I/O Format Checking

backend.check_io() runs the backend against sample input and checks that the outputs meet Packflow’s format requirements: correct types, matching row counts, and JSON-serializable values. If all checks pass, the outputs are returned for inspection.

The optional tests.py file included in generated projects contains a test_check_io test that calls backend.check_io() with sample data. Update the sample inputs to match the backend’s expected input format before running.

Rules

Packflow’s validators enforce a series of rules to ensure fully normalized inputs/outputs. For reference, the rules are listed below:

  • Correct Input Format:

    • When calling check_io() or invoking a backend directly, inputs may be provided as a single dictionary or a list of dictionaries. Single-dictionary inputs are normalized to a list by the framework before reaching the backend.

    • The pipeline entry-point (transform_inputs if defined, otherwise execute) receives a list of dictionaries for passthrough and records preprocessors, or an ndarray for numpy. Preprocessing still runs during .check_io() - data reaches the entry-point in the format determined by input_format.

  • Inputs and Outputs are same length:

    • The number of output dictionaries must equal the number of input dictionaries.

    • Tip: Build in exception handling and return an empty dictionary if a row fails, when possible

  • Correct Output Format:

    • The final pipeline step (transform_outputs if defined, otherwise execute) must return a list of dictionaries equal in length to the input list. Intermediate steps may pass data in any format.

  • Outputs must be JSON Serializable:

    • Output cannot contain non-native types (e.g., Numpy Arrays) that are not serializable with the json library.

Packflow’s built-in validation helpers check these rules and will warn if some conditions are not met – however, it is recommended to keep these items in mind when developing.

See also

Packflow Inference Backend - for a full explanation of the pipeline I/O contract, including how single-dict inputs are handled at the framework level and where list-of-dict is required.

Running I/O Format Checks

Use the Inference Backend’s built-in .check_io() method to run all format checks against sample input data.

This can be accomplished programmatically:

 1 # -- Import backend --
 2 from main import Backend
 3
 4 backend = Backend()
 5
 6 # Replace this with data relevant to the Inference Backend implementation
 7 sample_data = {"number": 5}
 8
 9 outputs = backend.check_io(sample_data)
10
11 # If all checks passed, the outputs will be returned and can be visually inspected
12 print(outputs)

Best Practices

  • Use realistic sample data: When running I/O format checks, use sample data that closely resembles the expected input during actual inference.

  • Integrate with unit tests: Packflow’s format checks are most effective when run alongside tests written for the backend’s expected behavior.

Warning

Packflow’s format checks run only when explicitly invoked via the .check_io() method. They do not run automatically during inference, i.e. in .__call__() or any other pipeline method.

Next Steps

Continue to the Preparing for Distribution documentation page for instructions on how to prepare a Packflow project that can be shared across environments for replicable inference.