In this post we discuss how to test the throughput of PromptFlow pf-serve module and key learnings doing so. We explore the impact on throughput and performance the different WSGI and ASGI hosting methods have and the importance of engineering your Python nodes with the async await pattern for I/O.

This post discusses a pattern to prevent a class of prompt injection attacks in LLM-based solutions. It emphasizes the importance of building strong foundational patterns to mitigate risks and avoid potential pitfalls. By implementing this pattern, teams can enhance the security of their tool-based solutions.