Today, we're excited to announce that Fiddler Guardrails is now available natively in NVIDIA NeMo™ Guardrails. This integration brings the industry's fastest guardrails directly to the NVIDIA ecosystem, providing <100ms response times to moderate LLM conversations for hallucinations, toxicity, and jailbreak attempts — all while maintaining enterprise-grade security in your own environment. And now, you can experience these capabilities firsthand with our new Fiddler Guardrails free trial.
By integrating with the NeMo Guardrails orchestration platform, this collaboration with NVIDIA reflects our commitment to helping more enterprises launch accurate and safe LLM applications into production.
Industry’s Fastest Guardrails in NVIDIA NeMo Guardrails
NVIDIA NeMo Guardrails is a scalable AI guardrail orchestration platform that enables organizations to implement and orchestrate multiple AI guardrails, ensuring the safety, security, accuracy, and topical relevance of their LLM interactions across popular generative AI frameworks.
Today, the native NVIDIA NeMo Guardrails integration further bolsters our partnership to jointly provide our enterprise customers with:
- Industry-Leading Speed: The only solution in the NeMo ecosystem that moderates hallucinations, toxicity, and jailbreak attempts with <100ms response times.
- Cost-Effectiveness: Reduced costs by leveraging Fiddler Trust Models, delivering a cost-effective Guardrails solution.
- Secure Environment: Fiddler Guardrails is the only guardrails within the NVIDIA ecosystem that can be deployed in the customer’s cloud and VPC environment.
- Enterprise Scalability: Scale LLM applications with Fiddler Guardrails, processing 5+ millions of events daily.
The native integration offers engineering teams a seamless way to implement Fiddler Guardrails within NeMo Guardrails with minimal setup requirements.
- Simply obtain a Fiddler platform key, set it as an environment variable, and update your config.yml file with your endpoint and desired hallucination, toxicity, and jailbreaking thresholds.
- Once configured, Fiddler Guardrails will moderate inputs and outputs from your LLMs without disrupting service during API failures.
- You can easily customize the moderation sensitivity by adjusting the threshold values to match your specific use case requirements.
- More information on how to set up the Fiddler Guardrails integration is available on the NVIDIA NeMo Guardrails GitHub.
LLM Metrics: The Foundation for Observable, Secure AI
Metrics are the essential foundation for ensuring LLM applications can be properly observed and governed. Without comprehensive metrics, enterprises lack the visibility needed to understand LLM performance, detect issues, and ensure alignment with business objectives.
That's why Fiddler has built the Fiddler Trust Service, which is powered primarily by Fiddler Trust Models that track 50 out-of-the-box and custom metrics. Based on the metric performance, Guardrails, LLM and agentic monitoring solutions help protect your LLM applications.
The Fiddler Trust Service is powered by two different scoring approaches:
The Fiddler Trust Service delivers 50 LLM metrics to enable enterprises to monitor key hallucination, safety, and domain-specific metrics, ensuring AI observability and security for safe LLM deployments in production environments.
Strategic Partnership: Fiddler & NVIDIA Collaborate to Secure Enterprise LLM Deployments
As a key partner, Fiddler has collaborated closely with NVIDIA from the inception of NeMo Guardrails and NVIDIA Inference Manager (NIM). This partnership leverages Fiddler’s deep AI expertise to create complementary technologies that address the core challenges of GenAI and LLM deployments: speed, cost, quality, security, and scalability.
Our journey began with an integration that captures prompts, responses, metadata, and rails executed in NeMo Guardrails and publishes them to the Fiddler AI Observability and Security platform. This helped AI and application engineers gain insights into rails execution, define rich alerts for rail activities, and monitor issues like hallucination, safety violations, prompt injection attacks, and operational metrics.
Most recently, we announced our integration with NVIDIA NIM, which ensures efficient scaling of secure LLM deployments by logging and routing processed prompts to Fiddler for monitoring, actionable insights, and enhanced operational oversight.
Experience the Fastest Guardrails for Free
We're excited to offer a free trial of Fiddler Guardrails, giving you hands-on experience with the industry’s fastest solution, delivering <100ms response times* to protect your LLM applications against hallucination, toxicity, and jailbreak attempts.
Free Trial Includes:
- 14-day access to Fiddler Guardrails
- 200 API requests per day
- Real-time moderation against hallucination, toxicity, and jailbreak attempts
- Comprehensive developer documentation
- Guided walkthrough to get you started
Getting Started:
- Sign up for the trial at fiddler.ai/guardrails-trial
- Get your API key
- Start protecting your LLM applications
Ready to take the industry's fastest Guardrails for a spin? Sign up for the free trial today.
