Skip to main content
State is structured records: a collection name, a key, and JSON data. Collections are named groups of records (for example orders or items). Add creates an empty collection with no schema. Empty: no collections yet — add one, or wait until the Workflow writes State. A client chat can change a collection that is not listed on the workflow. A collection listed on the workflow changes from a client chat only when Clients can write state is on. The owner can always change both. Select a collection to inspect Records or Schema.

Records

Browse records or open one for full JSON. Edit opens the JSON for that key. Save replaces the record. The key stays. If the record changed after you opened it, save is refused. Open it again. You can clear the collection (keep name and schema) or delete the collection. Neither can be undone. Add record opens a panel with Key and JSON Data. An existing key is refused. When a schema is saved, the record must match it. Import loads a csv, xlsx, or xls file into this collection. Choose the column that becomes the record key. Importing that key again updates the record. A schema is optional. When a schema is saved, a row that does not match it is rejected and nothing from that file is written. A file larger than 10 MB, or a sheet with more than 50,000 rows, is rejected. The file is not added to Files. A csv uses commas and is read as UTF-8, or as Windows-1251 when the file is not valid UTF-8. When a workbook has several sheets, choose one. The first row is the column names. First row contains column names is on. Turn it off when the file has no names: the columns are col1, col2, and every row is data. Export downloads this collection as csv. The first column is Key, then the record fields. An empty cell means the field is absent. A nested value is JSON text; importing that file again stores it as text. Importing the same key updates the record. A large collection is several csv files, each at most 50,000 records and 10 MB, in key order. Export downloads every part. When there is more than one part, the download is a zip of those files. A single record larger than 10 MB is not exported.

Keys

A key starts with a letter or digit in any language, then letters, digits, spaces, and _ . : -, up to 256 characters. Shanghai:Moscow:40HC and Шанхай Москва are both valid keys. It cannot contain @, so an email address stays in a column. Spaces at the start and end are dropped. The same word in different Unicode forms (for example, й as one character or as и plus a mark) is one key.

Schema

Optional JSON for the fields of one record. Paste it in JSON Schema. The same JSON goes in Schema on Collections.
Here origin and price must be present. dest may be missing. status is one of two strings. Each tag is a string. Any other field is rejected. Validate sample checks a pasted record against the JSON above, even if you have not saved it yet. Clear removes the schema and keeps the collection. Without a schema, records stay free-form. With a schema, a saved record must match it. A record that does not match is rejected, and nothing from that save is written. The message names the record key, each failing field, and the reason, such as qty: required or price: expected type number, got string. An agent gets the same message, so it can fix those fields and save again.

Keywords

A keyword that is not in this table is refused. That includes $ref, pattern, format, allOf, anyOf, and oneOf. minLength above maxLength, minimum above maximum, or minItems above maxItems is refused.

Limits

Import from a spreadsheet extract with Import from a file after Extract tables. Write records with Save records. Change the same fields on matching records with Update records. Query with Read records.

When records already exist

Save on Schema, for a collection you added on the Instance, replaces the schema and keeps every record, including ones that do not match. The next save, import, or Workflow write must match. A collection with the Workflow badge is changed on the Graph. Stored records are not opened. The new schema is used only when it still allows every value the previous schema allowed. A new field that is not required is used even if a stored record already has another value under that name. An empty collection accepts any schema. No schema is always used. The draft is still saved. A collection that does not fit keeps the previous schema. Stricter additionalProperties means other fields were allowed and now are not, or an open object becomes a schema.

Keep records

Keep for days on a Graph collection is a whole number from 1 to 3650, or empty. Empty keeps records. The count is days since the last change, not since the record was created. A later write keeps the record. Saving a shorter number, or the first number on a collection that already has records, deletes records past that age on the next cleanup. The Graph stores the number. Cleanup runs in the background. An empty declared collection stays in the list. A collection added on the Instance, and not on the Graph, is kept.

Workflow

Collections from the Graph are already in the list. The badge Workflow marks them. The schema is read-only. Save and Clear are not on Schema. If the Graph left the schema empty, Schema does not show an editor. Delete collection is not on Records. Clear collection still removes the records and keeps the collection. Add creates a collection that is not on the Graph. Updating the Instance to a version whose schema does not fit stored records is refused. The Instance stays on the current version. A collection removed from the Graph stays on the Instance. You can edit its schema and delete it.