n8n AI Safety Suite with Groq LLM and Guardrails

This n8n workflow demonstrates a robust AI safety suite, integrating Groq's fast language model with multiple layers of Langchain Guardrails. It's designed to process an initial input, apply several distinct guardrail checks, and then use a Groq LLM for processing, ensuring that the AI interaction adheres to predefined safety and content policies.

What it does

This workflow provides a comprehensive example of how to implement multi-layered AI safety checks before and after interacting with a large language model.

1. Manual Trigger: Initiates the workflow upon manual execution.
2. Edit Fields (Input): Sets up an initial input, likely representing a user prompt or data to be processed.
3. Guardrails (Input): Applies the first layer of safety checks to the initial input. This could involve sanitizing data, checking for PII, or detecting harmful content.
4. Split Out: Divides the output from the first guardrail into individual items, allowing for parallel or sequential processing of each item.
5. Groq Chat Model: Processes the (now guarded) input using the Groq LLM, generating a response.
6. Guardrails (Output): Applies a second layer of safety checks to the output generated by the Groq LLM. This ensures the LLM's response itself is safe and compliant.
7. Sticky Note: A placeholder for comments or documentation within the workflow, indicating the purpose of the "Input" section.

Prerequisites/Requirements

• n8n Instance: A running instance of n8n.
• Groq API Key: An API key for the Groq language model, configured as a credential in n8n.
• Langchain Guardrails: Ensure the @n8n/n8n-nodes-langchain package is installed and configured in your n8n instance to utilize the Guardrails node.

Setup/Usage

1. Import the Workflow: Download the provided JSON and import it into your n8n instance.
2. Configure Credentials:
   • Set up your Groq API Key as a credential within n8n.
3. Customize Nodes:
   • Edit Fields (Input): Modify the data in this node to define the initial input you want to test with the AI safety suite.
   • Guardrails (Input) and Guardrails (Output): Configure the specific guardrail policies within these nodes according to your safety requirements (e.g., PII detection, content moderation, prompt injection prevention).
   • Groq Chat Model: Select the desired Groq model and any specific parameters for your LLM interaction.
4. Execute Workflow: Run the workflow manually to observe the multi-layered guardrail checks and Groq LLM processing in action.