ChatGPT - Comprehensive AI shop assistant

Installation instructions, documentation and FAQs for the Shopware plugin

Installation instructions

Install the extension via the extension area.

OpenAI API key

  1. Go to https://platform.openai.com/.
  2. Log in.
  3. Create a new project.
  4. Go to your profile and go to the "API keys" section https://platform.openai.com/settings/organization/api-keys.
  5. Create a new API key via "Create new secret key".
  6. Top up credit so that the OpenAI API works. https://platform.openai.com/settings/organization/billing/overview
  7. Optional: We recommend that a usage limit is set so that no unnecessary costs are incurred in the event of an error https://platform.openai.com/settings/organization/limits
Create a new API key

Setting up the extension in Shopware

API credentials

  • Enter the API key, this should start with "sk-proj".
  • Select your desired AI model.

Frontend Assistant

A selection is only available if a front-end wizard has been created.

  • To create a frontend assistant, go to Catalogues -> OpenAI Assistant Frontend.
  • Select "Add new assistant".
  • The Assistant ID is created automatically.
  • Select your desired sales channel.
  • Name your assistant so that it is easily recognisable.
  • Choose whether you want a formal or less formal tone.
  • Store additional information for the assistant, such as separate instructions or an FAQ.
  • Select the AI model here.
  • The temperature indicates how random the AI's answers are. Further information here: https://platform.openai.com/docs/api-reference/assistants/modifyAssistant
  • Top P is an alternative sampling method similar to temperature. Further information here: https://platform.openai.com/docs/api-reference/assistants/modifyAssistant#assistants-modifyassistant-top_p
  • The design colour of the frontend assistant is based on the "primary colour" in the theme.
  • The frontend wizard texts can be customised using the text modules:
    • The text modules can be found under Settings, in Shopware 6.6 Settings->Shop.
    • The search window in the text module set can be used to search for "biloba.openAI".
    • The following values can then be adjusted here:
Frontend assistant text modules
Creating a new frontend wizard
Changed colour design of the frontend wizard (primary colour)
Changed colour design of the frontend assistant

Settings "All META tags product / category (configured)"

Settings for the creation of SEO data in mass processing. These settings are only for mass processing or for the "All META tags (configured)" button in single mode.

Creating META information for images

  • Automatische Generierung ausgewählter META-Informationen beim Erstellen oder Aktualisieren von Bildern: diese Einstellung ermöglicht es, dass die META-Informationen für Bilder direkt beim hochladen oder beim aktualisieren eingefügt werden. Dies funktioniert automatisch, bitte prüfe deine OpenAI Limits.
    • This option works NOT with the product context. In order for meta data to be generated with the product context, the article must first be saved with the desired image. This is necessary for the assignment.
  • Selection of languages to be generated (default language is always generated): a list of the languages in which the META information is generated.
  • Checkboxes which information is processed in mass processing and automated creation.
  • Wenn du einen Htaccess für deinen Shop aktiv hast, musst du diese Anmeldedaten hier eintragen, damit die Bild META-Informationen erstellt werden können. Dies ist nötig, da die Bilder von OpenAI analysiert werden und ein Zugriff benötigt wird.

General settings

AI processing single entity

Product description

  • Go to Catalogues -> Products and select your desired product.
  • If there is a product description, this and the product title are used.
  • If no description is available, please enter a text of 50-1000 characters. This will be used to create the description.

Products SEO data

  • Go to Catalogues -> Products and select your desired product.
  • Meta title, meta description and keywords can be created in the SEO area.
  • Use the corresponding button to create your desired SEO data or all three.

Products Images Meta Data

  • Go to Catalogues->Products and select your desired product.
  • Go to the pictures section.
  • Meta data can be generated here for images that have already been saved.
  • Please note that meta data is only generated if the article has been saved with the desired images.

Category Description

  • Go to Catalogues -> Categories and select your desired category.
  • If there is a product description, this and the product title are used.
  • If no description is available, please enter a text of 50-1000 characters. This will be used to create the description.

Category SEO data

  • Go to Catalogues -> Categories and select your desired category.
  • Meta title, meta description and keywords can be created in the SEO area.
  • Use the corresponding button to create your desired SEO data or all three.

Bulk text editing

  • Go to Catalogues -> Open AI - Mass update.
  • The desired area to be edited can be selected in "Area".
    • Product description
    • Product META information
    • Image alt and title text
  • Individual settings are available for the various areas
  • Filter specifies whether something should be taken into account when selecting the areas to be processed, such as only processing entities that have not yet been processed by the AI
  • Languages specifies which language is to be edited.
  • Categories offers a category selection. This is used to narrow down the entities to be edited. Select all categories with Shift+click to edit all entities.
  • The entities found, which are processed, are listed below.
  • In the "Currently running bulk jobs" area, it is displayed whether a job is currently running
Selection of available areas

Prompting

Overview

The Prompting module allows you to customise and modify AI text generation. It is possible to extend or replace the default prompts. A basic understanding of prompting is required, so this module is only recommended for „experts“.

Access to the prompting module

The prompting module can be accessed in two ways:

  1. About the settings: Navigiere zu Einstellungen → Plugins → Open AI Prompts.
  2. About the mass update: Im Massenupdate-Modul findest du einen Link „Prompts konfigurieren“ direkt in der Konfigurationsansicht.

Available areas

The following areas are available to you in this app:

  • Product description: Generation of product texts
  • Product meta title: SEO-optimised product titles
  • Product meta description: SEO descriptions for products
  • Product meta keywords: Relevant keywords for products
  • Category description: Category texts
  • Category metadata: SEO data for categories
  • Media metadata: Alt texts and titles for images, product context can be set here. Please note.

Configuration options per area

1. Additional prompt (optional)

  • Special instructions can be added here, which are sent to the AI in addition to the standard system prompt.

2. Available variables

  • The available variables are displayed for each area.
  • Variables can be clicked directly to insert them into the additional prompt.
  • The variables are automatically inserted at the cursor position in the text.
  • Available variables depending on the area:
  • ##targetLanguage## – Target language
  • ##productName## – Product name
  • ##productText## – Product description
  • ##productNumber## – Item number/MPN
  • ##productManufacturerName## – Manufacturer name
  • ##productManufacturerDescription## – Manufacturer's description
  • ##categoryName## – Category name
  • ##categoryText## – Category description
  • ##lengthFrom## / ##lengthTo## – Text length specifications
  • ##keywords_amount## – Number of keywords

3. AI model selection

  • The OpenAI model can be selected for text generation.
  • Different models have different strengths and costs:
  • GPT-5 Mini: Fast and cost-effective, ideal for large volumes
  • GPT-5: Balanced performance and quality
  • GPT-5 Nano: Compact model for simple tasks
  • GPT-4.1 series: Older models, still available

4. System prompt settings

  • The standard system prompt contains the basic instructions for the AI.
  • Option „Standard-System-Prompt deaktivieren“: kann verwendet werden, wenn die komplette Kontrolle übernehmen werden soll. Bitte nur aktivieren, wenn dumit Prompting vertraut bist.
  • The current system prompt is displayed for transparency.

Important information for prompting

Caution when adjusting prompts:

  • Prompts sollten nur angepasst werden, wenn du dich mit Prompting auskennst.
  • All adjustments are made at your own risk.
  • Teste Änderungen zuerst im Einzelmodus oder bei der Massenbearbeitung mit wenigen Elementen.
  • The prompt that was actually sent can be checked in the logging module in the detailed view.

Best practices for your own prompts

  1. Sei spezifisch: The more precise your instructions, the better the result.
  2. Nutze Variablen: Verwende die verfügbaren Variablen für dynamische Inhalte.
  3. Teste schrittweise: Beginne mit kleinen Änderungen und teste die Ergebnisse.
  4. Examples are helpful: Gebe der KI Beispiele für den gewünschten Stil.
  5. Note the length specifications: Nutze die Längenvariablen für präzise Textlängen.

Examples of additional prompts

For product descriptions:

Write in an enthusiastic tone and use bullet points for the most important features. Start with a compelling introductory sentence that addresses the main problem that ##productName## solves.

For meta titles:

Always include the manufacturer name ##productManufacturerName## at the beginning of the title. Use action-oriented words such as "Buy", "Order" or "Discover".

For category descriptions:

Structure the description in 3 paragraphs: 1. Introduction to the category 2. Main advantages of the products in this category 3. Call to action to browse

Logging and control

Nach der Textgenerierung kann im Logging-Modul überprüft werden:

  • Which prompt was actually sent to the AI.
  • How the variables were replaced.
  • What result was generated.

This helps you to continuously optimise your prompts.

Tips for troubleshooting

  • Texts that are too short/long: Passe die Längenangaben in der Plugin-Konfiguration an.
  • Wrong tone: Ergänze spezifische Ton-Anweisungen im zusätzlichen Prompt.
  • Missing information: Stelle sicher, dass alle benötigten Produktdaten gepflegt sind.
  • Inconsistent results: Verwende spezifischere Anweisungen oder wechsel das Modell.

OpenAI - Log

The OpenAI log records every job that has run for the corresponding areas. If a job does not run, the log is the first point of contact.

Go to Catalogues -> OpenAI - Log

Entity Log

The individual entities that are processed by the AI are entered here and sorted by area.

Bulk Log

The bulk log jobs are entered here.

Assistant Frontend Log

The customer messages are entered here with the front-end assistant and you can see how the assistant reacts to them in the detailed view.

Detailed view of messages between customer and front-end assistant

Common errors - entries in the log:

  • error_no_credits: indicates that no credit is stored in the OpenAI account.

MCP-Server Verwaltung

Was sind MCP-Server?

MCP-Server (Model Context Protocol) sind externe Dienste, die der KI zusätzliche Fähigkeiten verleihen. Während der Textgenerierung kann die KI diese Server kontaktieren, um Echtzeitdaten abzurufen – zum Beispiel aktuelle SEO-Keywords, Suchvolumen oder Wettbewerbsanalysen.

Server hinzufügen

Um einen neuen MCP-Server hinzuzufügen:

  1. Klicke auf „Server hinzufügen“
  2. Gib einen Namen für den Server ein (z.B. „SEO Keywords“)
  3. Wähle den Server-Typ aus der Liste
  4. Trage die erforderlichen Zugangsdaten ein (API-Key, URL, etc.)
  5. Speichere die Konfiguration

Server testen

Nach dem Speichern kannst du die Verbindung testen:

  1. Klicke auf das Test-Symbol neben dem Server
  2. Der Server wird kontaktiert und die verfügbaren Tools werden abgerufen
  3. Bei erfolgreicher Verbindung siehst du eine Bestätigung

Server mit Prompts verknüpfen

Damit ein MCP-Server bei der Textgenerierung verwendet wird:

  1. Gehe zu den Prompt-Einstellungen
  2. Wähle den gewünschten Prompt-Typ (z.B. „Produkt Beschreibung“)
  3. Im Abschnitt „MCP-Server“ wähle die Server aus, die verwendet werden sollen
  4. Speichere die Einstellungen

Die KI nutzt dann automatisch die verknüpften Server während der Generierung.

Fehlerbehebung

Server-Verbindung fehlgeschlagen:

  • Überprüfe die Zugangsdaten (API-Key, URL)
  • Stelle sicher, dass der Server erreichbar ist
  • Prüfe, ob dein API-Kontingent noch nicht erschöpft ist

Keine Tools verfügbar:

  • Der MCP-Server antwortet möglicherweise nicht korrekt
  • Kontaktiere den Anbieter des MCP-Servers

Generierung dauert länger:

  • MCP-Server-Anfragen können die Generierungszeit verlängern
  • Dies ist normal, da zusätzliche Daten abgerufen werden

In-App Käufe

Was ist Prompts Plus?

Prompts Plus ist ein optionales Upgrade, das erweiterte Funktionen für deine KI-Textgenerierung freischaltet. Das wichtigste Feature ist die MCP-Server-Integration, mit der du externe Datenquellen einbinden kannst.

Enthaltene Funktionen

  • MCP-Server-Integration: Verbinde externe Tools und Datenquellen
  • Erweiterte Prompt-Optionen: Verknüpfe Server mit spezifischen Prompts
  • Echtzeit-Datenabfragen: Die KI greift während der Generierung auf aktuelle Daten zu

Kauf und Aktivierung

  1. Der Kauf erfolgt direkt über den Shopware Store
  2. Nach dem Kauf wird das Feature automatisch aktiviert
  3. Du siehst den Status in der „In-App Käufe“ Übersicht im Log-Bereich

Abonnement-Informationen

  • Prompts Plus ist ein monatliches Abonnement
  • Die verbleibende Laufzeit siehst du in der „In-App Käufe“ Übersicht
  • Das Abonnement verlängert sich automatisch über den Shopware Store

Systemvoraussetzungen

  • Shopware Version 6.6.9.0 oder höher
  • Aktive Internetverbindung für MCP-Server-Kommunikation

FAQs

AI Shop Assistant: General questions

This can have several causes:

  1. An already installed plugin has used an incorrect database type: https://forum.shopware.com/t/unknown-database-type-enum-requested/98651
  2. You could insert the following once in the Shopware configuration: https://symfony.com/doc/current/doctrine/dbal.html#registering-custom-mapping-types-in-the-schematool

Please ensure that the API key has a credit balance.

In this case, the application has not been installed correctly. Most of the time, this occurs in conjunction with other plugins that cause an error in your database configuration. If you uninstall our app, you should receive an error message from the database that gives an indication of which plugin caused the error. If you cannot interpret the error message, please let us know. (Contact form).

50 AI calls are activated for the test phase. If you are seriously interested in the APP and would like to continue using it after the test phase, you are welcome to contact us. (Contact form)

To configure the app, please follow the steps below:

1. Upload the app in the administration under "Extensions/My Extensions".

2. Install and activate the app.

3. Go to the app configuration and add the required ChatGPT API key.

4. Choose which AI model you use

5. You can specify which metadata should be changed. This option is intended for mass updating.

For online shops with a wide range of products, a higher temperature setting of 0.6-1 is often beneficial to generate varied and interesting descriptions. For shops with many similar products, a lower temperature setting of 0.2-0.4 should be used to ensure consistency and clarity. It can be useful to test different temperature settings to find out which best suits the specific requirements and style of the online shop.

This is a known problem. It is a Shopware error that prevents the apps from being deactivated in certain situations. Shopware has already been informed about this.

As a workaround, there is currently only the option of setting the "allow_disable" flag to 1 in the "App" table of the database.

The problem is expected to be fixed with Shopware 6.7.1.0. The issue for the error can be found here: https://github.com/shopware/shopware/issues/9068

Open AI: General questions

The OpenAI account can be accessed via this link: https://platform.openai.com/ .

Here you will find the management of the credit, limitation and the API keys, which are necessary for the extension!

  1. A new project must be created first.
  2. A new API key can be created here and assigned to a project.

This key can be entered in the configuration of the extension.

Credits are required for the OpenAI API to function properly. As with other services, this is a "prepaid" method. You top up the credits and these are automatically "spent" as soon as the extension is used and a request is sent.

To top up credits with OpenAI, a payment method must be set up via your own profile.

The payment methods can be found under: Organisation -> Billing. A new payment method can be added here via "add payment details". If this is done, credits will also be topped up.

Credits are billed in dollars. OpenAI uses "tokens" - one token corresponds to approx. 4 characters.

Overview table from OpenAI https://openai.com/api/pricing/:

If, for example, only a certain budget of credits per month is available or for protection, limits can be set.

Monthly limits can be set via the profile: Organisation -> Limits, as well as an e-mail notification as soon as a certain amount is reached.

OpenAI provides usage in different "tiers". Initially, every user is in the "Free" tier, which severely restricts the use of the AI models to a few requests per minute. As soon as $5 has been paid for, these restrictions are greatly reduced and the extension can be used more efficiently.

Further information on the limitations can be obtained directly from OpenAI: https://platform.openai.com/docs/guides/rate-limits/usage-tiers?context=tier-free

Which current "tier" has been reached can be seen in the account here: https://platform.openai.com/settings/organization/limits

If there is interest in having someone else check settings and limits, members can be invited to join the organisation.

You can find the area here: https://platform.openai.com/settings/organization/members

A new person can be added here via "invite". There are two role options to choose from, here you can specify whether the new member can only view certain data or whether they can contribute to the organisation. As the first option is quite limited and certain tasks cannot be assigned, it is advisable to assign the "Owner" option.

For technical reasons, the extension only works on the domain for which it was purchased.

Staging and test environments are not supported.

We recommend that you always use the latest model.

For cost reasons, the predecessor model can also be used.

What is reasoning?

Reasoning is a function that instructs the AI model to go through a „thought process“ to avoid hallucinations and increase comprehensibility before giving an answer.

Reasoning from the ChatGPT5.1 model:

Since the ChatGPT5.1 model, OpenAI has offered reasoning that has significantly improved response times and the quality of responses.

Nein, das Plugin funktioniert vollständig ohne Prompts Plus. Die MCP-Server-Integration ist ein optionales Upgrade für erweiterte Anwendungsfälle.

Das Plugin unterstützt alle MCP-Server, die dem Model Context Protocol Standard entsprechen. Gängige Beispiele sind SEO-Tools, Keyword-Recherche-Dienste und Analyse-Plattformen.

Die Generierung kann etwas länger dauern, da zusätzliche Daten von externen Servern abgerufen werden. Die verbesserte Textqualität durch die Echtzeitdaten kompensiert dies in der Regel.

Ja, für die Abfrage werden relevante Informationen (z.B. Produktname, Keywords) an die konfigurierten MCP-Server gesendet. Achte darauf, nur vertrauenswürdige Server zu verwenden.

Hinweis: Biloba IT übernimmt keine Verantwortung für Daten, die an fremde Server übertragen werden.

Ja, du kannst jeden MCP-kompatiblen Server einbinden. Trage einfach die entsprechenden Zugangsdaten in der Server-Verwaltung ein.

Die Textgenerierung funktioniert auch ohne MCP-Server. Wenn ein Server nicht erreichbar ist, wird der Text ohne die zusätzlichen Daten generiert.

No solution found?

Describe your problem to us - we'll take a look at it and get back to you as soon as possible.

Submit a support request