🔑 1. Activation – License Entry
In this section, you enter:
• Pricelist AI License
• AI Expiration Date
• WhatsApp Component License (if purchased)
• WhatsApp Component Expiration Date

👉 This is only for enabling additional AI features.

If the WhatsApp license is not active, WhatsApp functions do not appear.

⚙️ 2. Setup – General Settings

🔧 Mode
• You have two modes: Master and Slave.
• These are typically used in multi-site or multi-instance systems.
If you are unsure which to use → leave it in Master (default).

🔐 OpenAI API Key
The API key provided to the client should be entered here.

It is used to generate:
• AI content,
• automatic prompt filling,
• topic/keyword processing,
• automatic Caspar functions.

🌐 Allowed Hosts (separated by 😉
Used to specify which domains are authorized to call the AI.

Typical example:
https://miosito.it;https://admin.miosito.it

🔎 Also used to prevent AI calls from duplicate sites or test environments.

⚡ Enhanced Mode
Internal option that enables advanced features.
If activated:
• enables advanced debugging,
• enables detailed logging,
• enables certain internal optimizations.

👈 Normally disabled, used only for testing.

3. Melchior – AI Management of the Price List
It is the most important part. It is divided into 4 tabs.

📌 3.1 Prompt Generator
From here, you select the price lists on which the AI should generate or update content.
• The AI will work only on the selected price lists.
• This is for clients who have multiple price lists (e.g., 2024, 2025) and want the AI to respond only with a specific price list.
👉 I select “Price List 2025” → The AI responds exclusively with the prices from the 2025 price list.

🔔 Also shown here are:
• errors,
• missing topics,
• unconfigured blocks.

📊 3.2 Topic Stats
Show AI-generated statistics:
• total number of prompts per topic,
• reserved prompts,
• non-reserved prompts,
• any deficiencies.

📌 The topics come from the Gutenberg blocks of the site.

🏷️ 3.3 Keywords
Each topic has a list of keywords used to recognize sections of the price list.
• Here the AI stores the associations created automatically.
• You can modify them manually (Edit Keywords button).

🔎 It is used to help the AI understand:
• categories,
• collections,
• models,
• segments of the price list.

3.4 RAG – Melchior RAG Sync
Here you can start the RAG synchronization.

📌 What is it for?
Creates a vector memory of the pricelists → the AI can respond coherently and retrieve information directly from the pricelist database.
It has two buttons:
• ▶️ Run RAG
• ⛔ Force Stop (to force the stop)

📦 4. Caspar – Standard sliders
Page where the user defines the tags for technical data.
Example:
• Width →
• Gross weight →
• Number of packages →

👉 These tags are used in AI prompts.
👉 The AI automatically populates them when generating technical content.
⚠️ It is crucial to use only item tags here, not header tags.

🚀 5. Production – Production Mode Once everything is ready: • switch to Production mode • lock critical changes • ensure stability for real use Used when the price list is final.

The Pricelist AI block allows you to create customized prompts directly within WordPress pages, enabling the artificial intelligence to learn additional instructions and provide consistent responses based on the site’s content and configured price lists.

It is used for:
• providing additional information to the AI;
• teaching details not present in structured data (codes, tables, prices, images, etc.);
• linking the prompt to a topic to make it part of the learning system;
• defining priorities and behavior of the prompt during response generation.

Block Fields

1. Content
In this field, you can enter free text that will be used by the AI as instructional basis. It may include:
• additional explanations;
• clarifications;
• technical notes;
• textual descriptions of elements not present in the catalog data (such as an image, a product feature, exceptions, constraints, etc.).

Important:
In this block, it is not possible to use tags like , , etc. It is a purely descriptive space.

2. Load Saved Prompt / Save Prompt 💾
These two functions work together with the “Content” field.
• Save Prompt: saves the content written in “Content” so it can be retrieved in the future.
• Load Saved Prompt: quickly loads a previously saved prompt.
They are convenient when:
• frequently reusing the same instructions;
• creating variants of the same prompt;
• working on many similar pages.

3. Topic 🔖
Each AI block must be associated with a topic.
The topic determines which semantic area the prompt belongs to (e.g., bases, wall units, model, collection, etc.).

This allows the AI system to:
• organize the learned material;
• know when to use that prompt based on the end user’s query;
• improve the accuracy of responses.

👉 A block can have only one selected topic.

4. Priority ⭐
Priority is used to define the order in which the AI considers various prompts related to the same topic.
Example:
• Priority 0 = main prompt
• Priority 1, 2, 3… = supporting or supplementary prompts

The higher the priority, the greater the importance of the prompt during response generation.

5. Reserved 🔒
Indicates how the prompt should be treated by the AI system:

• No → the prompt is “general” and can be used freely.
• Yes → the prompt is reserved, thus used only in specific contexts (e.g., internal tests, particular pricing logics, prices, or sensitive information).

This system helps to keep separate:
• prompts intended for responses to the end user;
• prompts intended for tests, private technical descriptions, or non-public data.