AI-Powered n8n Workflow Credential Search

• Stores n8n workflow and credential mappings in a local SQLite database.
• Utilizes n8n's API to fetch workflow and credential information.
• Employs an AI agent (Langchain with OpenAI) for natural language querying.
• Allows complex queries like 'Which workflows use Slack and Google Calendar?'
• Provides a mechanism to link back to specific workflows via generated URLs.

Step 1. Store Workflows Credential Mappings to Database

1. Manual Trigger: Click "Test workflow" to initiate the process.
2. n8n Node: This node queries your n8n instance using its API to retrieve all workflows.
3. Map Workflows & Credentials: This node processes the data from the n8n node, extracting the workflow_id, workflow_name, and a structured list of credentials from each workflow's nodes.
4. Save to Database: This Python code node creates and populates a SQLite database named n8n_workflow_credentials.db with the extracted information.

Step 2. Use Agent as Search Interface

1. Chat Trigger: This node is where you'll interact with the AI agent. It listens for incoming messages.
2. OpenAI Chat Model: Configured with your OpenAI API credentials, this node provides the language model for the agent.
3. Window Buffer Memory: Manages the conversation history for the AI agent.
4. Query Workflow Credentials Database: This toolCode node provides the AI agent with access to query the SQLite database created in Step 1. It's configured with the database schema and instructions on how to use it.
5. Workflow Credentials Helper Agent: This agent node orchestrates the interaction between the user's input, the OpenAI model, and the Query Workflow Credentials Database tool. It interprets your natural language queries and translates them into SQL queries to retrieve the relevant information. It can also generate links to workflows if requested.

Example Queries:

• "Which workflows are using Slack and Google Calendar?"
• "Show me workflows that use Airtable."
• "List workflows that have 'notification' in their name."