May 2026

In the last month, Sensible introduced a powerful new feature for bringing dynamic external context into document extractions and released a new advanced JsonLogic operator.

New feature: Bring external context into document extraction for validation and transformation

You can now attach extra_data, a flat key/value object, to an asynchronous extraction request and read its values inside an extraction config. Use this feature to bring pipeline context into the extraction step, such as expected values from a system of record, fields from a prior extraction, or output from an LLM agent.

Common use cases include cross-document validation (extract values from a loan application, then pass them as extra_data into a subsequent bank statement extraction to check consistency) and incorporating external data from third-party lookups. To compare extra_data values to extracted fields, use the Custom Computation method for deterministic comparisons or LLM-based methods with the Source IDs parameter configured for semantic comparisons. When you attach extra_data to a portfolio extraction, Sensible passes the same object to every document in the portfolio.

For more information, see the Extra data computed field method.

New feature: Random number generation in JsonLogic

In addition to the existing JsonLogic operators for transforming extracted data, Sensible released the new Random operator. The operator returns a random float in the range [0, 1) and takes no arguments. For example, you can use it to flag roughly half of your extractions for human review by comparing the result to 0.5. For more information, see JsonLogic extensions.