> For the complete documentation index, see [llms.txt](https://blog.mantago.cc/en/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://blog.mantago.cc/en/other/database.md).
# Database
Before the AI Recommendation Engine can start suggesting products to your customers, there's one thing to nail down first: getting all your product data into your **Database**. And this database isn't just for recommendations — it can also serve as the product data source for **GAI** (generative Q\&A): build it once, and both AI recommendations and GAI can draw from it. The richer your database, the closer the bot's picks and answers land to what shoppers actually want. Let's build one from scratch together. Just follow along!
> 💡 TL;DR: There are three ways to build a database. **Build it manually** (Item Management, adding products one by one), **connect an e-commerce platform** (CYBERBIZ / SHOPLINE / shopify, auto-synced), or **connect your own API** (Customer URL). This data feeds both the **AI Recommendation Engine** and **GAI Q\&A**, so the more complete your fields and the clearer your item descriptions, the sharper the recommendations — and the better GAI answers product questions too.
***
#### Step 1: Open Database and Click Add
1. From the left-hand menu, head into **Database**.
2. You'll land on a list of all your databases. Each one shows its **status** (Running / Pending Update / Error), **product count**, **update frequency**, and **last sync time**, so you can read the health of every database at a glance.
3. Click **Add** in the top-left corner to start a new one.
#### Step 2: Pick a Data Source
Once you click Add, the **Add Database** window pops up. This is where you decide how your database gets built, and MantaGO gives you three ways to do it:
* **Build it manually (Item Management)**: Add products one by one. Great when you don't have many items, or when you want full hands-on control over the content.
* **Connect an e-commerce platform (CYBERBIZ / SHOPLINE / shopify)**: If your store runs on one of these, connect it directly and let your products sync in automatically.
* **Connect your own API (Customer URL)**: Got your own API? Pick this and let MantaGO pull products from your end on a schedule.
We'll walk through the two most common picks in detail: **Item Management** (manual) and **Customer URL** (your own API). The e-commerce integrations work much the same way, so once you can pick a source and map your fields, you're all set.
***
#### Method A: Build It Yourself with Item Management
If you don't have many products, or you just want to control every entry yourself, Item Management is the most direct route.
**First, name your database**
1. In the source window, click **Item Management**.
2. In the **Data Source Name** field, type something easy to recognize (say, "Mid-Autumn Gift Boxes").
3. Click **Create** in the top-right corner, and your database is ready to go!
**Add items one by one**
Once your database is built, you'll land on the management page. Now let's get some products in:
1. Click **Add Item** in the top-left corner.
2. In the **Add Item** window, fill in these required fields:
* **Item Name**: What the product is called.
* **Item URL**: The product page link, so customers can click straight through to buy.
* **Item Description**: The product's selling points, uses, and who it's for. The AI leans heavily on this, so be specific.
3. Want a product image? Click **Upload Photo** to add one.
4. Click **Save** in the top-right corner, and that's one item done!
> Quick tip: Below the form you'll also spot **Add Auto Tag**, **Add Category Tag**, and **Add Content Hashtag**, which let you attach **Item Tags (category)** and **Hashtags (description)** to your items. Both matter a lot for AI recommendations, and we'll dig into them in the Advanced section below. Get your basics in first; you can always come back and add these later.
**Advanced: Customize item fields**
When the default fields (Item Photo, Item Name, Item Description, Item URL) aren't enough, add your own:
1. On the management page, click **Setting** in the top-right corner.
2. In **Item Field Settings**, type a field name into the input box below (for example, "Special Price") and click **Confirm** to add it.
Now every item can carry one more detail you care about, and the AI can factor it into recommendations too.
**Advanced: Meet Item Tags and Hashtags, and feed your items smarter**
With the basics in place, the secret to sharper AI recommendations comes down to two things: **Item Tags** and **Hashtags**. Both help the engine decide "should I show this product to this shopper?", but they play very different roles. One handles **categorization**, the other handles **description**.
| Term | Role | What it does |
| ------------ | -------------- | ------------------------------------------------------------------------------------------ |
| **Item Tag** | Categorization | Files a product into the right category and hierarchy, like setting it on a specific shelf |
| **Hashtag** | Description | Spells a product out through specific traits, so it's clear what it is and who it's for |
Picture a mobility aid, and the difference clicks right away:
* **Item Tag (categorization)**: `Assistive Devices > Mobility Aids > Manual Wheelchair`, `High-Back Reclining`, `Eligible for Long-Term Care Subsidy`. This tells the engine what category the product is, where it sits in the hierarchy, and what conditions it qualifies for.
* **Hashtag (description)**: `Max load 100 kg`, `Seat width 45cm`, `Headrest included, reclinable`, `Good for those who can't sit long`, `Easy transfers`, `Weighs about 15 kg`. This breaks the product's specs and features into bite-sized descriptions, so the engine can match a shopper asking for "a wheelchair that reclines and holds enough weight."
Easy way to remember it: **Tags put a product in its place; Hashtags let it speak.** Keep both in good shape and your AI recommendations land accurate and on point. Back in the Add Item form, you'll use **Add Category Tag** for Item Tags and **Add Content Hashtag** for descriptions.
**Advanced: Generate Content Hashtags**
Don't fancy adding Hashtags one by one? Click **Generate Content Hashtag** in the top-right corner, and MantaGO will auto-generate descriptive tags for your items, rounding out their features so the AI Recommendation Engine can match shopper needs more easily.
**Advanced: Export and Import for batch item creation**
When you've got a lot of items, there's no need to build them one at a time:
* **Export**: Download your existing items as a file, handy for backups or for batch editing in a spreadsheet.
* **Import**: Upload a prepared file all at once and a whole batch of items appears in a flash, way faster than doing it by hand. The **required fields** for import are **Item Name, Item Description, and Item URL**. All three must be there, or the import won't go through.
> 💡 The fastest way to create lots of items at once: build one entry first, hit **Export** to grab the file format, fill in all your products in a spreadsheet following that format (don't forget the three required fields: Item Name, Item Description, Item URL), then **Import** them all in one go.
***
#### Method B: Connect Your Own API with Customer URL
If your products already live in your own system and you've got an API ready, don't move everything by hand. Let MantaGO pull from your end on a schedule instead: less hassle, and no stale data to worry about.
**Fill in the connection basics**
1. In the source window, click **Customer URL**.
2. Fill in the **Data Source Name**, again something easy to recognize.
3. Set these connection parameters:
* **API URL**: Your product API's address (for example, `https://api.example.com/product`).
* **Update Frequency**: How often to pull (say, every 1 day), keeping your data fresh on a regular cadence.
* **Authentication**: Pick **No Auth**, **Basic Auth**, or **API Key**, depending on what your API needs.
* **Path**: If the product data sits at a certain level inside the response, point to it here.
**Map the API fields**
This step is how MantaGO makes sense of what your API returns. On the left are MantaGO's **field names**; on the right, fill in the matching **API fields** from your side and pair them up:
1. In the **API Field Settings** area, map each field:
* `id` → the id field in your API
* `Item Name` → e.g. `title`
* `Item Description` → e.g. `description`
* `Item URL` → e.g. `link`
2. Need more fields? Click **Add Field** to add them (the screen shows how many you're using, e.g. 4 / 10).
3. If your API returns results in pages, under **Pagination Settings** map "Page Number" to `page` and "Items Per Page" to `perpage`, so MantaGO pulls everything.
4. Once it's all mapped, click **Create Data Source** in the top-right corner.
From here, MantaGO auto-syncs at the frequency you set, keeping your database always current. That matters a lot for the AI Recommendation Engine, since it'll only ever recommend products that exist and are valid right now.
> ⚠️ When mapping fields, the names on the right have to match what your API actually returns **exactly**, case and all. Get one wrong and no data comes through. If the database list shows an **Error** status, it's usually a connection setting or a field mapping that's off, so come back and check this page.
***
#### What's next?
By now your database has something in it. Whether you built it by hand, connected an e-commerce platform, or linked an API, the goal is the same: **before you switch on the AI Recommendation Engine, make sure MantaGO has a complete, up-to-date product list to work with.**
The better you keep your database, and the more on-point your item descriptions, the closer the bot's in-chat picks and answers land to what shoppers want. This list feeds both the **AI Recommendation Engine** and **GAI Q\&A** — build it once, and it pays off in both places. Lay a solid foundation first, and the smart recommendations and generative answers that follow will slot right in and hit the mark.
Head into the backend and build your first database now! MantaGO comes with a **14-day free trial**, so feed in your product data as you explore, and get ready to let your bot do the recommending for you. 🚀
