This workflow integrates WordPress content with a Retrieval-Augmented Generation (RAG) pipeline to create an intelligent chatbot powered by OpenAI and Supabase. It fetches posts and pages from your WordPress site, converts them into embeddings, stores them in a vector database, and enables real-time chat functionality where users can ask questions and receive context-aware answers directly on your website.
✨ Features
- WordPress Content Integration
- Automatically fetches and updates posts and pages.
- Filters only published and unprotected content.
- RAG with Supabase Vector Store
- Converts content into embeddings using OpenAI’s
text-embedding-3-smallmodel. - Stores embeddings in Supabase for semantic search.
- Supports upserts (insert new or update existing documents).
- Converts content into embeddings using OpenAI’s
- Smart Content Processing
- HTML → Markdown conversion.
- Token splitting for optimized embedding generation.
- Metadata tagging (title, URL, content type, publication & modification dates).
- Chatbot Functionality
- Webhook-based chat trigger for real-time user queries.
- Retrieves matching documents from the vector store.
- AI agent (OpenAI
gpt-4o-mini) generates natural language answers. - Forces metadata citation (URL, content type, publication & modification dates) in answers.
- Supports multilingual responses (answers in the visitor’s language).
- Memory & History Tracking
- Stores conversation history in Postgres for continuity.
- Logs embedding creation and workflow executions.
- Scalable & Automated Updates
- Scheduled checks for new/modified WordPress content.
- Automatically re-embeds and updates documents in Supabase.
