If you have used AI Engine (Meow Apps) or AI Puffer / AI Power, Wyregraf will
feel familiar — it is a self-hosted, bring-your-own-key AI chat plugin — but a few core
ideas are different. This guide maps the concepts across so you know where to look.
What's the same #
- You bring your own API key. Like AI Engine, you paste an OpenAI or Anthropic key
and pay the provider directly. There is no per-message markup or hosted middleman. - A chat widget on your pages. You drop the assistant onto the site with a block, a
shortcode, a widget, or the floating launcher. - Per-role usage limits. Query, token, and cost quotas per WordPress role (and for
guests) work the way AI Engine's limits do — see Wyregraf → Options → Usage Limits.
What's different — and better #
Your content is a real, local vector index #
AI Engine and AI Puffer can answer from your content, but Wyregraf's whole design is
"chat with your site." It converts your posts, pages, PDFs, scraped URLs, and code into
embeddings stored in your own WordPress database — no external vector service
(Pinecone, Qdrant, a hosted "AI" endpoint) is involved. You control what is indexed, per
chat, through Namespaces and Embeddings profiles.
Concept map: what other tools call a "knowledge base," "data store," or "embeddings
index" is, in Wyregraf, an Embedding (a piece of indexed content) organized by
Namespace and searched by a Chat.
Answers cite their sources #
Every grounded answer links back to the exact source content it used. If you have relied
on AI Engine's chatbot and worried about made-up answers, Wyregraf's citations — plus the
optional grounded-only mode that refuses to answer when it has no matching content —
are the direct fix.
Retrieval is hybrid and tunable #
Wyregraf combines keyword and vector search and can generate summaries and questions per
chunk for better recall. The Knowledge Gaps screen shows you what visitors asked that
the index couldn't answer, so you know what content to add.
A suggested migration path #
- Install and connect a provider. Follow
Installation & Requirements then
Getting Started. Reuse the same OpenAI/Anthropic key you already
have. - Index your content. Create an Embeddings profile and index the posts/pages you
previously pointed your old tool at. See
Creating embeddings. - Create a chat and place it. Create a chat scoped to the
namespace you indexed, then display it with the block,
shortcode, or floating launcher. - Set your limits. Recreate your old per-role quotas under Options → Usage
Limits. - Retire the old plugin. Once the Wyregraf chat is answering well, deactivate the
previous plugin. Wyregraf's index and settings are independent, so there is nothing to
export or import between them.
There is no automated importer — the two tools store their data differently — but because
you reuse the same provider key and the same source content, re-indexing in Wyregraf is
quick.
