[FEATURE]: Support Manual Management and Rolling Cleanup of Local Image Data

[bitifirefly]

📝 Feature Description [Please make everyone to understand it]

MineContext currently stores local image data generated during the recognition process. Over time, these image files can accumulate and occupy significant disk space. There is currently no built-in mechanism to manage, rotate, or clean up these local image assets.

This issue proposes adding support for manual management and scheduled rolling deletion of local image data, while retaining only the extracted semantic text after recognition.

👍 What problem does this feature solve

Prevent uncontrolled growth of local storage usage

Enable long-running or always-on deployments without manual disk maintenance

Reduce storage cost and operational risk

Align with privacy-by-design principles by minimizing retained raw data

In most use cases, once image recognition is completed, only the semantic text results are needed for downstream tasks. Persisting raw image data is unnecessary.

👾 What does the proposed API look like

Introduce configurable local image data management with the following capabilities:

Manual Management

Allow users to manually delete stored image files via configuration or CLI

Optionally expose image storage status (count, size)

Rolling / Scheduled Cleanup

Support automatic deletion based on:

Time (e.g. delete images older than N days)

Storage limit (e.g. keep total image size under X GB)

Count limit (e.g. keep only the latest N images)

Cleanup should not affect already-extracted semantic text

Data Retention Strategy

Persist only the recognized semantic text by default

Image retention should be optional and configurable

🚑 Any additional information

No response