It consists of two main components :

- Continuous red teaming, to automatically generate dynamic and context-aware attacks, ensuring comprehensive risks coverage

- Evaluation runs, to continuously enrich your golden dataset and align your testing strategy with business context, helping prevent regressions and maintain quality standards

This is intended for both the technical teams developing LLM agents (Data Scientists, ML Engineers, etc.), the product teams responsible for these agents, and the subject matter experts in the business domains addressed by these agents.

With the Giskard Hub :

- You save time by drastically cutting down on tedious, manual and non-collaborative testing

- You avoid sacrificing quality through overly restrictive bots, with excessive moderation

- You ensure unbiased evaluations, independently run by a GenAI testing specialist

- You significantly reduce the risk of reputational damage, financial loss or legal issues for your company