AxiaChat AI – Free AI Chatbot (Answers Customers Automatically)

Description

Transform your WordPress site into a 24/7 AI-powered customer service hub. AxiaChat AI delivers instant, intelligent responses using OpenAI (GPT-4, GPT-5), Anthropic Claude, and Google Gemini.

🚀 Key Features

Smart Training with RAG
Train your bot on YOUR content using Retrieval Augmented Generation. Unlike generic chatbots, AxiaChat AI learns from your website pages, PDFs, and documentation to provide accurate, contextual answers specific to your business.

Multiple AI Providers & Models
Choose between OpenAI, Claude, or Gemini. Create multiple specialized bots—one for sales, one for support, each with unique configurations and training.

MCP Server Integration
Connect Model Context Protocol servers to extend your bot beyond Q&A. Perform actions, fetch live data, and integrate with business systems.

Enterprise Features
✅ Context-aware responses (local or Pinecone embeddings)
✅ PDF ingestion for knowledge bases
✅ AI tools: web search, email notifications, custom actions
✅ Usage limits and cost controls
✅ GDPR-compliant consent management
✅ Full UI customization (colors, avatars, positioning)
✅ Floating widget or inline shortcode
✅ Complete conversation analytics

🎯 Real Benefits

• Reduce support workload by automating common questions
• Capture leads 24/7 even when your team is offline
• Improve engagement with instant, helpful responses
• Scale support without scaling headcount

Perfect for e-commerce, service businesses, SaaS, educational sites, and corporate websites.

Need help? Get free setup assistance: https://wpbotwriter.com/log-a-support-ticket/

Tutorial: Create Your First Bot with AxiaChat AI

Learn how to install and set up your first AI chatbot in just 5 minutes

In this quick video, we cover:
– Create & customize your bot (text + voice)
– Train it with your site pages, PDFs, and FAQs
– Full appearance control + moderation

Features

🤖 AI & Model Support

• Multiple AI Providers: OpenAI (GPT-4, GPT-5), Anthropic Claude, Google Gemini
• MCP Server Integration: Connect Model Context Protocol servers for extended capabilities
• Provider-native Web Search: Enable live internet lookups with cited sources (OpenAI, Claude, Gemini)
• Multiple independent bots with unique configurations per bot (model, temperature, instructions, UI)
• Advanced AI Tools: per-bot capabilities including web search, email notifications, custom actions with admin logs

🎓 Smart Training & Context

• Retrieval Augmented Generation (RAG): Train your bot on your actual content
• Context modes: embeddings (local or Pinecone) / page-specific / none
• PDF ingestion: Upload and process PDF documents into your knowledge base
• Auto-sync: Automatically detect and update changed content
• Smart chunking: Overlapping content chunks for better retrieval precision
• Easy Config Wizard: Guided setup scans your site and creates optimized context automatically

🎨 Customization & Deployment

• Floating global widget OR inline embedding via shortcode
• Full UI control: custom colors, avatars, positioning, window controls
• Draggable, minimizable, closable chat panel with state persistence
• Multiple deployment options: per-page shortcodes with attribute overrides
• 9 avatar designs included + custom avatar upload
• Responsive design: works perfectly on desktop and mobile

📊 Usage Control & Analytics

• Conversation logging with ON/OFF toggle (user/session tracking; IP only if limits enabled)
• Daily usage limits: per-user/IP and global caps
• Flexible limit behaviors: disable input or hide widget when exceeded
• Cost tracking: monitor token usage and API costs
• Detailed conversation logs: review all interactions with filtering and search

🔒 Security & Compliance

• GDPR-compliant: built-in consent management with in-stream consent bubble
• Privacy controls: optional conversation logging, IP anonymization
• WordPress security best practices: nonces, capability checks, prepared queries, output escaping
• Encrypted API keys: secure storage of sensitive credentials
• Data ownership: all processing on your server, no external SaaS proxy
• Local vendor assets: Bootstrap & Icons bundled—no CDN dependencies

🛠️ Developer-Friendly

• Action hooks: aichat_after_response, aichat_conversation_saved for custom integrations
• Extensible architecture: provider system for adding new AI services
• Translation ready: text domain axiachat-ai (Spanish included)
• Clean uninstall: removes options safely (conversation data preserved)
• Well-documented code: follows WordPress coding standards

🌐 Language & Accessibility

• Works with any language (customizable responses)
• RTL support for right-to-left languages
• Spanish translation included (es_ES)
• Translation ready with .pot file for easy localization

Need a custom capability? Tell us what your bot should do! We love building useful integrations for the community and will gladly add features that help others. Request a capability →

Usage

Basic Usage

After installation, your chatbot is ready to start helping your visitors! Here’s how to make the most of AxiaChat AI:

For Site Owners:
– Create specialized bots for different purposes (sales, support, FAQs)
– Train bots with relevant content using the Context manager
– Monitor conversations and improve responses based on real user interactions
– Set usage limits to control API costs
– Enable/disable logging as needed for privacy compliance

For Visitors:
– Click the floating chat icon to start a conversation
– Ask questions in natural language—the AI understands context
– Receive instant, accurate answers based on your site’s content
– Get cited sources when web search is enabled
– Accept GDPR consent to start chatting (if enabled)

Advanced Configuration

AI Tools & Capabilities:
Enable advanced features per bot in the Tools section:
– Web Search: Let the bot search the internet for current information
– Email Notifications: Send notifications to admins or authorized recipients
– Custom Actions: Use MCP servers to extend functionality
– Domain Restrictions: Limit web search to trusted domains only

Window Controls:
Customize the chat experience:
– Closable: Users can dismiss the chat window
– Minimizable: Compact to icon when not in use
– Draggable: Let users move the window around the screen
– Start Minimized: Begin in compact mode

Usage Limits:
Protect your API budget:
– Set per-user or per-IP daily message limits
– Configure global daily limits across all users
– Choose behavior when limits are reached (disable input or hide widget)

Context Modes:
Choose how the bot accesses information:
– Embeddings: Uses RAG with your trained content (most powerful)
– Page: Answers using only the current page content
– None: Pure AI responses without site-specific context

Need help with setup? Our team provides free configuration assistance to ensure your bot works perfectly for your use case: https://wpbotwriter.com/log-a-support-ticket/

AI Tools (Capabilities, Macros & Rules)

Supercharge Your Chatbot with AI Tools

AxiaChat AI goes beyond simple Q&A. The optional AI Tools layer allows your assistant to take actions, access live data, and integrate with external services—all while maintaining strict security controls.

🌐 Provider-Native Web Search

Enable real-time internet lookups so your bot can answer questions about current events, latest information, or topics beyond your site content.

Supported Providers:
– OpenAI: GPT-5* models with native web_search capability
– Claude: 4.x models with web_search_20250305 tool
– Gemini: google_search integration (Live API only)

Features:
– Cited sources in responses
– Optional domain allowlist per bot (restrict to trusted sites)
– Automatic integration—no complex configuration needed

When to Use:
– News sites needing current information
– Product comparison queries
– Research and fact-checking
– Supplement your knowledge base with web data

📧 Email Notifications

Send intelligent email notifications triggered by conversations.

Safety Features:
– Sends to site admin by default
– Rate limiting to prevent spam
– Policy gates for security
– Client emails require explicit server-side authorization via filter

Use Cases:
– Notify team when high-priority leads are captured
– Send follow-up emails based on conversation topics
– Alert staff to urgent support requests

🔌 MCP Server Integration

Model Context Protocol (MCP) integration enables your chatbot to connect with external tools and data sources.

What You Can Do:
– Fetch live data from APIs
– Query databases
– Integrate with business systems
– Create custom workflows
– Extend capabilities without coding

How It Works:
1. Add MCP server configuration in the addon settings
2. Enable desired tools per bot
3. The AI automatically calls tools when appropriate
4. Results are seamlessly integrated into responses

📋 Rules Engine (Optional)

Define simple conditions to trigger automatic actions:
– Ask follow-up questions based on user responses
– Speak a message at specific points
– Call tools automatically when certain conditions are met

📊 Tools Logging

Every tool call is logged for transparency and debugging:
– Execution duration
– Output excerpts
– Error tracking
– Searchable admin interface

Access logs: AxiaChat AI > Logs > Tools

🔒 Security by Design

• Sandboxed execution: Tools run in isolated contexts
• Policy gates: System-level restrictions prevent abuse
• No secrets exposed: API keys and sensitive data never sent to AI models
• Admin authorization: All tool capabilities require explicit enablement

Note: On OpenAI GPT-5* (Responses API), AxiaChat normalizes function tools to the required schema and passes the native web_search tool when enabled.

Easy Config Wizard

Version 1.1.2 introduces an optional “Easy Config” wizard that appears after initial activation (or until completed). It performs:
1. Site scan: collects up to the most recent 200 published posts, pages (and products if WooCommerce detected).
2. Context creation: creates a local embeddings context record.
3. Batch indexing: chunks are created by generating embeddings for discovered items (10 per batch via AJAX) using your OpenAI API key.
4. Bot linking: links the default bot (slug “default”) to the new context with context mode = embeddings.

Smart Mode (experimental):
If available the wizard now uses a smart discovery mode prioritizing:
* Homepage and its internal links
* Legal / FAQ / About / Policy pages (slug heuristics)
* WooCommerce top categories and a sample of recent products (if WooCommerce active)
* Fallback to a few recent posts/pages if the set is too small
Legacy behavior is still used internally as fallback.

Real Chunking:
From 1.1.2+ the indexing process splits long content into multiple overlapping chunks (~1000 words, ~180 overlap) each embedded separately, improving retrieval precision. Existing single-row entries will be transparently replaced on re‑index.

Selective Indexing:
The wizard lets you deselect discovered pages/products before indexing to reduce token usage and noise. At least one item must remain selected.

Notes:
* The wizard hides automatically after completion (flag stored in option aichat_easy_config_completed).
* You can abort and later ingest additional or different content via the Context screens.
* If you already manually configured a bot/context you can ignore the wizard and mark it complete by finishing or deleting the option.
* Current limit (for safety) is 200 items; extend manually if needed.

Troubleshooting:
* If embeddings fail, check that your OpenAI API key is valid and that the server can reach api.openai.com.
* Browser console (when AICHAT_DEBUG true) will include extra debug logs.

Shortcode Reference

Basic:
[aichat id=”bot-slug”]

Aliases: also bot="bot-slug".

Optional overrides (if supported):
* title=”Custom Title”
* placeholder=”Ask me anything…”
* layout=”floating|inline”
* position=”br|bl|tr|tl|bottom-right|bottom-left|top-right|top-left”
* class=”extra-css-class”

Example:
[aichat id=”support-bot” title=”Support Assistant” layout=”floating” position=”bottom-left”]

Data Storage

Custom tables:
* wp_aichat_conversations – message, response, timestamps, page_id, session_id, bot slug, optional user_id, IP (binary) if limits active
* wp_aichat_contexts – context definitions
* wp_aichat_chunks – content chunks & embeddings
* wp_aichat_bots – bot configuration

Options store API key, limits, GDPR, widget settings. Tables are not auto‑dropped on uninstall (safety).

Privacy

User prompts (and selected context snippets) are sent to OpenAI. Content may contain personal data if users type it. Inform users and obtain consent where required. Logging can be disabled; if enabled, data stays on your server.

GDPR / Compliance Notes

• Data Processor: OpenAI
• Lawful basis: legitimate interest or consent (consult legal counsel)
• IP Storage: Binary IP kept only when enforcing per-user/IP limits
• Right to erasure: Delete rows manually (future export/anonymize tool planned)
• Consent: Optional in‑stream consent bubble blocks input until accepted
• Recommendation: Add a privacy note near the chat input

Security

• Nonces on AJAX (aichat_ajax, delete actions, etc.)
• Capability checks (manage_options) for admin screens
• Prepared statements / sanitization for user input
• Escaped output in admin & front‑end templates
• API key stored in an option (not exposed publicly)

Performance

• Assets only enqueued when needed; versioned with filemtime
• Embedding/PDF ingestion can be heavy—schedule during low traffic
• Lightweight front‑end footprint otherwise

Bundled Libraries

• Bootstrap (local)
• Bootstrap Icons (local)
• smalot/pdfparser (LGPLv3) for PDF extraction

Translation

• Text Domain: axiachat-ai
• Domain Path: /languages

External Services / Data Disclosure

This plugin can connect to the following third‑party APIs depending on which features you enable. You (the site owner) must supply the API keys. No keys are bundled and no traffic is proxied through a vendor server controlled by this plugin author.

1. OpenAI

Used for: chat completions / responses, embeddings (context indexing), moderation (safety checks).

Endpoints used (HTTPS):
* https://api.openai.com/v1/chat/completions (legacy Chat Completions)
* https://api.openai.com/v1/responses (Responses API – new unified endpoint if configured)
* https://api.openai.com/v1/embeddings (document/post/PDF embedding + wizard indexing)
* https://api.openai.com/v1/moderations (content moderation)

Data Sent:
* User prompt text (per message) and limited conversation history (trimmed for token control)
* System / policy instructions (security + privacy policy + bot instructions)
* Optional retrieved context snippets (only the selected top‑N chunks or page excerpt – never the full original document)
* Embedding requests: raw chunk text produced from your site’s content or uploaded PDFs
* Moderation: only the user prompt text (not the entire history)

Data Retention (Your Server):
* Conversation log rows (if logging enabled) including user prompt, model reply, timestamps, bot slug, session id, optional user id. IP (binary) only stored when per‑IP limits are turned on. Disable logging to stop storing new rows.
* Embeddings table stores numeric vectors generated from your content (not reversible plaintext) plus the original chunk text for retrieval.

Recommendations:
* Update your privacy policy to disclose sending user prompts and limited site content to OpenAI for processing.
* Disable logging or periodically purge if you process personal data.

Legal / Docs:
* Terms: https://openai.com/policies/terms-of-use
* Privacy: https://openai.com/policies/privacy-policy
* Usage Policies: https://openai.com/policies/usage-policies

Opt‑Out / Control:
* Remove the OpenAI API key in Settings to stop all OpenAI calls (bots will refuse to answer).
* Disable conversation logging.
* Limit context ingestion to non‑sensitive pages.

2. Anthropic (Claude)

Used for: alternative chat completions via Claude models (messages API) when a bot provider is set to Anthropic/Claude.

Endpoint used (HTTPS):
* https://api.anthropic.com/v1/messages

Data Sent:
* A rewritten message array: system instructions + user prompt + condensed prior turns (trimmed) + optional retrieved context snippets.
* Model identifier, max tokens / temperature style parameters.

Headers:
* x-api-key (your key) and anthropic-version (currently 2023-06-01 set in code).

Retention (Your Server):
* Same as OpenAI notes for conversation logging (the provider choice does not change local storage schema).

Legal / Docs:
* Terms: https://www.anthropic.com/legal/terms-of-service
* Privacy: https://www.anthropic.com/legal/privacy
* Usage Policy: https://www.anthropic.com/legal/aup

Opt‑Out / Control:
* Leave the Claude API key blank; those bots will fallback/refuse if provider requires it.
* Switch provider per bot back to OpenAI.

3. Pinecone (Optional Remote Vector Store)

Only used if you explicitly create a Context with remote type “Pinecone”. Local context mode (default) never contacts Pinecone.

Endpoint Pattern:
* Region/index specific HTTPS endpoints you enter (example placeholder: https://controller.pinecone.io and index query/upsert endpoints under *.pinecone.io). The plugin validates host against an allowlist containing pinecone.io (filter extendable).

Used For:
* Upserting embeddings (during context indexing / syncing)
* Querying similar vectors when answering a question in that context.

Data Sent:
* Vectors (embedding arrays) and …