User manual — Xabia Agent Core

Product version: Xabia Agent Core v1.0.209 (August 2026)
Index of manuals: https://xabia.ai/docs/
Online PDF: https://xabia.ai/docs/manual-usuario-xabia-core.pdf
Online HTML: https://xabia.ai/docs/manual-usuario-xabia-core.html

Quick Installation Guide

1. Download the Xabia Agent Core ZIP file ( xabia-agent-core-1.0.209.zip or equivalent retail package).
2. In WordPress, go to Plugins → Add New → Upload Plugin , select the ZIP file and click Install Now → Activate .
3. Open Xabia Agent and configure AI Connection : paste the XABIA--… , choose Xabia Cloud (recommended) or Own Infrastructure , and save.
4. Create an agent from New Agent , enter name, greeting and basic instructions.
5. Choose a data source (CSV, SQL, addon or multi-source), click Connect / Map when appropriate and save the agent.
6. Click Sync Data and, if using vector search, Train AI .
7. Publish the agent with the shortcode or activate the native interface from Appearance → Show on site without shortcode .
8. Test it in Playground and on a real page before opening it to the public.

Upgrading from a previous version: Upload the new ZIP file from Plugins , replace the installed version, and keep the existing settings. If you are using /xabia-box/ , go to Settings → Permalinks → Save after upgrade. After installing 1.0.72+ with WPML, visit a front-end page to synchronize the chat UI translations; clear the cache (LiteSpeed, etc.). With a remote SQL source, after upgrading to 1.0.164+ , re -sync (and submit to the Hub if applicable) to regenerate the tabs with the mapped attributes appended .

Quick guide: what to do in each situation

Use this table as a roadmap .

Each row indicates the exact screen and the order of steps.

1. Menu structure in WordPress

After activating Xabia Agent Core , the Xabia Agent menu appears on the desktop with a “superhero” icon.

Options depending on what your site has installed.

• If you have Xabia Central active, you can see Federation (Xabia Central) as a submenu to link multiple sites or nodes when using that feature.
• If you have the Nexus Federation addon, it may appear as Federation or Nexus Federation (see the manual for that addon if you use it).

To change almost all settings you need to be logged in as a WordPress administrator (or with an equivalent profile that allows you to manage plugins and site options).

2. Main screen: agent list

When you open Xabia Agent without editing a specific agent, you will see:

2.1 Title and shares

• New agent: Create a new profile and open the editor. When you save for the first time, the system assigns it a stable identifier (derived from the name); this is the code you will later paste into the web page with the shortcode (see below).

2.2 Existing Agent Cards

Each card shows:

• Name that the agent will see on the panel.
• ID: a code like ID: mi-asistente that must match what is in the shortcode: .
• Edit: opens all agent settings.
• Delete: Removes the agent and all the knowledge it has generated (what it learned during synchronization and training). This cannot be undone with a single click.

If you delete an agent, you will lose its configuration, data mappings, and associated memory. If you uploaded a CSV file for that agent, the file can remain on the server; if you no longer need it, you can delete it from Media or via FTP, depending on your situation.

2.3 “Available Addons” Block

Cards with the plugins (addons) that Xabia knows: name, brief description, and link to install or manage. Activating or deactivating a plugin is still done in WordPress Plugins , just like with any other plugin.

Important: an add-on may be installed but inactive, or active but require other configuration (add-on license, tables, etc.).

2.4 Pause an agent (list)

Each card in the list has three actions: Edit , Pause / Activate , and Delete .

• Pause: the agent stops being displayed on the site (the shortcode and the floating trigger do not appear).

  The configuration and memory are preserved.

• If it is paused, the card displays the label Paused and the button changes to Activate .

2.5 Agent Chat Interface (Appearance tab)

Since version 1.0.28 , the interface is configured per agent in Edit Agent → Settings / Appearance , under the “Chat Interface (Avatar and Panel)” section (after Chat Colors). Elementor and HTML blocks on the page are not required.

From v1.0.57 onwards there are clearly separate modes; from v1.0.192+ onwards there is also the embeddable launcher :

The “Show only on these pages” and “Exclude these pages or posts” rules belong to native mode. In shortcode/launcher mode, the chat or button appears only where the shortcode is pasted.

Official Kinetic Avatar (v1.0.47)

The default trigger reproduces the Xabia doll in inline SVG layers (head, white sockets and eyes/satellite), scaled to a 125×125 px box:

The avatar looks at the cursor with a depth effect: the head moves very little, the eye sockets a little more, and the eyes/satellite area moves more (GSAP animation). You can also use a custom image (PNG/SVG from Media).

Web behavior (without Elementor)

Options in the Appearance panel

Save with Save Agent . In native mode, the plugin automatically injects the trigger and assets into wp_footer . In shortcode mode, paste or where appropriate.

Chat appearance (v1.0.197–1.0.201)

The panel uses a stream layout (without message bubbles), square corners, and controls (send/microphone/mute) that are revealed when you hover over the writing area. User text can inherit the avatar’s accent color. Bot responses support basic Markdown ( **negrita** , bulleted lists) since version 1.0.201 .

Embedded chat with shortcode: halo and focus

The chat generated by the shortcode retains the aesthetics of the native panel: animated light border and soft shadow. By focusing on the text field, the plugin can raise the chat above the page and display a blurred overlay that obscures the background to improve readability.

The overlay closes when you click outside the chat, press Escape , or use the close button on mobile. This layer is independent of the native avatar backdrop: the shortcode remains an embedded chat, not a floating trigger.

The route /xabia-box/ (Smart QR / totem) displays the chat in full screen without a floating trigger.

Smart QR, code generator and totem mode are included in the Core — see the Smart QR manual for the complete guide with examples.

2.6 Online documentation (xabia.ai/docs)

The product manuals in PDF and HTML formats are published at:

In the development repository, the sources are located in xabia-agent-plugins/documentation/ and are regenerated with ./scripts/build-modular-manuals-pdf.sh .

3. AI Connection (Global Card)

Here you decide how your website connects to artificial intelligence and the xabia.ai service when using the recommended mode.

Use the Save Settings button on this same page (not to be confused with Save Agent within an agent).

3.1 Connection mode

You have two options:

After changing modes , check each agent: in the Cloud you don’t need to worry about the “engine”; in your own infrastructure you need to check which provider each one uses and if the paths/keys are correct.

3.2 License and balance

Information block (tokens and expiration)

• Remaining tokens: This is what’s displayed after the system checks your license. If it’s been a while since you last synced, a bar may appear until you tap Update Balance .
• Expiration: the date until which your license is active according to what the system receives.
• Checked: when the last check was done against the central service.

The Update Balance button prompts Xabia again and refreshes this information. If you haven’t saved a license yet, it will notify you beforehand.

Without a valid license or insufficient balance , the chat may not respond or may display a message about insufficient balance or license depending on the service settings.

“Xabia License for this site” field

• It is a password-type field;

  You can use the eye icon to see only what you are typing now .

• If you already have a saved license and leave this field blank when saving, the old license will not be deleted . To permanently remove the license, use the “Delete saved license” checkbox (see below).
• Only paste here when you want to change the license to a new string that has been sent to you.

License saved: you will only see asterisks and maybe the last few characters on the screen (for security).

Yellow alert: If the system detects that the saved file looks like a panel ID (for example, it starts with lic_… ) and not the long activation string that usually starts with xabia_ , it will warn you: the service probably won’t accept it. The usual solution is to delete the saved license, save it again, and paste the complete key from the email or your purchase receipt.

“View license saved in WordPress” button

It shows what’s actually stored on your WordPress site (without any strange spaces or truncation). Only trusted administrators should use it: anyone who sees the license could use it on another site if they copied it.

How is the connection protected?

When the license is correct, the plugin uses it to securely identify your site to xabia.ai when sending requests. You don’t need to understand the technical details: if you change the license, the new one takes effect upon saving. If you accidentally paste the asterisks from the screen instead of the actual key, the system will try not to overwrite the saved, valid license.

Checkbox “Delete saved license from this site”

• By selecting it and clicking Save settings , the license for this site will be deleted.
• Effect: The chat will stop using the Xabia service associated with that key until you paste another valid license.
• It does not automatically delete your OpenAI or Google keys from the advanced section; only the Xabia license.

3.3 API Keys and Google Cloud (own account)

This section only appears if you selected “Own Infrastructure “. It’s collapsed to avoid overwhelming you: open it when you’re ready to configure credentials.

OpenAI — key (global here)

• It’s the secret key to your OpenAI account .
• It is useful when the agent is configured to use OpenAI and when it does not carry a different key just for it.
• Treat it like a banking password: if someone copies it, they can make charges in your name. Change it from the OpenAI dashboard if you think it’s been leaked.

Google Cloud (Vertex AI) — path to the service account JSON file

• It must be the full path on the server to the JSON file that Google gives you for the service account with Vertex/Gemini permissions.
• If a specific agent does not fill in anything, this global route will be used.

Common errors: misspelled path, server cannot read file, account does not have Vertex permissions, or project/region does not match what you use in Google.

Google Cloud — Maps key

• For maps in the chat or visual elements that integrate Google Maps, if your project uses it.

Nexus Federation — “Activate Bridge Mode Only”

You will only see this if you have the Nexus Federation add-on installed.

Important: If you are using Xabia Cloud , checking or unchecking this box on this screen may not save as expected; the actual value will remain the one the site already had. To safely change Bridge mode, use your own infrastructure or the screen indicated in the Nexus manual.

What is Bridge mode (in a nutshell): Your WordPress site acts as a data server for other sites in the federation. It’s not the normal mode for a hotel or shop that just wants a chat feature on its page. If you accidentally activate it on a public site, the usual chat shortcode might not display correctly.

Consult with whoever the federation asked you to.

Tip: If you don’t know what the federation is for, don’t just select Bridge mode.

4. Agent Editor: Overview

When you click Edit or New Agent , the form uses tabs. The default tabs for Core are:

1. General — basic data, AI engine only if using your own infrastructure , sources (CSV/SQL…) and how to label columns for the bot.
2. Smart QR / Totems (Core) — landing page, tunnel URLs, QR generator per entity, kiosk mode. Detailed Smart QR manual .
3. Personality — colors, voice reading, how much context to use, response limits, and the assistant’s overall “script”.
4. Conversation log — recent history saved if enabled.

Addons can add extra tabs within the same agent (for example, Avirato ).

Press Save Agent at the end to apply changes; until then, only drafts are displayed.

5. “General” tab

5.1 Agent’s Name

• Name visible on the panel for you and your team.
• If you create the agent for the first time today, the shortcode identifier is generated from the name. If you only rename an existing agent, the shortcode identifier is usually retained (the old pasted URL remains unchanged).

5.2 Shortcode

After saving an existing agent, the following is displayed:

 By pasting it into a WordPress page 

Publish the chat of that agent (if the agent is not Paused ).

• If “Display on site without shortcode” is enabled, the shortcode can coexist with the native interface, but the floating avatar is controlled by the Appearance settings.
• If “Display on site without shortcode” is disabled, the shortcode will only show the chatbot embedded on that page, without a floating avatar or native visibility rules.

Some add-ons support extra options within the same shortcode (language, totem mode, assistant name displayed in messages, etc.): check these in the manual of each add-on you use. The lang attribute, when used, serves as the interface and fallback language; it does not lock the language in which the AI writes the response (§10.11).

5.3 Xabia Cloud Block (when the UI is in Cloud mode)

Informative message: the agent uses the Xabia service; the license and balance are managed on the plugin’s main screen , not here within the agent.

In this view you will not see duplicate engine or Google Cloud options: the system uses what it already has saved for that agent and what is consistent with Xabia Cloud .

5.4 Intelligence Engine (only appears in proprietary infrastructure)

Intelligence Engine:

Google JSON file per agent (optional): if empty, the global path to general settings is used.

OpenAI key for this agent only (optional):

• If you fill in and save, this agent uses that key and not the global one.
• If you leave it blank when editing and the agent already had a key, the previous one is usually preserved .
• Depending on its global mode and license, if there is no local key where needed, the system may route through the Xabia proxy ;

  It depends on the case.

Practical tip: If one agent uses your direct OpenAI account and another doesn’t, separate charges may appear for each provider — note which one uses which if more than one manages it.

6. Source of information (where the assistant learns from)

Here you decide what data the system should “read” before creating the memory used by the chat (FAQs, catalog, structured content, etc.). This is separate from the automatic context that some add-ons add to each message (for example, booking availability).

The plugin will remind you on screen that active add-ons can provide additional information without only going through this ; what you choose in this list is the knowledge base that you control with CSV, SQL or connectors.

6.1 CSV Files

• Useful for table catalogs: products, points of interest, listings exported from Excel as CSV.
• Uploading CSV saves the file to the site and links it to this agent.
• Delete CSV removes the linked file and the reference.
• Explore/scan CSV detects columns and prepares the mapping (what each column means to the bot).
• It generally distinguishes whether the columns are separated by a comma or a semicolon.

Remember: Until you click Sync Data — and, if you’re using semantic similarity search, Train — the assistant may not “see” those rows correctly in the responses.

6.2 WordPress Database (Local SQL)

A query to the WordPress database itself. Use the wildcard {prefix} when indicated by the theme: the system will substitute it with the actual prefix of your tables (usually wp_ or another prefix configured by your hosting provider).

Tip: Very broad queries can slow down the site or retrieve excessive data. Test first in a backup/development environment or with reasonable limits.

6.3 External database (remote SQL)

Fill in Host , Database Name , Username , Password and query when the catalog lives outside of WordPress.

If the external database is a WordPress site with Modern Events Calendar , you can use the remote MEC preset . The Core will populate the SQL query for events, apply the recommended mapping ( Evento , Fecha , Hora , Lugar , Link , Precio , Descripcion , Categorias_Tags , Imagen_URL ), and save sql_preset = mec_remote .

Use this preset when you are not going to install the MEC add-on on the chat site. If you do have Xabia MEC active, the recommended flow is Add-on Connectors → MEC with remote SQL credentials (§6.4), the same as Woo.

Security: credentials are stored with the agent; ideally, the remote server should only accept connections from your website’s IP address.

6.4 Add-on Connectors

It is only available with a compatible add-on installed and active . If it appears as unavailable, first install that plugin from Plugins .

After selecting it, use Connect and Map to review the columns it exposes before using them.

The MEC and Woo add-ons can also work with a remote database using the Host/DB/username/password fields. In that case, the mode remains native (because it provides rules, mapping, and domain actions), but the SQL read is performed against another WordPress site.

For remote MEC (Core ≥ 1.0.202 + MEC addon), if the chat site does not have a local Modern Events Calendar, the assistant never issues local booking buttons ( [ACTION:BOOK:ID] ): it directs the visitor with [ACTION:URL:Link] using the Link column of the remote event.

6.5 Multiple sources at once (“multi-source”)

You can combine up to two sources (e.g., remote SQL + CSV).

• Each part carries its own data and its own column mapping .
• In remote areas, an automatic prefix can often help, using the same criteria as {prefix} .

Tip: More sources offer flexibility, but also more places to make mistakes; review each block carefully.

7. Tools within the Data tab

7.1 Test the SQL query (one row)

It performs a test read ( only the first row ) of your query to show you which columns it detects and help with mapping. Avoid using it during peak traffic periods on uncontrolled production sites: it’s a testing tool.

7.2 WordPress Content Assistant (“Curating Console”)

A step-by-step assistant that:

1. List the content types of the active source (not those of the WordPress chat if the source is different).
2. Suggest which fields are interesting for the bot (title, text, metadata, taxonomies…).
3. You can transfer an SQL query and a reasonable mapping to the agent form .

Segregation by source (Core ≥ 1.0.162):

The modal displays a source warning (e.g., “Remote Database”). In deep schema, Woo products include _price / _sku / stock meta tags; MEC includes calculated slots where applicable.

What is “Entity”: mark the field that uniquely identifies each element (useful in QR experiences or when the visitor “chooses” a specific card).

It also has a shortcut to add metadata for a content type without going through the entire wizard.

7.3 Column mapping (repeatable attributes)

💡 Think of it simply: AI is incredibly intelligent, but it can’t guess which number in your Excel spreadsheet is a phone number, which is a price, or which cell is a link to an image. By assigning the Phone or Image role in the mapper, you’re giving it the keys to your business: you’re teaching it to populate its contact list for when a customer asks for a company’s phone number.

For each row, choose:

• CSV/SQL column or field .
• User-friendly labeling (“Price”, “Reservation phone number”…).
• Visual role in the chat when appropriate to show rich card: title, text, date, image (photos/gallery), logo (corporate brand), phone, website, email, map, etc.
• IDENTITY (ENTITY) if that field distinguishes unique records in your project.

Key roles for company catalogs (Core ≥ 1.0.118):

The Core does not assume fixed field names for a specific client. Everything is resolved through this mapping.

Some typical WordPress columns display a “brain” icon to highlight that they are good candidates for main knowledge text — you don’t have to touch it if you don’t need to.

8. Smart QR and totem mode (included in Core)

Since version 1.0.59 , all the power of Smart QR (entity tunnel, code generator, /xabia-box/ , kiosk mode) is part of Xabia Agent Core . There’s no need to install or license a separate plugin.

8.1 Three-step summary

1. Synchronize knowledge with at least one column marked as IDENTITY (ENTITY) (§7.3).
2. Publish a page with and select it as Landing Page in Edit Agent → Smart QR / Totems .
3. Click Generate QR next to the desired entity, download the PNG and print it.

The scan opens the chat with context from the entity (greeting and limited RAG). Example of generated URL:

 8.2 Totem mode (kiosk-type public screens) https://su-dominio.com/asistente/?ente_id=sala-impresionismo

In Smart QR / Totems you can activate Totem Mode and define inactivity minutes : after that time without using the chat, the session returns to the initial greeting (useful on a tablet at reception).

You can also use the full-screen path:

 In the shortcode you can force minutes with https://su-dominio.com/xabia-box/?x_project=ID_DEL_AGENTE

totem="10" :

 Summary: 

Smart QR = posters and links per entity; totem mode = kiosk on a fixed screen. You don’t have to use either if you only want classic chat on the web.

8.3 Extended documentation

Museum, hotel, retail cases, POI ?xqr= , migration of the old xabia-smart-qr plugin and troubleshooting: manual-usuario-xabia-smart-qr.md (PDF/HTML in xabia.ai/docs/manual-usuario-xabia-smart-qr.pdf ).

9. Sidebar (only saved agents, not “new”)

The right-hand column (memory + Playground) scrolls with the page (Core ≥ 1.0.165): it is not fixed at the top. This allows you to review the entire form without struggling with stuck panels.

9.1 Agent Memory (figures in the sidebar)

• Synchronized records: lines added after the last Synchronize data .
• Ready for smart search: how many pieces already have the “semantic” part prepared if you use that function.
• Tokens today: how many that agent has spent today compared to their daily limit (see below in Personality).

9.2 Button “1. Synchronize data”

Reread your CSV/SQL/multi-source or connector according to what you configured in General . Do this after changing source files or the query.

9.3 “2. Train” button (when vectorization is applied)

Prepare the text snippets for similarity search . If something goes wrong (Google permission, insufficient Cloud balance, etc.), the system will display a clear message to correct it.

9.4 “Delete vector memory”

Clears the mathematical memory associated with this agent (does not necessarily delete the CSV file on disk). Use it when you want to start from scratch after moving a lot of data.

9.5 Testing Zone (“Playground”)

A lab chat just like the one the visitor will use, but without publishing a page yet.

Ideal for fine-tuning before opening to the public.

10. “Personality” tab

Here you give the assistant a “face and voice.” Some add-ons may add more options at the bottom of this screen.

10.1 Assistant’s name (in messages)

The short name that visitors will see next to each bot message (default is “Xabia”). Do not confuse this with the kinetic avatar of the floating trigger (§2.5). You can use avatar_name="…" in the shortcode to customize it.

10.2 Colors and typography

• Identity color: buttons and highlighted chat details.
• Chat background color.
• Text size in pixels (choose what is readable on mobile and computer).

10.3 Voice (browser reader)

The visitor can use read aloud if their browser allows it.

• Voice type (default / more feminine / more masculine): This is a suggestion to the system; the result depends on the device.
• Speed: from slow to fast within the allowed range.
• Filters to make it sound natural: you can remove bold text, asterisks, technical marks like [ACCIÓN:…] or emojis, and add specific text that you don’t want to be read (for example, very long links).

In chat, the speaker button toggles between muted voice and read-aloud. When an action button is active or highlighted (voice, microphone, or send), the icon changes to white on a colored background to improve contrast.

10.4 Context Confidence Index

Value between 0 and 1 (on screen something close to 0.20 is usually suggested).

In simple terms: how closely does your website content need to match the question for the model to use it? If you set this value too high , the assistant will be more selective : it might become critical or generic because it discards pieces that don’t seem like a sufficient match. If you lower the value, it will accept more related context; sometimes, a bit more noise appears if the data isn’t well-organized.

Tip: Adjust slowly and test in the Playground before opening to the public.

10.5 How long can each answer be?

Control the maximum text size that each response can generate (the system has a reasonable range saved).

Increasing this limit: responses can be longer and more detailed , but they also use more tokens and may take longer. Decreasing the limit: responses are more efficient and direct, but may not delve as deeply into the topic.

10.6 Daily consumption limit per agent

Maximum number of tokens that this agent can use in a day (the counter closes according to the “technical” day of the service, usually at universal time).

If this limit is exceeded, the chat may enter maintenance mode for that agent even if the overall wallet has a balance: it’s a useful safeguard against spikes or accidental abuse.

10.7 How many pieces of knowledge to send in each question (“context”)

Number of fragments of your knowledge base that can be mixed with the question on each turn (usually in a wide range; if you leave it empty, the system will use a default value).

• More pieces: contextually richer , but more expensive and sometimes more “scattered” if the content is not right.
• Fewer chunks: more economical and direct, but the assistant may leave out data that was in other less prioritized chunks.

10.8 Intelligent (“vector”) search and sensitivity

• Enable vector search: This is only recommended if you have already performed a training exercise and see that there is available memory in the sidebar. If nothing has been trained, the help text will warn you.
• Similarity threshold (0–1): how similar a piece must be to the question to count.

  Very high values may cause you to “find nothing” on related but not identical topics.

10.9 Initial greeting

Simple HTML text that the visitor sees when opening the chat. WordPress will apply its usual security rules regarding allowed tags.

10.10 General instructions (personality and rules)

Here you write who the assistant is , the tone (formal, friendly, brief), and the limits (“does not give opinions on politics”, “do not invent prices without consulting data”, etc.).

Practical tip: You don’t need to list every synonym in your industry; with good data and a well-synchronized memory, the answers will be more natural. Use this chart for stable business performance .

10.11 Languages and multilingual responses

Xabia responds in the language of the user’s last message . If the synchronized knowledge is in Spanish but the visitor asks in English, French, or Basque, the model writes the response in that language.

The language configured by WordPress or by the shortcode’s lang attribute primarily affects the static interface (placeholders, buttons, browser voice). It does not force the AI to always respond in that language.

Avoid rigid phrases in the master prompt such as “respond exclusively in Spanish” unless you want to block multilingual behavior.

10.12 Multilingual site with WPML (greeting + chat texts)

Want your assistant to speak Basque or English automatically? With WPML + String Translation , Core ≥ 1.0.72 and a Cloud license, follow this quick 3-step path:

1. Write your initial greeting in Spanish within the Personality tab and save the agent .
2. Xabia will detect your active languages and (thanks to your Cloud license) will automatically translate the greeting in the background.
3. Go to WPML → String Translation and check that your translations are ready in the Xabia AI context. It’s that easy!

Next, clear the site cache (LiteSpeed, Cloudflare…) and try the URLs / , /eu/ , /en/ (or the ones you use) in incognito mode.

A must-have WPML setting (only the first time)

In WPML → Settings , scroll down to String translation → Change the original language of strings → choose Spanish .

If you don’t see that link: WPML → String Translation → Utilities and set Spanish as the original language for the plugin strings. This will prevent the greeting from appearing with the English flag even if the text is in Spanish.

Fixed chat texts (“Type here…”, buttons)

When you visit the site after installing or updating, Xabia usually populates the chat interface translations (placeholder, Send, Chat, etc.). If any text doesn’t change language: clear the cache, go back to the front end, and check the String Translation. If it’s still empty, contact support; sometimes a forced resynchronization is necessary.

If something doesn’t add up

No WPML or Cloud license? You can translate the greeting manually using String Translation; the automatic step simply won’t run.

11.

Catalog, remote passport and entity tracking (Core ≥ 1.0.118 / 1.0.164+)

Since version 1.0.118 , Xabia Agent Core has included a system-agnostic catalog engine : listings and record data by company, product, or point of interest are independent of field names specific to a particular client. Since version 1.0.164 , if your data resides in a remote database , Xabia will automatically save an appendix with each record containing the attributes you mapped in the General section (phone, email, website, etc.) when synchronizing.

11.1 What it solves

11.2 What you need to configure (only once)

1. Data source consistent with your catalog (CSV, local/remote SQL, multi-source, addon…).
2. Mapping of columns with at least one IDENTITY (ENTITY) field and contact/image roles (§7.3).
3. A recognizable activity category or taxonomy in your project — the Core deduces it from the mapping, without fixed names of a vertical.

Content within the same WordPress chat: listings and information can be read live (no sync required for that shortcut).

Data in remote SQL / Hub (Core ≥ 1.0.164): You must sync (and send to the Hub if using cloud search) so that phone, email, etc., travel with each record. Xabia doesn’t create fields: it only uses what you mapped in General and has a value in the row.

11.3 Typical Conversational Flow

1. Visitor: Listing question (multiple entities).
2. Assistant: concise vignettes; closes by inviting you to choose or ask for more details.
3. Visitor: “[name]’s phone number” / “first contact”.
4. Assistant: card data (annex or passport).
5. Visitor: “and some pictures.”
6. Assistant: Displays the image only if the Image role is mapped and has a value.

The system remembers the list based on the chat history sent in each message.

11.4 List vs. card mode (Core ≥ 1.0.165)

Agnostic behavior: it only depends on its mapping, not on a specific client.

11.5 If something goes wrong in the Playground

After a list, try in the same chat: “Phone number of the first one?” or “Any image?”. If it doesn’t respond well: Core ≥ 1.0.166 , Phone / Image roles in the mapping (§7.3) and, with remote SQL, Sync again.

Technical details for integrators: DEVELOPMENT.md §11–12 .

11.6 Hub vs local catalog vs remote SQL

A Hub without trained memory does not block the local listing;

Yes, it limits deep semantic responses until re-synchronization.

11.7 What’s New 1.0.162 – 1.0.168 (summary)

12. “Conversation Log” tab

When activated, the recent history appears here: date, visitor message, and assistant response.

Privacy (GDPR): Check with your legal representative whether you can store conversations that include personal data and whether you must inform in your privacy/cookies policy as appropriate in your country.

13. Prices, licenses and packages

📢 Note on fees and licenses: To ensure you always see current offers, the updated price list, token packs, and renewal conditions are managed in real time on our official website: https://xabia.ai/precio .

In practice you will find there:

• Core License (first year and renewal), with complimentary tokens according to the current offer.
• Token packs (Starter, Business, Enterprise or other published denominations).
• Addons ( Avirato , Woo , MEC , etc.) and custom federated network projects.

Top-ups are done in Xabia Agent → Wallet . Add-ons add features , but conversations with AI still use tokens unless an add-on indicates otherwise (“light” responses without a model).

14. Wallet

Menu: Xabia Agent → Wallet .

13.1 What does this screen show?

• Current balance of tokens associated with this installation when the service confirms it.
• A customer or license identifier that you’ll also see in some payment links. It’s not necessarily the same text as the secret string xabia_… that you paste into AI Connection . If you need to be sure, use “View saved license” on the main screen.
• Histogram of consumption over the last few weeks to get a visual idea of the spending rate.
• The “Renew year” message usually appears when the expiry date is approaching (typical window of about thirty days).

13.2 How to recharge tokens

1. Check that in AI Connection you have the correct long license (the one that was the email or the activatable key, usually in xabia_… format).
2. Log in to Wallet as administrator.
3. Click Top-up on the pack that suits your audience (see current prices and sizes at xabia.ai/price ).
4. Complete the payment on the payment gateway that opens. The system will then send the necessary data to associate the purchase with this site .

   You usually need to wait a few minutes and then tap Update Balance on the main screen until it matches.

Purchased from another computer: there is usually no problem as long as it is the same domain + license combination that the store expects for its site.

Multisite or a trial copy (staging): pay close attention to the URL from which you have installed Xabia: the purchase may be linked to that specific machine.

13.3 Renew only the license

The renewal button should continue the existing license , not create a new one. The current price is always displayed at xabia.ai/precio . If the date is still outdated after waiting a while and clicking ” Update balance ,” open a support ticket with your proof of payment.

13.4 Low Balance Notice

If your balance drops below tens of thousands of tokens, you may see a prominent message reminding you to check your wallet before the public experiences cutoffs due to insufficient credits.

15. Addons Screen

Paid add-on grid (Avirato, Central, MEC, Woo… depending on installation): status, links to plugins , and, for those using Hub subscriptions, a license form specific to the add-on (does not replace the Xabia license on the main site ). Smart QR/totems do not appear here: they are integrated into the Core (Agent tab).

• Shop: usual link to polar.sh/digixop .
• Customer portal (manage subscription): polar.sh/digixop/portal .
• If an addon license is invalid or the subscription is about to expire, the warning appears here ; in the public chat of addons like Avirato, the visitor receives generic contact messages, not technical license errors.

The installer’s ZIP file and the contract continue to arrive through the commercial channel or the usual mail from Xabia.

16. Practical order to avoid getting lost

1. Cloud vs. own infrastructure mode, license and balance (main screen).
2. Create agent — name and copy it into the shortcode where you want chat.
3. Choose a data source — map it carefully; test it in a staging environment if available.
4. Synchronize data → look at side statistics → if using smart search, train until counters paint consistent.
5. Refine Personality → try Playground → publish page with shortcode → monitor spending in Wallet.

17. Common problems (what to check before writing to support)

18. Frequently Asked Questions

Do I need to sync for it to list companies instantly?

Not for native listings and contact/image tracking (Core ≥ 1.0.118): Core queries live WordPress based on its mapping. Yes if you want a vector RAG for open-ended catalog questions.

How many agents can I create with one license?

In typical retail use, one or more agents on the same domain are included at no extra cost per agent. Always confirm the exact limits in your welcome email , on the official pricing page, or with your sales representative if it’s a large project.

What exactly is a “token”?

It’s the unit used to measure how much language processing a conversation entails ( incoming text plus outgoing text when using the model). Training the smart memory also consumes tokens. When the response comes from a fixed template (without a model), it often doesn’t use any.

Do I need to paste my OpenAI key if I use Xabia Cloud?

Generally not . With cloud services and a license with a balance, regular traffic goes through Xabia. The OpenAI key comes into play mainly in on-premises infrastructure .

Why doesn’t it “remember” my tables or PDFs if I have them on the server?

You must incorporate this knowledge using Sync (and Train if you enabled similarity search). Add-ons that provide live data (reservations, etc.) are a separate channel and do not replace this step if you want the assistant to cite your documentation.

Can I use staging copy with the same license as the production copy?

It depends on contractual conditions : two different public addresses sometimes count as two facilities .

Ask before moving actual traffic.

If I delete the license, will my agents be deleted?

They don’t disappear from the list nor is their configuration automatically deleted, but the chat will stop working properly in the Cloud until you paste another valid license .

Is there a balance in the wallet but the visitor sees “maintenance mode”?

Check the daily token limit for each agent under Personality . The Tokens Today sidebar shows whether you’ve already reached that limit for the current technical day; the next day, that counter resets to zero for the agent, even if your global wallet still has available balance for other agents or other days.

Release Notes (Core)

Core v1.0.208 (August 2026)

• Updates: Fixes the warning in Plugins when WordPress had not yet created the transient update_plugins .
• Updates: Hub revalidation every 5 minutes if the local cache matches the installed version.

Core v1.0.207 (August 2026)

• Automatic updates also in LITE mode (without an active PRO license).
• Version panel in the LITE admin with “Check now”.

Core v1.0.206 (August 2026)

• Checkout Polar: “Subscribe” links send domain and license_key to the Hub.
• Addons: badge «Hub Polar» below the panel title.

Core v1.0.205 (August 2026)

• Retail: If Core starts without a license (limited UI), it allows you to paste the Xabia license and activate PRO without WP-CLI.

Core v1.0.202 (August 2026)

• Chat actions: relative images in [ACTION:IMG:…] ; Amelia backup fallback; strict remote MEC rules ( [ACTION:URL:Link] only).
• Production-aligned monorepo: Full API and Woo/MEC/Avirato add-ons.

Core v1.0.201 (August 2026)

• Chat: Basic Markdown in replies; bubble-free UI stream; talking avatar/launcher; boot questions; full ZIP packaging.

Core v1.0.168–1.0.171

• Chat latency, Document-to-RAG, remote passport, listings, WPML/DTP (see technical history in DEVELOPMENT.md).