Security

I think using open-source llms and embed models improves and make your data stay at your side which is a big PLUS.


Apart from this you can use nemo-gaurdrails to verify user input , mask any important details like card numbers etc in case going with closed llms like openai or azure.

https://github.com/NVIDIA/NeMo-Guardrails

Llamaindex implementation: https://medium.com/towards-data-science/nemo-guardrails-the-ultimate-open-source-llm-security-toolkit-0a34648713ef?sk=836ead39623dab0015420de2740eccc2

I didn't know about this. Going to take a look. Thnks for sharing!

Hello @memorIA , I'm Riccardo, Co-Founder of Glaider. I believe Glaider could be a valuable asset for your needs. Our platform is purpose-built to safeguard your applications from:

• Prompt Injection attacks: shields against direct and indirect attacks that can manipulate your prompts.
• Data leakage: ensures the safety of sensitive data when connecting to critical resources.
• Toxic language: keeps outputs aligned with ethical standards and company policies.

Integration is a breeze! With our API, you can secure any LLM with just a few steps.

Want to give it a try? DM me or reach out at riccardo@glaider.it for a trial, we’re just starting and would love some solid feedback!