File Processing & Routing

Between receiving a file and delivering it, a workflow may need to validate its contents, change its format, encrypt it, or choose a destination. Design these as distinct stages so that a failure leaves enough information to understand what happened and decide what can safely run again.

Establish the Input and Output

For each stage, define the files it accepts and what it produces. A conversion might read one CSV and produce one JSON file. An unzip operation can produce many files. A validation might report a problem without changing the original at all.

Use separate locations for received files, working output, files awaiting review, and delivery-ready results when those states have different permissions or handling. The folder names are your convention; moving a file into a folder called approved only means approval if your process and access controls make it so.

Preserve the original when you need to investigate a parsing error or repeat a conversion. A copy kept in the same live workflow is useful working history, but an independent backup has a different purpose and retention policy.

Select and Route Files

Automations v1 selects files through paths and patterns and can copy or move them with new destination names. Those actions organize files; they do not rewrite their contents. Several v1 Automations triggered together do not execute in a guaranteed order.

For a production process that includes external conversion software, let that software write completed results to a dedicated output location, then transfer those results. The scripts and application workflow guidance explains where that coordination can live.

Automations v2 connects processing and routing nodes, including transformations and conditional branches. It is an opt-in early beta and is not recommended for business-critical production workflows. Evaluate a v2 design with test inputs and isolated destinations under its beta terms.

Transform or Validate Contents

TransformScript maps structured data between formats and layouts. For example, it can convert a supplier's CSV column names and date values into the JSON structure an accounting application expects. It runs in the v2 Transform node, subject to the beta terms, or through the interactive Transform File action.

Expectation content validation uses TransformScript to check a delivery without rewriting the files. A missing required field can therefore be a validation failure rather than a transfer failure. Decide whether a failed check should require correction by the sender or a review within your organization.

Test transformations with malformed records, empty input, unexpected values, and realistic file sizes as well as a valid sample. Matching the output format does not establish that its business values are correct.

Apply Folder Processing Deliberately

A folder setting and an Automation action have different triggers. Automatic GPG processing runs after uploads and copies into a configured folder. Moving or renaming a file into that folder does not trigger encryption or decryption.

GPG output appears as a new file and can trigger subsequent file-creation Automations. Account for the output filename and existing-file behavior when designing that next step. On a Remote Server Mount, GPG requires buffered uploading through Files.com before delivery to the remote system; it is not a direct stream that bypasses temporary storage.

Use AI Where Judgment Is Needed

AI Tasks can assess text-based documents and produce reports or take actions according to a prompt. They run unattended, their output can vary, and they do not automatically retry. Use fixed rules for a required record layout or exact filename calculation; use AI where interpreting the content is the work.

If someone must approve the result, make review a separate stage with a clear handoff. An AI Task does not pause mid-run for human approval. Record how rejected or uncertain results are handled before connecting the output to automatic delivery.