Converters#

Converters service provides bulk CSV ingestion by parsing uploaded CSV files and publishing their rows as platform messages. It is intended for batch import of historical or bulk-collected sensor data without requiring device-level publishing.

Two input formats are supported:

• SenML — the first row is the header where column names become measurement names. The first column must be a Unix timestamp. All remaining columns are treated as numeric measurement values.
• JSON — column names become JSON field names. The value from the column named by the thing's profile transformer time_field setting is parsed as a numeric Unix timestamp and stored as a Created field in the payload. The first column is excluded from the payload. Values are auto-parsed as numbers where possible, otherwise treated as strings.

Large files are processed in batches with a 30-second pause between batches to avoid overloading the message broker. For SenML, a batch is flushed every 50,000 SenML records (one record per value column per row). For JSON, a batch is flushed every 50,000 rows.

Authentication#

Requests must be authenticated with the device's internal key using the Authorization: Thing <key> header.

SenML CSV#

The first column must be a Unix timestamp. Subsequent columns are measurement names taken from the header row.

Example file (readings.csv):

time,voltage,current,power
1709635200,120.1,1.2,144.12
1709635260,119.8,1.3,155.74

curl -s -S -i -X POST \
  -H "Authorization: Thing <thing_key>" \
  -F "file=@readings.csv" \
  https://localhost/converters/csv/senml

Rows are collected and published as batched SenML messages. Each SenML record carries the measurement name (n), value (v), and timestamp (t) from its column and row.

JSON CSV#

Column names become JSON field names. If the profile transformer time_field is set and matches a column name, that column's value is parsed as a numeric Unix timestamp (seconds or nanoseconds as a float) and stored as a Created field in the payload. The first column is always excluded from the payload. Values that cannot be parsed as numbers are kept as strings.

Example file (events.csv), assuming time_field is set to time in the profile transformer config:

time,temperature,humidity,status
1709635200,21.5,60,ok
1709635260,22.1,58,ok

curl -s -S -i -X POST \
  -H "Authorization: Thing <thing_key>" \
  -F "file=@events.csv" \
  https://localhost/converters/csv/json

Rows are collected and published as batched JSON messages. The time column value is parsed as a float and stored as Created in the payload. The remaining columns — temperature, humidity, and status — become fields in the JSON payload. The resulting record looks like:

{"Created": 1709635200, "temperature": 21.5, "humidity": 60, "status": "ok"}

For the full API reference, see the API documentation.