Hi folks, I am developing ai project for my software. I am using azure ai model so i have azure content filters for saftey. I want to add the nemo guardrails in my project to enhance more security. My project is based on agentic model not RAG model. But following are my conditions/ limitations:
- I dont want to host the nemo guardrails as seperate server or dont want to use any third party api service for nemo guardrails.
- I want to use my defined llm in nemo guardrails not the other llm.
- I just want to use the nemo guardrails meachanism to define the rails file (colang file) and use it. NO extra functionality. It should use my defined llm for processing. For example i will define some input and output rails using colang files, then use my own llm model for validation (no nemo/third party llm)
- When i installed the nemo guardrails in my python project, i saw there are so many dependencies required for project like uvicorn, watchdog and many more. Honestly why these dependencies required when i just want to use nemo guardrails in simple way.
a. Is there any better way to install nemo guardrails in python project?
b. I dont wnat to maintain these extra dependencies in my project if i dont use them?
c. does nemo guardrails supports only specific dependencies download in installation and work on it?