A summary index stores information that corresponds to a document. OpenSearch Retrieval Engine Edition can use the docID to obtain the location in which the information is stored and provide you with that storage location.
The schema of a summary index resembles that of a forward index, but the two serve different purposes and have different performance characteristics.
Configure a summary index
The following example shows a complete summary index configuration in schema.json:
"summarys": {
"summary_fields": ["id", "company_id", "subject", "cat_id"],
"compress": false,
"parameter": {
"compress_type": "uniq|equal",
"file_compress": "simple_compress1"
}
}Parameters
| Parameter | Description | Valid values | Default |
|---|---|---|---|
summary_fields | Fields to store in the summary index. Supports all data types. | Any field name | — |
compress | Compresses the summary index using zlib. | true, false | false |
parameter | Wrapper object for advanced compression settings. Requires V3.9.1 or later. | — | — |
compress_type | Encoding method for the summary index. For details on uniq and equal, see Forward index compression. | uniq, equal | — |
file_compress | File-level compression. Specify the alias of a compressed file as defined in schema.json. | A valid alias | — |
Choose a compression method
Two compression approaches are available and can be used independently or together.
| Method | Description | Recommended for |
|---|---|---|
compress (zlib) | Compresses the entire summary index at the zlib level. | Simple setups or low-query-rate scenarios. |
file_compress | Compresses at the file level using a named alias. Delivers better performance for compression than compress. | Production workloads where read performance matters. |
For production use, set compress to false and configure file_compress instead.
Usage notes
summary_fieldsaccepts fields of any data type.When a field of the TIMESTAMP data type is added to a summary index, the field is stored as an attribute. The system processes the time format centrally when the default time zone is in use.
The
parameterfield, includingcompress_typeandfile_compress, requires OpenSearch Retrieval Engine Edition V3.9.1 or later.Define the compressed file alias in
schema.jsonbefore configuringfile_compress.