AI Conversational Sales & Support Agent (n8n + OpenAI)

Unlock Dynamic Customer Interactions with this AI Agent

This AI Agent, 'Jovelino', acts as a sophisticated chatbot for your business, powered by OpenAI Assistants. It's engineered to understand user inquiries through natural language, provide personalized responses by initially processing lead data if available, retrieve real-time product or service details from your MySQL database, and consult external knowledge bases or APIs via HTTP requests. The agent maintains conversation context using a PostgreSQL-backed memory, ensuring seamless and intelligent interactions. Its core abilities include conversational AI, data retrieval, and tool utilization to handle sales inquiries, provide customer support, and guide users effectively.

Key Features & Benefits

• Intelligent Conversations: Leverages OpenAI Assistants for natural, context-aware dialogue, enhancing user engagement.
• Personalization Engine: Can process initial lead data to prime the AI's memory, enabling more tailored interactions from the start.
• Dynamic Data Access: Connects to MySQL databases using a dedicated tool to fetch up-to-date product/service information on demand during the conversation.
• External Knowledge Integration: Employs tools to query external APIs and knowledge bases, providing comprehensive answers beyond its core training.
• Persistent Chat Memory: Utilizes PostgreSQL to remember past interactions within a session, leading to more coherent and contextually relevant conversations.
• Extensible Toolset: The main OpenAI Assistant's capabilities are augmented with custom tools for database lookups and API calls, making it highly adaptable to specific business needs.
• Automated Lead Nurturing & Support: Frees up human agents by handling initial queries, qualifying leads, and providing 24/7 assistance.

• An n8n instance (Cloud or self-hosted).
• OpenAI API Key with access to assistant-compatible models (e.g., GPT-4, GPT-3.5 Turbo).
• Credentials for a PostgreSQL database (for chat memory storage).
• Credentials for a MySQL database (for product/data lookup).
• (Optional) URLs and any necessary authentication details for external APIs or knowledge bases used by the HTTP Request tools.

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Configure the 'Chat Trigger' node: Customize the initialMessages (e.g., greeting from 'Jovelino') as needed.
4. Configure the 'OpenAI' nodes (named 'OpenAI' and 'OpenAI2'):
   • Select or create your OpenAI API credentials.
   • The workflow uses specific Assistant IDs (asst_numdCoMZPQ6GwfiJg5drg9hr for initial lead data processing and asst_x2qfc7EuoPv7XGOL84ClEZ3L for the main conversation). You will likely need to recreate these Assistants in your OpenAI account, defining their instructions, and enabling the necessary tools (e.g., for database queries, knowledge base access). Ensure the tool descriptions and functionalities match what the n8n workflow expects.
5. Configure the 'Postgres Chat Memory' nodes ('Postgres Chat Memory' and 'Postgres Chat Memory1'):
   • Select or create your PostgreSQL credentials.
   • Ensure the tableName (e.g., 'aimessages') is appropriate. Note the different contextWindowLength settings for initial data priming vs. ongoing conversation.
6. Configure the 'Products in Daatabase' (MySQL Tool) node:
   • Select or create your MySQL credentials.
   • Review and adapt the SQL query to match your database schema and the parameters your OpenAI Assistant tool is configured to provide (e.g., {{ $fromAI("cityQuery") }}).
7. Configure the 'Knowledge Base' (HTTP Request Tool) node:
   • Update the url and any required authentication. Customize the toolDescription to accurately reflect its purpose for the AI Assistant.
8. Configure the 'External API' (HTTP Request Tool) node:
   • Update the url, method, jsonBody, and authentication as needed. Customize the toolDescription.
9. Review the 'If' node logic: This node routes the flow based on the presence of leadData. Ensure this aligns with how you plan to pass initial information (e.g., from a CRM or form) to the chat.
10. Review 'Edit Fields' nodes ('Edit Fields1', 'Edit Fields2'): These format the input for the AI. Adjust the templating if your input data structure or desired AI priming message differs.
11. Test the workflow thoroughly by simulating conversations, including scenarios with and without initial leadData, and testing all integrated tools.
12. Activate the workflow.