A datapoint is a single completion by the machine learning model (such as GPT-4) which is stored in Humanloop and enriched with additional metadata like initial inputs, user feedback, correction through the UI or API.

When you sign up for your trial of Humanloop you'll have access to all features of your chosen plan. At any point during the trial you may choose to cancel your subscription without any charges from our side.

It can take minutes to create a new AI endpoint from the playground. To set up an end-to-end system where your AI features in production are using all Humanloop features around logging and feedback collection may take a few hours. We have guides to help you get started quickly.

Fine tuning is customising the model by further training on your own data. It can significantly improve performance.

Your data is encrypted, private to your account and never shared with third parties. Our platform is regularly penetration tested to ensure we meet the best security standards. Our enterprise plans offer additional dedicated cloud deployment options.

Yes, you can export all your data at any time.

Yes – you bring your own API key from the AI providers and pay for your usage with them separately. This means that you can use any custom terms you have with them and any fine tuned models created are accessible outside of Humanloop.

JSON and CSV formats are supported for uploading data via the user interface or the API. See our documentation for further details.

Our APIs and SDK allow you to use our tools with your own model provider where relevant. See our documentation for more details.

Join our community Discord server or send us an email at team@humanloop.com.