documents for an end user and use them as context - everything about a person via
DELETE /end-users/{id}or in the interface under End users, see Manage and delete end users via API
When you delete an end user, you get an erasure receipt. Billing and usage rows are kept so your invoices stay correct.
Automatic memories
If you use the Brain, it automatically remembers important details from conversations as memories. This is on for new end users. If you don't want that, turn it off per end user with PATCH /end-users/{id} and "memory_extraction_enabled": false.
The model providers
Your requests go to the provider of the chosen model for processing. That provider's terms apply to how it handles the data. We make no commitments of our own about retention or training at the providers. Which providers we use is stated in the privacy policy on flowmodels.one.
Privacy tips
- Use a neutral ID from your system as
external_identifier, not an email address or a name. - The
userfield only serves abuse tracking. An email address doesn't belong there either. - Delete files via
/filesas soon as you no longer need them. They have no automatic deletion period. - Webhook deliveries don't contain response text. For finished background jobs, they contain only the identifier and status, and your server fetches the result itself afterward.