Getting Started with Wyregraf

Wyregraf adds an AI assistant to your WordPress site that answers questions from your own content. It works in three steps: you connect an AI provider, Wyregraf converts your posts and pages into embeddings (a searchable AI index stored right in your WordPress database), and a chat widget on any page answers visitors' questions from that index — with citations back to the source content. No external vector database or third-party search service is involved; everything except the AI provider calls stays on your server.

What you need #

• An OpenAI API key for creating embeddings (and optionally for the chat model), or an Anthropic API key for the chat model. Both are entered once, in a profile, and used everywhere.
• Content — the posts, pages, or other content you want the assistant to answer from.

The Wyregraf menu #

After activating the plugin, look for Wyregraf in your WordPress admin sidebar.

Start with Wyregraf → Setup. The guided setup checks each result rather than merely recording
that you clicked Next: it helps create and validate your provider profiles, index a source, create a
chat, try a question against your content, and copy the shortcode or enable the site-wide launcher.
You can leave and return without losing progress because completed steps are derived from the
profiles, index, and chat that actually exist.

1. The menu holds Setup, All Chats, All Prompts, All Embeddings, Namespaces, and Options — plus, with Wyregraf Pro active, All Agents and All Networks.
2. Add New Wyregraf Vector creates a chat — the assistant you will place on a page.
3. Each chat in the list is a separate assistant with its own content scope, prompts, and look.

The Options page #

Go to Wyregraf → Options. The page is split into tabs.

1. Getting Started links to the guides you are reading now. Embeddings & Vector Profiles and LLM Profiles are where you connect your AI provider — that is the first stop after activating. Help, Licenses, and System Info round out the page.

The pieces, in order #

The Setup wizard leads through these same pieces. Use the linked guides when you want the full
set of options or prefer to configure them manually:

1. Create an Embeddings & Vector profile — your embeddings provider key plus how content is chunked and indexed.
2. Create an LLM profile — the chat model that writes the answers.
3. Create embeddings from your posts and pages — or, with Pro, from PDFs, URLs, and code.
4. Create a chat and tune what it may answer.
5. Place the chat on a page with the shortcode.

Once that works end to end, Summaries & Questions can sharpen retrieval quality, the accuracy features add hybrid search, enrichment, and verified citations, and Wyregraf Pro's Agents & Networks let you build multi-step assistants.