See llms.txt for all machine-readable content.

Back to Templates

Build a WordPress RAG chatbot with OpenAI, Qdrant or MongoDB

Created by

Created by: Paolo Ronco || paoloronco
Paolo Ronco

Last update

Last update 7 days ago

Categories

Share

Quick overview

Build a WordPress chatbot that answers questions from your site content using Retrieval-Augmented Generation. n8n handles indexing and chat orchestration, while Qdrant or MongoDB Atlas provides vector search and OpenAI generates context-grounded responses.

How it works

  1. Content indexing — WordPress content is retrieved, cleaned, chunked, embedded, and stored in the selected vector database.
  2. Intent routing — Incoming questions are classified so conversational requests, content searches, and optional profile/skills queries can follow different paths.
  3. Semantic retrieval — Relevant chunks are retrieved by meaning rather than exact keyword matching.
  4. Reranking — Cohere can rerank retrieved documents before response generation to improve relevance.
  5. Response generation — OpenAI generates an answer using the retrieved context and configured assistant behavior.
  6. WordPress delivery — The supplied WordPress integration exposes the chatbot through a responsive frontend and shortcode.
  7. Security and logging — The architecture includes webhook authentication, input handling, server-side secrets, and privacy-oriented logging controls.

Setup

  1. Purchase the complete package — Obtain the workflows, WordPress plugin, test frontend, supporting scripts, and complete documentation.
  2. Choose a vector backend — Configure either the Qdrant path or the MongoDB Atlas path according to the included deployment guide.
  3. Configure AI credentials — Connect OpenAI and any optional reranking services used by your selected setup.
  4. Import and configure workflows — Configure the indexing workflow first, then the real-time chatbot workflow and their required credentials.
  5. Index WordPress content — Run the indexing process and verify that embeddings and document metadata are available in the vector store.
  6. Deploy the frontend — Install/configure the WordPress integration, set the webhook and authentication values, and place the chatbot shortcode where required.
  7. Test retrieval and security — Validate normal conversation, RAG queries, unavailable answers, authentication, and frontend behavior before production deployment.
  8. Cohere API credentials for reranking (optional)
  9. MongoDB-based profile/skills knowledge path (optional)
  10. Standalone test micro-website (optional)
  11. Additional content sources or vector collections (optional)

Requirements

  • WordPress site with REST API access
  • n8n instance
  • OpenAI API credentials
  • Qdrant or MongoDB Atlas vector database
  • HTTPS for production deployments
  • Ability to install/configure the supplied WordPress integration

Customization

  • Knowledge sources — Extend indexing beyond WordPress posts or add separate collections for specialized data.
  • Retrieval — Adjust chunking, top-K retrieval, metadata filters, vector backend, and reranking behavior.
  • Assistant behavior — Modify intent classification, prompts, response style, language handling, and fallback behavior.
  • Frontend — Customize WordPress styling, labels, placement, responsive behavior, and shortcode integration.
  • Security — Adapt authentication, request validation, rate limiting, logging, and privacy controls to your deployment.
  • Models and providers — Change supported LLM, embedding, or reranking components where the workflow architecture allows it.

Additional info

The complete purchased package includes two n8n workflows, WordPress integration, test frontend, setup/customization documentation, and supporting tools.
The public repository preserves promotional images and frontend screenshots under assets/; the paid workflow package itself is distributed through the purchase channels.