Home
Your comprehensive resource for deploying, customizing, and scaling AI safety testing for LLM-based applications.
Whether you’re a CI/CD developer, application owner, test developer, or compliance professional, these guides will help you harness Moonshot’s full potential—from seamless pipeline integration to creating custom tests tailored to your use case.
Before jumping straight into the technical guides, below are some user personas that we think Moonshot will be helpful towards. Do take a moment to consider which user persona you belong to!
If you are a CI/CD developer tasked to run safety tests on your LLM-based AI applications as part of your CI/CD workflow before production, you've come to the right place!
- Check out this deployment guide to understand how you can easily deploy Moonshot in any CI/CD pipelines.
- For organizations using AWS CodeBuild, you can check out this guide instead.
- If you are interested to contribute guides for other CI/CD platforms, kindly contact our Support Team at [email protected].
- There will be certain configurations required to operationalize Moonshot in your pipeline, so be sure to check out the configuration guide here and user guide to run Moonshot test.
As an Application Owner, ensuring the safety of your LLM-based AI applications for your users will be an utmost priority. But we know that choosing the right safety tests for the right application use case can be immensely challenging -- and getting it wrong carries real reputational and even financial risks!
As such, the first batch of tests we've included are meant for testing Q&A-type applications -- if your application use case is for users to input a question and the LLM to generate an output, the tests here will be right up your alley!
- For a start, check out the
Test Methodologysection in our Wiki here to understand the different test cases for each of the four risk areas listed below, as well as the methodology for deriving the corresponding tests: - You can also check out this developer guide to understand how you can quickly
git cloneour repository and test out the tool! - We've made it simple for you to use Moonshot and configure the tests to fit your use case:
- To connect to your application end point, you can check out this create custom endpoint guide.
- To configure your tests, you can check out this configuration guide.
- To run the tests, you can check out this user guide.
Are you an AI practitioner tasked to develop tests specific to the application use case in your organization? We've created the following to assist you:
- Expert Guide to develop your own Moonshot-compatible custom benchmark tests and red teaming agents.
- Every test requires proper evaluation to be effective, so be sure to check out this guide.
- Other guides that you may find helpful:
- To connect to your application end point, you can check out this create custom endpoint guide.
- To configure your tests, you can check out this configuration guide.
- To run the tests, you can check out this user guide.
For folks from the audit or internal compliance team, you are not left out! We have also developed a web-based application that is aligned with the AI Verify Testing Framework -- Process Checks for Generative AI. This application enables you to assess the responsible implementation of AI system against 11 internationally recognised AI governance principles and generate a summary report for audit and validation.
For more information, be sure to check out the framework here.
This application is available as a separate Docker image that compliance folks can easily download and deploy in your laptop.
- To understand how you can use the Process Checks application, you can check out this onboarding guide.
- Alternatively, you can download the Quick Start Guide here to get yourself up and running.
Alternatively, you can go straight to our onboarding and users guides under the Moonshot User Guides section.