Finally — An AI Agent You Can Actually SELL (Free n8n Template)

Key Concepts

• RAG (Retrieval-Augmented Generation) Pipeline: A system that retrieves relevant information from a knowledge base (like API documentation) and uses it to generate more accurate and context-aware responses.
• Naden: A no-code platform used to build and automate workflows, including AI agents.
• vectoriz.io: A platform that helps build RAG pipelines by converting unstructured data into vector embeddings and storing them in a vector database.
• Vector Embeddings: Numerical representations of data (text, images, etc.) that capture their semantic meaning, allowing for efficient similarity searches.
• Slack Bot: An automated application that interacts with users within Slack channels.
• API Documentation: Technical manuals that describe how to use and integrate with an Application Programming Interface (API).
• HTTP Request Node: A component in Naden that allows sending HTTP requests to external APIs.
• AI Agent: An autonomous program that uses AI to perform specific tasks, such as answering questions or providing support.

Building an Intelligent RAG AI Agent for Technical Support

Problem Statement

Sales and support teams often struggle to quickly answer technical questions, especially those related to API documentation and integrations. Salespeople get stuck in meetings, and support agents waste time digging through documentation.

Solution Overview

The video demonstrates how to build an intelligent RAG AI agent using no-code tools (Naden and vectoriz.io) to address this problem. This agent connects directly to API documentation, stays up-to-date automatically, and can instantly answer technical questions through Slack.

Naden Workflow Setup

1. Importing the Template: The presenter provides a free Naden template that can be downloaded from the AI Workshop Light community (link in description). The template is imported into Naden using the "Import from file" option.
2. Workflow Structure: The workflow is set up to listen for Slack triggers and interact with vectoriz.io's API documentation.
3. Slack Trigger: The workflow is triggered when the bot is mentioned in a Slack channel (e.g., "@NadenBot"). The "Trigger on bot or app mentions" option is used to keep the Slack channel clean.
4. HTTP Request Node: This node sends a POST request to the vectoriz.io API to retrieve relevant information based on the user's question.
   • URL: The endpoint URL for the vectoriz.io retrieval API is obtained from the "Connect" section of the RAG pipeline in vectoriz.io.
   • Headers: The request includes an "Authorization" header with a token generated in vectoriz.io. The token can be set to expire after a specific duration.
   • Body: The request body is a JSON object containing the question, the number of results to retrieve ("num_results"), and a re-ranking option ("rerank"). The question is dynamically populated with the text from the Slack trigger.
5. AI Agent (Chat Model): The retrieved content is passed to an AI agent (using a chat model like GPT-4.0-mini) with a specific prompt.
   • Prompt: The prompt instructs the AI agent to generate a concise and helpful response based on the user's question and the retrieved content, formatted for Slack. The prompt includes instructions for a friendly tone and Slack-friendly formatting.
   • Simple Memory: The "event_ts" (timestamp) from the Slack trigger is used for memory management to prevent errors.
6. Slack Response: The AI agent's response is sent back to the Slack channel as a reply to the original message.
   • Message Type: Set to "Simple Text Message."
   • Text: The text is the output from the AI agent.
   • Reply Message: The "reply to message" option is enabled, and the "event_ts" from the Slack trigger is used to ensure the response is a reply to the specific question.

Building the RAG Pipeline in vectoriz.io

1. Creating a New RAG Pipeline: A new RAG pipeline is created in vectoriz.io.
2. Source (Web Crawler): The source is set to "Web Crawler" to ingest API documentation from a URL.
   • Adding a New Container: A new container is created with a name (e.g., "vector docs test") and the URL of the API documentation (e.g., "docs.vectoriz.io").
   • Allowed URLs: Additional URLs or prefixes can be added to allow the crawler to access specific sections of the documentation.
3. Extractor: The extraction strategy is set to "Fast" for simple and fast extraction. The chunking strategy is left as "Default."
4. Embed: The built-in vectoriz.io embedding model (OpenAI v3 small) is used.
5. Vector Database: The built-in vectoriz.io vector database is used.
6. Deploying the RAG Pipeline: The RAG pipeline is deployed, which initiates the process of crawling the URL, extracting the content, creating vector embeddings, and storing them in the vector database. The platform provides real-time updates on the progress.

Testing the Workflow

The presenter demonstrates the workflow by asking the bot a question in Slack ("What is a RAG pipeline?"). The bot retrieves the answer from the vectoriz.io documentation and responds in a reply thread, including a link to the source documentation.

vectoriz.io Configuration Details

• Connect Endpoint: The endpoint URL for the retrieval API is found in the "Connect" section of the RAG pipeline.
• API Token: An API token is generated in vectoriz.io and used for authentication in the HTTP request node. The token can be set to expire after a specific duration.
• Sync Schedule: The sync schedule determines how often the RAG pipeline is updated with new or changed content. On the free plan, this is manual. Paid plans allow for automatic updates (e.g., every 12 hours, daily).

Benefits and Use Cases

• Improved Efficiency: Sales and support teams can quickly access accurate information, reducing the time spent searching for answers.
• Enhanced Customer Service: Customers receive faster and more accurate responses to their technical questions.
• Reduced Burden on Engineering Teams: Sales and support teams can answer many questions themselves, reducing the need to involve engineering teams.
• Monetization Opportunity: This solution can be sold to companies that struggle with technical support.

Monetization and Further Learning

The presenter mentions a paid community and a "Launch Your AI Agency with Naden" course that provide more in-depth training on Naden, AI agent development, and monetization strategies. The course covers topics such as pricing services, creating contracts, and selling to clients.

Conclusion

The video provides a practical guide to building an intelligent RAG AI agent using no-code tools. This agent can significantly improve the efficiency of sales and support teams by providing instant access to technical information. The presenter emphasizes the monetization potential of this solution and encourages viewers to explore further learning resources.