About AI
What models you can call, how a call is counted, and what happens to what you send.
AI lets your code call open models, such as GPT OSS and Nemotron, through one API. It belongs to a project like every other product, and uses the same API keys, access levels, and bill.
One API, compatible with OpenAI’s
The API takes the same requests and returns the same answers as OpenAI’s Chat Completions API, with streaming, images, and tool calling. To use OpenAI’s SDKs, change the base URL to your project’s, and the API key to a Digitospace key. See Use OpenAI’s SDKs.
Models have names of their own, such as openai/gpt-oss-120b. A name never changes, so your code keeps working when we change how a model is run. See Models for every model, what it accepts, and how much it can read and write.
How calls are counted
Every call is counted in AI units. Each model has a number of units for every million tokens it reads and every million it writes, so you can work out what a call uses before you send it. Every answer says how many units it used. See AI units.
Limits
Your plan sets how many units a project can use each month, and how many requests and tokens it can send each minute. You can also set a lower monthly limit of your own, so a mistake or a leaked key can never use more than you meant. A call that would go over the month’s units is refused before it reaches the model. See Track usage and set limits and Limits.
What happens to what you send
- We do not keep your prompts or the answers. We keep a record of each call for 90 days: when it ran, the model, who made it, and the tokens and units it used, so you can see your usage.
- Each model runs in the country Models names for it: Germany, in the EU, for GPT OSS, and the United States for the Nemotron preview models. Each answer says where it ran, in the
X-Ai-Processed-Inheader. - What you send is not used to train models.
Who can do what
An API key with the AI permission Call models can send chat calls and list models. View usage and limits lets it read the project’s usage and limits. Changing a project’s own limit is kept for people with Developer access or more; no API key can do it. People with View access can see usage and limits. See API keys and Team members and access.
Last updated