Model settings¶
A model is used to describe the structure of the data stored in a business application. For most usage, it is mandatory to declare at least one model.
It is composed of the following options :
- Structures and types of all the datas, declared in the form of a tree
- Optional "natural language synonyms" for the assistant system
- Advanced options for the backend elasticsearch supporting the data storage (purge, indices rotation, ...)
Facts are the backbone of a Myrtea application, but they can only be attached to a single model.
Models must be built based on the target facts, using a top-down approach.
Data structures and Datatypes¶
Supported field types¶
| Type | Category | Comment |
|---|---|---|
string | Primitive | Any alphanumeric chain |
integer | Primitive | Round number |
float | Primitive | Floating-point number |
datetime | Primitive | Date and time |
object | Advanced | Complex object with nested fields |
Field Leaf¶
| Attribute | JSON Type | Comment | Example |
|---|---|---|---|
name | String | Field name | label |
type | String | Field type (based on the previous table) | string |
synonyms | Array of String | Liste de synonymes métier pour le modèle | ["label", "title"] |
Field Object¶
| Attribut | JSON | Values | Description |
|---|---|---|---|
name | String | Nom du champ | |
type | String | Object | Décrit le type du champs |
synonyms | Array of String | Liste de synonymes métier pour le modèle | synonyms dans un field objet ? plutôt dans l’id ? |
fields | Array of Object | Liste de champs du modèle | |
keepObjectSeparation | Boolean |
Elasticsearch Options¶
Rotation mode- "Cron" is the only choice for now. The indices rotations is based on a cron expression.
- It is advised to stay on a daily rotation for now (cron :
0 0 * * *)
Purge activation- Can be enabled or not
- Based on a second parameter "Index count before purge" to specify how many indices must be kept before any purge.
Alias patch depth- Used to specify the depth of a specific "Patch Alias" which target indices used for update on "old" data (stored in indices other than the current one)
Advanced options- Any other elasticsearch index options (ie: number_of_shards, number_of_replicas, ...)