# Indexes

An index is an entity, like a table in SQL, with a specific schema definition. It gathers a collection of documents with the structure defined by the schema.

An index is defined by an unique identifier uid that is generated by MeiliSearch (if none is given) on index creation. It also has a name to help you track your different indexes.

# Index UID and name

The uid of an index is its unique identifier. It is the :uid parameter on every indexes/:uid route.

The name is the only mandatory information needed on index creation.

It helps you track your different indexes with a human-readable name.

    "name": "Movie",
    "uid": "12345678",
    "createdAt": "2019-11-20T09:40:33.711324Z",
    "updatedAt": "2019-11-20T10:16:42.761858Z"

# Schema definition

A schema is a representation of the attributes of the document and how MeiliSearch will handle these fields.

In the schema definition, each attribute has one or multiple of the following properties :

  • identifier: The unique identifier of a document (e.g id)
  • indexed: the search engine will search inside those fields.
  • displayed: Fields that will appear in the returned documents.
  • ranked: Ranked fields are used to create ranking rules.

As a result, it is possible to have indexed fields that are not displayed or displayed fields that are not indexed. Depending on your documents and your needs, this could be useful.


The only mandatory document field is the identifier.

# Fields order

The order of the fields represents their relevance in the search engine.

Thus, if a title field is defined before a description field, its content will be considered more relevant to a search query than that of a "description" field.

This means that if you search for something that matches in the description of the document A and in the title of the document B, the document B will be considered better than the document A. You can read more about these rules in the ranking section.


The order of the document fields has a huge impact on the relevancy. So please order fields from the most important to the less.

# Example

Take, for example, a movie collection. We have several fields:

  • id: The unique identifier of a movie.
  • title: Title of the movie, will be showcased
  • description: A description of the movie.
  • release_date: The release date of the movies.
  • poster: The URL of the poster or the image related to the movie.

This is the schema:

    "id": ["identifier", "indexed", "displayed"],
    "title": ["indexed", "displayed"],
    "description": ["indexed", "displayed"],
    "release_date": ["ranked", "displayed"],
    "poster": ["displayed"]

The fields are sorted by relevancy importance.

The id field is the unique identifier of a movie document. We want to display it (displayed) and give the possibility to users to search for the movie using his ID (indexed).

The title field is an important field, which is why it is ranked second in the order of fields. We want our search queries to search inside this field (indexed), and we also want to showcase it to the user (displayed).

The description field is exactly like the title field. Its 3rd position in the order of the fields puts it after Title in relevancy.

The release_date field is not indexed because the user does not usually search for a precise date. With the ranked property, we can create a custom ranking rule that will make the recent movies more relevant than the older ones in the search engine.

Finally, the poster field contains the image URL to the movie's poster. We do not want to search inside the URL, that's why we omitted the indexed property.

When no schema is given at the creation of an index, the schema is inferred. Inference follows strict rules to index correctly.

# Identifier

This property is given to the attribute that contains the unique key of each document.

If two documents added at the same time have the same ID, MeiliSearch will only save the last one.

If a document is added when there is already a document with the same ID in MeiliSearch, it will be updated.


Documents identifiers are always converted into strings. Only strings and integers are valid identifiers. It means that it is, for example, forbidden to use arrays or objects as an identifier.

# Indexed

It defines the fields that will be used to find the documents.

For example, an url field is not necessarily interesting for the search engine. So by omitting this property, the url will not be indexed but can still be returned inside the documents with the displayed property.

# Displayed

It defines the fields that will be returned when querying MeiliSearch. A non-displayed field will never appear in the search response.

For example, if the field contains a large amount of data, returning it would slow down server returns. By not storing it but still indexing it, the data will be used by the search engine but not returned in the documents.

# Ranked

This property allows the creation of a custom ranking rule. See custom ranking rules for more information.

# Inferred Schema

When creating an index, MeiliSearch expects a name for the index and a schema definition. If no schema is defined before adding documents, the schema will be inferred.

The inference of the schema is based on the first document added to MeiliSearch. Then, by following the inference rules, the schema will be created.

# Inference rules

The schema is inferred this way:

  • the order of the fields is the order of the schema fields
  • the identifier is the first field containing "id" (case insensitive)
  • every field is indexed and displayed

The order of the fields inside the document will determine their relevance in the search engine.

To determine the identifier, an attribute that contains the case insensitive string id is expected. Thus, _id, myId, for example, are correct keys.

This field will receive the identifier property, so it should contain the unique identifier of a document.


If the identifier field is missing, the inference will not be completed, and the documents will not be added.

Every other field will have the indexed and displayed properties.


By default, MeiliSearch infers the schema from the first document sent.