md2we-ai-assistant

Use MD2WE online AI endpoints for title suggestions, summary generation, and image generation.

When To Use

Use this skill when the user wants to:

• Generate five Chinese title options from an article draft
• Generate a concise Chinese summary for share cards or WeChat digest text
• Generate a 16:9 editorial illustration for the article

AI Config

Requests can pass ai_config explicitly, or rely on server-side defaults.

Supported environment variables:

OPENAI_API_KEY=...
OPENAI_BASE_URL=...
OPENAI_TEXT_MODEL=...
OPENAI_IMAGE_TOOL_MODEL=...

Server-side defaults only support OPENAIAPIKEY, OPENAIBASEURL, OPENAITEXTMODEL, and OPENAIIMAGETOOL_MODEL.

If text and image generation need different credentials or base URLs, pass them explicitly inside aiconfig in the request body. Do not rely on OPENAIIMAGEAPIKEY or OPENAIIMAGEBASE_URL as server-side defaults.

Service Check

export MD2WE_BASE_URL="${MD2WE_BASE_URL:-https://md2we.com}"
curl -sS "${MD2WE_BASE_URL}/api/health"

API Call Pattern

Title suggestions:

curl -sS "${MD2WE_BASE_URL}/api/ai/title-suggestions" \
  -H "Content-Type: application/json" \
  -d @<(jq -Rs '{
    markdown: .,
    focus_prompt: "更像媒体标题"
  }' article.md)

Summary:

curl -sS "${MD2WE_BASE_URL}/api/ai/summary" \
  -H "Content-Type: application/json" \
  -d @<(jq -Rs '{
    markdown: .,
    focus_prompt: "突出结论"
  }' article.md)

Image generation:

curl -sS "${MD2WE_BASE_URL}/api/ai/generate-image" \
  -H "Content-Type: application/json" \
  -d @<(jq -Rs '{
    markdown: .,
    focus_prompt: "科技感、克制、媒体头图"
  }' article.md)

Pass explicit AI config when needed:

curl -sS "${MD2WE_BASE_URL}/api/ai/summary" \
  -H "Content-Type: application/json" \
  -d @<(jq -Rs \
    --arg api_key "$OPENAI_API_KEY" \
    --arg base_url "${OPENAI_BASE_URL:-https://generativelanguage.googleapis.com/v1beta/openai}" \
    --arg text_model "${OPENAI_TEXT_MODEL:-gemini-2.5-flash}" \
    --arg image_model "${OPENAI_IMAGE_TOOL_MODEL:-gemini-2.5-flash-image}" \
    '{
      markdown: .,
      focus_prompt: "突出结论",
      ai_config: {
        text: {
          api_key: $api_key,
          base_url: $base_url,
          model: $text_model
        },
        image: {
          api_key: $api_key,
          base_url: $base_url,
          model: $image_model
        }
      }
    }' article.md)

Result Handling

• title returns suggestions
• summary returns summary
• image returns imageurl and revisedprompt

If the server reports missing AI credentials, surface that directly and stop. Do not invent fallback output.