Quick overview
Every plugin that promises “answers from your content” relies on the same trick: not retraining a large model on your site, but using retrieval-augmented generation (RAG). This guide explains RAG in plain English and then walks through setting up Hyve Lite, a free WordPress plugin that uses your posts and pages to answer visitor questions. Full disclosure: Hyve is our plugin, so you’ll get straightforward notes about what the free version can and can’t do.
What RAG actually is (simple three-step explanation)
– Turn the question into numbers: when a visitor asks something, the chatbot converts the query into an embedding — a numeric fingerprint representing its meaning.
– Search your content by meaning: your site content has been split into small chunks and pre-converted into embeddings. The system retrieves the chunks that are closest in meaning to the query.
– Generate the answer: the top-matching chunks plus the original question are sent to the AI model, which composes a reply grounded in your published text.
Two important consequences
– Good: answers update when your content updates. Change a post and re-sync the knowledge base; the bot’s replies will reflect the new text.
– Caution: the bot is only as useful as the content it can retrieve. Short or low-quality posts produce weak answers.
Set up a WordPress AI chatbot on your content (about 15 minutes)
Step 1 — Install Hyve Lite
– From the WordPress dashboard go to Plugins > Add New, search for “Hyve,” then install and activate Hyve Lite. Activation adds a Hyve menu in the sidebar where you’ll do the rest. No coding required.
Step 2 — Choose how to connect to the AI
– Use your own OpenAI key: in Hyve > Settings > Advanced paste your OpenAI API key and save. Messages go from your site directly to OpenAI and will consume your OpenAI quota.
– Use Hyve Connect (requires Pro): skip adding an API key. With a Pro license you connect Hyve to a managed backend and the plugin shows pooled usage limits for knowledge base chunks and chat messages.
Step 3 — Build the knowledge base
– Open Hyve > Knowledge Base and click Add Data. You’ll see a list of your posts and pages with an Add button next to each. Select the pages you want the assistant to use.
– Don’t add everything automatically. Pick the most useful, detailed, and relevant pieces: FAQs, user guides, policies, and core documentation work best. The quality and focus of the selected content determine the quality of answers.
Step 4 — Tune the assistant
– Go to Hyve > Settings > Assistant to pick the AI model and set sampling controls. Hyve offers models from higher-capability (and higher cost) options down to more economical ones; GPT-4o mini is a reasonable starting point if you want to balance cost and performance.
– Temperature controls randomness: lower values make replies more focused and deterministic (recommended for support-style, content-grounded bots); higher values add creativity but can stray from the source content.
Step 5 — Turn the chat on
– From Hyve > Dashboard enable chat on all pages, or place the Hyve Gutenberg block or shortcode on specific pages where you want the assistant to appear.
– Test on the site front end by asking a question that should be answered by one of your posts. If the bot quotes or paraphrases your content, the RAG pipeline is working.
Practical limits and trade-offs
– There’s no custom retraining: RAG is a runtime retrieval + generation pipeline, not a fine-tune of a model on your site.
– The free Hyve Lite knowledge base has a chunk cap (500 chunks). If you need larger knowledge bases, advanced reporting, or pooled managed access, that’s available in Pro.
– You still pay for the AI: either from your own OpenAI account or via Hyve Connect’s pooled plan if you buy Pro.
Bottom line
RAG is straightforward: split content into chunks, match queries to those chunks by meaning, and let a modern model generate an answer grounded in the retrieved text. Hyve Lite wraps that pipeline into an easy WordPress plugin so you can try a content-driven chatbot for little or no cost. For blogs, documentation sites, or small business sites with solid existing content, it’s a low-risk way to learn what visitors ask and to provide content-based assistance.
Try it and reflect
Grab Hyve Lite from the WordPress plugin directory and point it at your most helpful posts. Then watch what people ask — you’ll often discover questions and needs you hadn’t expected. Have you launched a chatbot on your content yet? What surprising visitor questions did you get?