Tests the prompt injection filter against various injection attempts. This test uses 100 strings generated by OpenAI's o3 model to simulate both RISKY and BENIGN prompts in this category.
Tests the sexual content filter against various prompts. This test uses 100 strings generated by OpenAI's o3 model to simulate both RISKY and BENIGN prompts in this category.
Tests the violence content filter against various prompts. This test uses 100 strings generated by OpenAI's o3 model to simulate both RISKY and BENIGN prompts in this category.
Tests the code output filter against various responses containing code snippets. This test uses 100 strings generated by OpenAI's o3 model to simulate both RISKY and BENIGN prompts in this category.
This test runs all prompts from tests/data/testprompts.txt through the main chat UI, using the Oscar (Jailbroken, SimpleBot) personality and all local guardrails filters (except rate limit). It is designed to quickly verify that SimpleBot and the local guardrails filters are working as expected. Results will appear in the main chat window.
Note: This test does not use any OpenAI-powered filters and will not rate limit requests.
This test runs all prompts from tests/data/testprompts.txt through the main chat UI, using the Bob (Tech Support, GPT) personality and all OpenAI-powered filters. It is designed to quickly verify that GPT and the AI-powered guardrails filters are working as expected. Results will appear in the main chat window.
Note: This test requires an OpenAI API key to be set in the Preferences panel.