PromptArmor is a service which checks LLM inputs for adversarial content before a completion is made. PromptArmor returns in realtime faster than LLMs, blocks 99% of known threat vectors, and takes less than 15 minutes to integrate.

PromptArmor integration is included in Baserun’s Python SDK as an Automatic Evaluation.

To use PromptArmor, first set the PROMPTARMOR_API_KEY environment variable to your PromptArmor API key. Similar to other automatic evaluation features, you can then use the baserun.evals.check_injection() to evaluate a prompt using PromptArmor.

For more details on how our evaluations work, see the Automatic Evaluation documentation for more details.