WP Manifestindependent plugin directory
manifest / unclassified / ai-provider-for-siliconflow

Bestony AI Provider for SiliconFlow

SiliconFlow provider for the WordPress AI Client

by Bestony · github.com/bestony/ai-provider-for-siliconflow · website

★ 0stars
0forks

Install

No release zip yet. The repository archive installs, but the folder name will carry the branch suffix and updates will not flow:

wp plugin install https://github.com/bestony/ai-provider-for-siliconflow/archive/refs/heads/main.zip
Readme github.com/bestony/ai-provider-for-siliconflow ↗

A sibling WordPress plugin that registers SiliconFlow with the WordPress AI Client. It supports text generation, documented vision models, and synchronous text-to-image generation.

Installation

  1. Copy this directory to wp-content/plugins/bestony-ai-provider-for-siliconflow/.
  2. Activate the plugin.
  3. Open Settings → Connectors, select SiliconFlow, and save an API key.

The SDK is supplied by WordPress 7.0 or the AI plugin. The provider is silent when the SDK is missing.

Configuration

The API key is owned by the AI Client registry. This plugin does not read the connectors option directly. Optional environment variables or PHP constants are:

Setting Default Purpose
SILICONFLOW_BASE_URL https://api.siliconflow.cn/v1 OpenAI-compatible API base URL
SILICONFLOW_DEFAULT_MODEL deepseek-ai/DeepSeek-V4-Flash Preferred text and vision model
SILICONFLOW_IMAGE_MODEL Kwai-Kolors/Kolors Preferred image model
SILICONFLOW_STRUCTURED_OUTPUT json_schema json_schema, json_object, or none
SILICONFLOW_MODEL_INPUT_MODALITIES catalog Set to text,image to force vision for chat models
SILICONFLOW_ENABLE_WATERMARK enabled Set to 0 only when downstream processing adds the required watermark
SILICONFLOW_IMAGE_SIZE_SQUARE model default Override image size for square requests
SILICONFLOW_IMAGE_SIZE_LANDSCAPE model default Override image size for landscape requests
SILICONFLOW_IMAGE_SIZE_PORTRAIT model default Override image size for portrait requests
SILICONFLOW_REQUEST_TIMEOUT 120 Request timeout in seconds
SILICONFLOW_CONNECT_TIMEOUT 10 Connection timeout in seconds

SiliconFlow-specific chat parameters such as enable_thinking, thinking_budget, reasoning_effort, min_p, and top_k can be passed through customOptions. The provider only declares standard options documented for all compatible chat models.

Image generation

The image endpoint receives prompt and image_size, and returns a remote image URL. The URL expires after one hour, so the calling feature must download it promptly. Only one remote image candidate is declared. Supported image custom options include negative_prompt, seed, num_inference_steps, guidance_scale, and cfg.

The provider sends X-Enable-Watermark: 1 by default. Set SILICONFLOW_ENABLE_WATERMARK=0 only when the downstream workflow adds the explicit watermark required by SiliconFlow.

Image editing, video, audio, embedding, reranking, and asynchronous jobs are not implemented. Those model IDs remain visible without capabilities.

Development

php scripts/selfcheck.php
php scripts/selfcheck.php --sdk=/path/to/php-ai-client/src
for file in $(find . -name '*.php' -not -path './.git/*'); do php -l "$file"; done

The self-check never makes a network request and does not require an API key.

License

GPL-2.0-or-later. See LICENSE.