Create an index

Create a search index over your JSON documents with FT.CREATE, choose the right field types, and understand how Redis indexes arrays.

This is step 2 of the Redis Search tutorial. You should already have the catalog loaded as JSON documents.

So far you can fetch a product only if you already know its key. An index — also called a secondary index, because it's a lookup structure maintained alongside your primary data — changes that: it tells Redis which fields to track and how, so you can ask questions like "which products cost less than $100?" or "which ones mention wireless?" and get answers quickly.

What an index does

When you create an index, you give Redis three things:

1. What to index — which keys belong to the index, selected by a key prefix (here, product:).
2. The data type — whether those keys hold hashes (ON HASH) or JSON documents (ON JSON).
3. The schema — which fields to index, the path to each one, and what type each field is.

Once the index exists, Redis keeps it up to date automatically. Any product: document you add or change after creating the index is indexed immediately, and the documents you loaded earlier are indexed right away.

Field types

Redis Search has a few core field types. Choosing the right one for each field determines how you can query it:

For the catalog, a good mapping is: name and description are TEXT (you want word search), brand and category are TAG (exact labels), and price, rating, review_count, stock, and release_year are NUMERIC. The features field is a list of exact labels, so it is also a TAG.

Create the index

Use FT.CREATE to define the index. Because the data is JSON, each field is identified by a JSONPath expression, and AS gives it a short alias to use in queries:

schema = (
    TextField("$.name", as_name="name"),
    TagField("$.brand", as_name="brand", sortable=True),
    TagField("$.category", as_name="category"),
    TextField("$.description", as_name="description"),
    NumericField("$.price", as_name="price", sortable=True),
    NumericField("$.rating", as_name="rating", sortable=True),
    NumericField("$.review_count", as_name="review_count"),
    NumericField("$.stock", as_name="stock"),
    NumericField("$.release_year", as_name="release_year", sortable=True),
    TagField("$.features[*]", as_name="features"),
)
index = r.ft("idx:catalog")
index.create_index(
    schema,
    definition=IndexDefinition(prefix=["product:"], index_type=IndexType.JSON),
)

A few things to notice:

• PREFIX 1 product: means "index every key that starts with product:". The 1 is the number of prefixes that follow.
• AS name, AS price, ... define the alias you use in queries (@name, @price). Without an alias, you would have to write the full JSONPath in every query.
• SORTABLE on a field lets you sort results by it efficiently. Add it to fields you expect to sort by, such as price and rating.
• $.features[*] ends in [*], which matters for arrays. More on that next.

You only create an index once. If you make a mistake, remove it with FT.DROPINDEX (this deletes the index, not your documents) and create it again.

Indexing arrays: the [*] you should not forget

The features field is a JSON array like ["wireless", "bluetooth", "waterproof"]. To index each element as its own tag, the JSONPath ends in [*]:

$.features[*] AS features TAG

This is the single most common point of confusion when indexing JSON, so it is worth understanding:

• With $.features[*], Redis indexes wireless, bluetooth, and waterproof as three separate tags. A query for @features:{waterproof} matches the document.
• With $.features (no [*]) on a JSON array, the behavior is not what you want for filtering element-by-element.

Check the index

After creating the index, you can confirm it picked up your documents. FT.INFO reports details about an index, including how many documents it contains:

info = r.ft("idx:catalog").info()
print("Documents indexed:", info["num_docs"])  # >>> Documents indexed: 12

Look for num_docs in the output; it should be 12, one for each product you loaded.

Next steps

Your data is indexed. Continue to searching and filtering to start asking questions of it.