Face Transform for Dify

Use the Ace Data Cloud Face Transform APIs in Dify workflows. Maintained by Ace Data Cloud. The plugin is free; API calls require your own authorized account and use the current service pricing.

Setup

Activate the service at Ace Data Cloud, check current pricing, and create an API token with the required service access.
Install from the official Dify Marketplace once the submission is approved and published. During review, use Dify's documented package/debug installation in a test workspace. A GitHub PR does not establish Marketplace availability or default installation.
In Dify's Plugins / Tools page, authorize this provider with Bearer Token (). Do not include credentials in prompts or exported workflows. Credential validation never generates media. For this synchronous service, saving validates the credential shape; the first real call verifies service access.

Tools

See CAPABILITIES.md for the current MCP comparison and parameter equivalents. All exposed inputs follow the current published API; model combinations and availability still depend on the service.

Run a workflow

For generation, use Start → generation tool → task retrieval → Output. Fill the prompt/text and model, and enter arrays/objects as JSON. Optional values can be left empty. The example requests in tests/contract-examples.json show valid shapes; example.org URLs are placeholders that must be replaced with your own accessible media.

A submission can return with . Save that ID, then use the retrieval tool with to read once, or 1–240 for a bounded wait. If still pending, query the same task again. Disable automatic retries on generation nodes. No paid request is automatically retried and no substitute model is selected.

, , , , , , and are available as Dify variables. Only a terminal successful result has ; intermediate previews remain pending. Batch queries preserve the state of each item. Terminal task failures raise a tool error. Synchronous search, text and management results are returned directly in /. The plugin does not execute model-generated tools.

The table maps service operations to Dify tools. Different MCP helper functions may use the same action selector or structured JSON input.

Task/query calls retry transport failures at most twice. Generation has one attempt and a 10-second connect / 60-second read timeout. After a timeout, inspect request history before resubmitting; the accepted task may still be running. Delete/archive operations require .

Branding and privacy

The plugin uses the exact existing system asset recorded in branding provenance, for both light and dark icons. No logo was generated or redrawn. Asset SHA256: .

Requests go directly to . The plugin passes reference URLs to that API and returns media links; it does not fetch arbitrary reference URLs. Dify may fetch/render output links under its own policies. Never submit media you lack permission to process. See PRIVACY.md.

API charges are recorded in Credits in your Ace Data Cloud account. Check the actual usage ledger; Dify execution counts are not a billing ledger. USD = Credits × your current package price / amount.

Development and evidence

Python 3.12 is required. Install , then run:

Source contracts, MCP mappings, brand provenance and offline cases are in . Recorded real Dify results state their exact coverage; they do not establish all models/options or Dify Cloud/Marketplace installation. See tests/README.md.

Source: https://github.com/AceDataCloud/FaceDify
Issues: https://github.com/AceDataCloud/FaceDify/issues
Contact: [email protected]
License: MIT
Simplified Chinese