Skip to content

user_guide.md

Latest commit

 

History

History
48 lines (37 loc) · 2.1 KB

user_guide.md

File metadata and controls

48 lines (37 loc) · 2.1 KB

User Guide - SETS WARP Backend

Overview

The backend provides essential services for the WARP application, including data contribution, knowledge updates, and access to the latest centrally-trained models.

Client-Side Integration

1. Contributing Data

Clients can contribute to the community knowledge base by sending data to the /contribute endpoint.

  • Endpoint: POST /contribute
  • Request Body:
    • install_id: Unique identifier for the client installation.
    • phash: Perceptual hash of the icon crop.
    • crop_png_b64: Base64-encoded PNG of the icon crop.
    • item_name: Confirmed name of the item.
    • warp_version: Version of the client application.

2. Fetching Knowledge

Clients can retrieve the latest merged knowledge base to improve local recognition.

  • Endpoint: GET /knowledge
  • Response: A JSON object mapping perceptual hashes to item names.

3. Checking for Model Updates

The client can check if a new centrally-trained model is available for download.

  • Endpoint: GET /model/version
  • Response: Contains available: true and a version (hash) if an update exists.

Admin Procedures

1. Merging Contributions

Confirmed contributions can be merged into the master knowledge.json using the admin_merge.py script or the /admin/merge endpoint.

  • Requirement: X-Admin-Key header with the secret ADMIN_KEY.

2. Manual Model Training

Training can be manually triggered by running admin_train.py with appropriate flags:

  • python admin_train.py --train --min 1

3. Automated Training

Automated training is handled by GitHub Actions and can be triggered by Hugging Face webhooks through main.py.

Support and Troubleshooting

  • Rate Limiting: Users are limited to a specific number of requests per day to prevent abuse.
  • Health Check: Verify if the backend is online at /health.

Related Documentation