AxiaChat AI for WordPress
Build multiple AI chatbots with independent settings, floating widget or shortcode mode, and enrich answers using your own content (PDFs, posts, custom chunks) via a lightweight RAG pipeline (local DB or optional Pinecone).
Shortcode Example
[aichat id=”sales” placeholder=”Ask about plans…”]
Customize position, color, avatar, placeholder, send button text and more.
Main Features
Multiple Bots
Each with its own model (OpenAI or Claude), temperature, system instructions and UI style.
Widget + Shortcode
Global floating chat or inline embeds where you need them.
Context (RAG)
Ingest PDFs or existing site content, chunk & embed for better answers (local or Pinecone).
Local Embeddings
Stored in your database—full ownership & portability. Optional remote vector store via Pinecone.
Consent Layer
Optional consent gate before any message is sent.
Usage Limits
Per user/IP & daily caps to prevent abuse.
Optional Logs
Conversation storage toggle for internal review (disable anytime).
Customizable UI
Colors, placeholders, open sentence, avatars, draggable/minimizable.
Security
Centralized sanitization, nonces, capability checks, size/MIME validation & no external proxy layer.
Claude Support
Anthropic Claude models selectable per bot (instructions + style preserved).
Smart Link Hints
Responses can embed a contextual top link via placeholder strategy.
Installation & Getting Started
- Install and activate the “AxiaChat AI” plugin.
- Go to Settings → AI Chat and add your OpenAI API Key.
- Create a bot (model, instructions, UI appearance).
- (Optional) Ingest context (PDFs or posts) to enhance responses.
- Insert the shortcode or enable the global floating widget.
Shortcodes & Attributes
[aichat id=”sales” title=”Sales Assistant” placeholder=”Ask about pricing…”]
[aichat id=”support” layout=”floating” position=”bottom-left”]
| Attribute | Values | Description |
|---|---|---|
| id / bot | slug | Identifier of the bot to load. |
| title | Text | Header title (inline mode). |
| placeholder | Text | Input placeholder text. |
| layout | floating | inline | Render as floating widget or inline block. |
| position | br | bl | tr | tl | Floating widget corner. |
| class | CSS class | Extra CSS class for styling. |
| button | Text | Custom send button label. |
| avatar | URL | Override bot avatar (if avatars enabled). |
Privacy & GDPR
Requests are sent directly to OpenAI or Anthropic (Claude) using your own API keys. You can disable conversation logging entirely. If IP-based limits are enabled, IPs are stored in binary form (not plain text). Add an appropriate privacy notice for your jurisdiction.
- Optional in-chat consent barrier.
- Data resides on your server (except model API calls).
- Embeddings & chunks stored locally.
FAQ
Do I need my own API keys?
Yes. Bring your own OpenAI and (optionally) Anthropic keys. The plugin does not proxy or bundle access.
Can I disable chat logging?
Yes. When disabled, no new rows are stored; existing history remains until you purge it.
How does the context (RAG) work?
Indexed content is chunked, embedded and similarity‑matched; the most relevant snippets are appended to the model prompt.
Performance impact?
Indexing PDFs can be heavier; schedule during low traffic times. Normal runtime queries are lightweight.
Does it support Pinecone?
Yes. You can switch a context to use Pinecone for vector search; local mode stays available for full control.
How are inputs sanitized?
A centralized set of helpers enforces bounded lengths, allowed MIME types, safe integers and size caps for patch/JSON bodies.
Start Now
Install the plugin, create your first bot and deliver contextual AI assistance to your users. Free, extensible, and privacy‑aware.