AI_AGENT_OS/docs/kci-processing-jobs/file-processing-conventions.md

File Processing Conventions

Use this file to document conventions shared by KCI processing jobs.

Mark unknowns instead of guessing.

File Locations

Location Type	Path or Description	Environment	Notes
Input	Unknown	Unknown	Unknown
Output	Unknown	Unknown	Unknown
Archive	Unknown	Unknown	Unknown
Error / reject	Unknown	Unknown	Unknown
Logs	Unknown	Unknown	Unknown

File Naming

Document any naming conventions used by processing jobs.

Examples to fill in:

• Prefix:
• Date format:
• Sequence number:
• Customer/account identifier:
• Extension:
• Temporary file naming:
• Completed file naming:
• Error file naming:

File States

Recommended state model:

received
validated
processing
processed
archived
rejected
error

Document the actual KCI state model when known.

PDF Processing Notes

Capture shared PDF rules here:

• Expected PDF source:
• Required page size:
• Required orientation:
• Naming rules:
• Merge/split rules:
• OCR rules:
• Validation rules:
• Output destination:

Retention

File Type	Retention Period	Cleanup Method	Owner
Input files	Unknown	Unknown	Unknown
Output files	Unknown	Unknown	Unknown
Logs	Unknown	Unknown	Unknown
Error files	Unknown	Unknown	Unknown

Reprocessing Principles

Before rerunning a job, identify:

• original input file
• output files from the failed or partial run
• database records created or changed
• archive state
• downstream systems already notified
• duplicate risks

Do not rerun a job until duplicate and partial-processing risks are understood.

Security And Privacy

• Do not put customer data samples in documentation unless they are sanitized.
• Do not commit credentials or connection strings.
• Document where configuration lives without exposing secret values.
• Avoid copying full production paths if they reveal sensitive structure; use environment labels when needed.