[FEATURE] Selective Metadata Redaction

[erik-sv]

Problem Statement

Some use cases require limiting or removing certain metadata fields before sharing AI-generated content, either for privacy or compliance reasons.

Proposed Solution

• Add support for --redact or --filter-keys flag in CLI and Python API.
• Users can specify fields to omit (e.g. model_id, timestamp, etc).
• Could be implemented as part of the encode or a separate redact function.

Alternative Solutions

• Developers manually remove fields before encoding.
• Not practical and prone to errors.

Use Cases

1. A publishing platform wants to remove internal session_id fields before exporting public content.
2. A research org wants to share output but redact user_id to comply with privacy rules.

Implementation Ideas

• Accept list of keys to exclude.
• Update encoder to drop fields before serialization.
• Add CLI option like: encypher redact --input text.txt --omit user_id session_id

Additional Context

Useful for privacy-conscious users and organizations.