> ## Documentation Index > Fetch the complete documentation index at: https://www.meilisearch.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # Documents > Documents are the individual items that make up a dataset. Each document is an object composed of one or more fields. A document is an object composed of one or more fields. Each field consists of an **attribute** and its associated **value**. Documents function as containers for organizing data and are the basic building blocks of a Meilisearch database. To search for a document, you must first add it to an [index](/resources/internals/indexes). Nothing will be shared between two indexes if they contain the exact same document. Instead, both documents will be treated as different documents. Depending on the [index's settings](/reference/api/settings/list-all-settings), the documents might have different sizes. ## Structure Diagram illustration Meilisearch's document structure ### Important terms * **Document**: an object which contains data in the form of one or more fields * **[Field](#fields)**: a set of two data items that are linked together: an attribute and a value * **Attribute**: the first part of a field. Acts as a name or description for its associated value * **Value**: the second part of a field, consisting of data of any valid JSON type * **[Primary Field](#primary-field)**: a special field that is mandatory in all documents. It contains the primary key and document identifier ## Fields A **field** is a set of two data items linked together: an attribute and a value. Documents are made up of fields. An **attribute** is a case-sensitive string that functions as a field's name and allows you to store, access, and describe data. That data is the field's **value**. Every field has a data type dictated by its value. Every value must be a valid [JSON data type](https://www.w3schools.com/js/js_json_datatypes.asp). If the value is a string, it **[can contain at most 65535 positions](/resources/help/known_limitations#maximum-number-of-words-per-attribute)**. Words exceeding the 65535 position limit will be ignored. If a field contains an object, Meilisearch flattens it during indexing using dot notation and brings the object's keys and values to the root level of the document itself. This flattened object is only an intermediary representation—you will get the original structure upon search. You can read more about this in our [dedicated guide](/resources/internals/datatypes#objects). With [ranking rules](/capabilities/full_text_search/relevancy/ranking_rules), you can decide which fields are more relevant than others. For example, you may decide recent movies should be more relevant than older ones. You can also designate certain fields as displayed or searchable. Some features require Meilisearch to reserve attributes. For example, to use [geosearch functionality](/capabilities/geo_search/getting_started) your documents must include a `_geo` field. Reserved attributes are always prefixed with an underscore (`_`). ### Displayed and searchable fields By default, all fields in a document are both displayed and searchable. Displayed fields are contained in each matching document, while searchable fields are searched for matching query words. You can modify this behavior using the [update settings endpoint](/reference/api/settings/update-all-settings), or the respective update endpoints for [displayed attributes](/reference/api/settings/update-displayedattributes), and [searchable attributes](/reference/api/settings/update-searchableattributes) so that a field is: * Searchable but not displayed * Displayed but not searchable * Neither displayed nor searchable In the latter case, the field will be completely ignored during search. However, it will still be [stored](/capabilities/full_text_search/how_to/configure_displayed_attributes#data-storing) in the document. The `fieldDistribution` object returned by the [`/stats` route](/reference/api/stats) is not affected by `searchableAttributes` or `displayedAttributes`. Even if a field is neither displayed nor searchable, it still appears in `fieldDistribution` with its document count. Rely on `displayedAttributes` and permissions, not on stats output, to keep a field name hidden from clients. To learn more, refer to our [displayed and searchable attributes guide](/capabilities/full_text_search/how_to/configure_displayed_attributes). ## Primary field The primary field is a special field that must be present in all documents. Its attribute is the [primary key](/resources/internals/primary_key#primary-field) and its value is the [document id](/resources/internals/primary_key#document-id). If you try to [index a document](/resources/internals/primary_key#primary-field) that's missing a primary key or possessing the wrong primary key for a given index, it will cause an error and no documents will be added. To learn more, refer to the [primary key explanation](/resources/internals/primary_key). ## Upload By default, Meilisearch limits the size of all payloads—and therefore document uploads—to 100MB. You can [change the payload size limit](/resources/self_hosting/configuration/reference#payload-limit-size) at runtime using the `http-payload-size-limit` option. Meilisearch uses a lot of RAM when indexing documents. Be aware of your [RAM availability](/resources/help/faq#what-are-the-recommended-requirements-for-hosting-a-meilisearch-instance) as you increase your batch size as this could cause Meilisearch to crash. When using the [add new documents endpoint](/reference/api/documents/add-or-update-documents), ensure: * The payload format is correct. There are no extraneous commas, mismatched brackets, missing quotes, etc. * All documents are sent in an array, even if there is only one document ### Dataset format Meilisearch accepts datasets in the following formats: * [JSON](#json) * [NDJSON](#ndjson) * [CSV](#csv) #### JSON Documents represented as JSON objects are key-value pairs enclosed by curly brackets. As such, [any rule that applies to formatting JSON objects](https://www.w3schools.com/js/js_json_objects.asp) also applies to formatting Meilisearch documents. For example, an attribute must be a string, while a value must be a valid [JSON data type](https://www.w3schools.com/js/js_json_datatypes.asp). Meilisearch will only accept JSON documents when it receives the `application/json` content-type header. As an example, let's say you are creating an index that contains information about movies. A sample document might look like this: ```json theme={null} { "id": 1564, "title": "Kung Fu Panda", "genres": "Children's Animation", "release-year": 2008, "cast": [ { "Jack Black": "Po" }, { "Jackie Chan": "Monkey" } ] } ``` In the above example: * `"id"`, `"title"`, `"genres"`, `"release-year"`, and `"cast"` are attributes * Each attribute is associated with a value, for example, `"Kung Fu Panda"` is the value of `"title"` * The document contains a field with the primary key attribute and a unique document id as its value: `"id": "1564"` #### NDJSON NDJSON or jsonlines objects consist of individual lines where each individual line is valid JSON text and each line is delimited with a newline character. Any [rules that apply to formatting NDJSON](https://github.com/ndjson/ndjson-spec) also apply to Meilisearch documents. Meilisearch will only accept NDJSON documents when it receives the `application/x-ndjson` content-type header. Compared to JSON, NDJSON has better writing performance and is less CPU and memory intensive. It is easier to validate and, unlike CSV, can handle nested structures. The above JSON document would look like this in NDJSON: ```json theme={null} { "id": 1564, "title": "Kung Fu Panda", "genres": "Children's Animation", "release-year": 2008, "cast": [{ "Jack Black": "Po" }, { "Jackie Chan": "Monkey" }] } ``` #### CSV CSV files express data as a sequence of values separated by a delimiter character. Meilisearch accepts `string`, `boolean`, and `number` data types for CSV documents. If you don't specify the data type for an attribute, it will default to `string`. Empty fields such as `,,` and `, ,` will be considered `null`. By default, Meilisearch uses a single comma (`,`) as the delimiter. Use the `csvDelimiter` query parameter with the [add or update documents](/reference/api/documents/add-or-update-documents) or [add or replace documents](/reference/api/documents/add-or-replace-documents) endpoints to set a different character. Any [rules that apply to formatting CSV](https://datatracker.ietf.org/doc/html/rfc4180) also apply to Meilisearch documents. Meilisearch will only accept CSV documents when it receives the `text/csv` content-type header. Compared to JSON, CSV has better writing performance and is less CPU and memory intensive. The above JSON document would look like this in CSV: ```csv theme={null} "id:number","title:string","genres:string","release-year:number" "1564","Kung Fu Panda","Children's Animation","2008" ``` Since CSV does not support arrays or nested objects, `cast` cannot be converted to CSV. ### Auto-batching Auto-batching combines similar operations in the same index into a single batch, then processes them together. This significantly speeds up the indexing process. Tasks within the same batch share the same values for `startedAt`, `finishedAt`, and `duration`. If a task fails due to an invalid document, it will be removed from the batch. The rest of the batch will still process normally. If an [`internal`](/reference/errors/overview#errors) error occurs, the whole batch will fail and all tasks within it will share the same `error` object. #### Auto-batching and task cancellation If the task you're canceling is part of a batch, Meilisearch interrupts the whole process, discards all progress, and cancels that task. Then, it automatically creates a new batch without the canceled task and immediately starts processing it.