Appearance
@inferedge/moss / SerializedIndex
Interface: SerializedIndex
Complete serialized representation of an index for backup and transfer purposes.
Contains all data necessary to recreate an index, including configuration, items, and pre-computed embeddings. Used by export/import operations.
Properties
dimension
dimension:
number
The dimensionality of the embedding vectors.
All vectors in the index will have this many dimensions. Higher dimensions can capture more semantic nuance but require more storage and computation.
embeddings
embeddings:
number
[][]
Pre-computed embedding vectors for all items in the index.
Each vector corresponds to an item at the same array position. These vectors enable fast similarity search without needing to recompute embeddings at query time.
The vectors are arrays of floating-point numbers with length equal to dimension
.
id
id:
string
The unique identifier of the index.
items
items:
Item
[]
All items currently stored in the index.
This array contains the original item data that was added to the index.
metric
metric:
string
The distance metric used for similarity calculations.
Common values:
- "cosine": Measures angular similarity (most common for text)
- "euclidean": Measures straight-line distance
- "manhattan": Measures grid-based distance
modelId
modelId:
MossModel
The embedding model used to generate vector representations.
This determines how text is converted to numerical vectors for similarity search.
textFieldId
textFieldId:
string
The field identifier used for text content processing.
Specifies which field of items contains the text content to be embedded.