Supported Formats
Maskura processes supported text objects through its WASM pipeline and Avro OCF through a typed schema-aware transform. The request Content-Type selects the format.
Currently supported
| Format | Content-Type | Notes |
|---|---|---|
jsonl | application/x-ndjson, application/jsonlines | One JSON object per line. PII/encryption applied per record. |
json | application/json | One complete JSON document, object, or array. |
csv | text/csv | Quote-aware record split; fields comma-delimited. |
tsv | text/tab-separated-values | Tab-delimited fields, same record model as CSV. |
text | text/plain | Free text; PII detection over the whole record. |
avro | application/avro, application/x-avro | Avro OCF. Schema-aware writes, multipart completion, and processed reads. |
Current behavior
- Supported text formats are transformed record by record on writes and processed reads. The same PII detection, redaction, envelope encryption, and stable encryption pipeline applies.
- Avro OCF processing is available in hosted Maskura. It uses a typed schema-aware transform (see Envelope Encryption) rather than the byte-oriented text plugin pipeline, and can select fields with
x-maskura-encrypt-fields. - Envelope-encrypted fields are decrypted by the private-key holder in the dashboard or SDK. Maskura does not decrypt customer content on the gateway read path.
- Other and unsupported binary Content-Types remain raw S3 passthrough today. They are stored and read without schema-aware transformation.
- Content-Type is the current format selector for hosted requests.
Binary format roadmap
Parquet
in development
Schema-aware columnar processing with encrypted staging for safe processed reads.
Avro
available
Schema-aware Object Container File processing for writes, multipart completion, and protected reads.
Hive Tables
coming soon
Partitioned table layout, schema, and _SUCCESS markers on top of the Avro and Parquet codecs.
PDF Redaction
coming soon
Render the text layer, run PII detection, and redact before storage.
Image Redaction
coming soon
Detect PII in images (cards, screenshots) and redact regions; optional OCR path.
ORC
coming soon
Columnar, like Parquet.
Guidance
Use JSON or JSONL for the general text plugin pipeline, orAvro OCF when you need a schema-aware binary container. Per-field envelope encryption is described inEnvelope Encryption. CSV and TSV are both supported. For Parquet/Hive-scale workloads, JSONL remains the pragmatic interim format until columnar support ships.