Notion to Weixin

Config (Optional)

Set a default author in ~/.agents/config.yaml:

notion_to_weixin:
  author: "Alice Wang"

You can also set a global default:

author: "Alice Wang"

Workflow (Sequential)

1. Resolve Notion page ID from the given title.
2. Export the page to Markdown.
3. Ensure author exists in Markdown front matter (and inject a byline if needed).
4. Process images (upload to Weixin and replace URLs in Markdown).
5. Prepare thumb_media_id:
   • If the Notion page has a cover, upload it as a Weixin thumb material and use that ID.
   • If no cover, use the last image material ID.
6. Generate digest with AI summarization (if user did not provide one).
7. Create a new Weixin draft using Markdown input (--format markdown) and always pass --css-path.

Inputs (Ask the user if missing)

• notion_title: exact Notion page title to publish.
• author: optional author name (used for Markdown front matter and node-wxcli --author). If missing, read from ~/.agents/config.yaml.
• css_path: required when using --format markdown; if missing, default to assets/default.css.
• thumb_media_id: only required if cover and image fallback both fail.
• digest: optional user-provided summary; if missing, generate via AI from the final Markdown.

Prerequisites

• notion-cli installed and authenticated (NOTION_TOKEN or notion auth set).
• node-wxcli installed and authenticated (node-wxcli auth set or node-wxcli auth login).
• curl and jq available.

Step 1: Resolve Page ID by Title

1. Copy references/notion-search.json to a temp file and replace __TITLE__.
2. Run notion search --body @query.json.
3. Choose the exact-title match. If multiple matches remain, pick the most recently edited or ask the user to confirm.
4. If no match is found, ask for a Notion page ID or URL and extract the ID.

Step 2: Export Page to Markdown

• Use notion pages export <page_id> --assets=download -o <workdir>/page.md when the page contains images.
• Use a workdir like /tmp/notion-to-weixin/<slug-or-timestamp>.

Step 3: Default Author Handling

• If author is not provided, read it from ~/.agents/config.yaml.
• Ensure author exists in Markdown front matter (and add a byline if you want it visible in content).

Step 4: Process Images (Recommended)

If the Notion page contains images, upload them to Weixin and replace Markdown image URLs so they remain valid for draft rendering.

1. Find image links in <workdir>/page.md. Notion exports local images under the assets folder.
2. For each local image file, upload it and capture the url:

node-wxcli material upload --type image --file <workdir>/assets/<image-file> --json | jq -r '.url'

3. Replace the Markdown image URL with the returned url:

![alt](<weixin-url>)

If you skip this step, images that reference local files or expiring Notion URLs may not render in Weixin.

Step 5: Prepare thumb_media_id

5.1 If Notion page has a cover

1. Fetch page metadata (use notion-cli):

notion pages get <page_id> > <workdir>/page.json

2. Extract cover type + cover URL:

cover_type=$(jq -r '.cover.type // ""' <workdir>/page.json)
cover_url=$(jq -r '.cover | if .==null then "" elif .type=="external" then .external.url else .file.url end' <workdir>/page.json)

3. If cover type is file, download via notion-cli; if external, use curl:

if [ "$cover_type" = "file" ]; then
  jq -c '.cover' <workdir>/page.json | notion files read --body @- --output <workdir>/cover.jpg
else
  curl -L "$cover_url" -o <workdir>/cover.jpg
fi
thumb_media_id=$(node-wxcli material upload --type thumb --file <workdir>/cover.jpg --json | jq -r '.media_id')

Note: Notion file URLs expire quickly. If download fails, re-fetch page metadata and retry.

5.2 If no cover exists

1. Get image material count:

image_count=$(node-wxcli material count --json | jq -r '.image_count')

2. If image_count > 0, fetch the last image and use its media_id:

offset=$((image_count - 1))
thumb_media_id=$(node-wxcli material list --type image --offset "$offset" --count 1 --json | jq -r '.item[0].media_id')

3. If no images exist, ask the user to provide a thumb_media_id or upload a thumb manually.

Step 6: Build digest with AI summarization

If digest is not provided, summarize the final Markdown content with AI and generate a concise Chinese digest:

• 1-2 sentences
• Prefer 60-120 Chinese characters
• Plain text only (no Markdown / emoji / line breaks)
• Focus on the article purpose and key outcomes

Step 7: Create Weixin Draft (Markdown input)

When using Markdown format, always pass --css-path. If css_path is missing, set:

css_path=${css_path:-assets/default.css}

node-wxcli draft add \
  --title "<notion_title>" \
  --author "<author>" \
  --digest "<digest>" \
  --format markdown \
  --content - \
  --css-path <css_path> \
  --need-open-comment=1 \
  --only-fans-can-comment=0 \
  --thumb-media-id "$thumb_media_id" < <workdir>/page.md

• If you need machine-readable output, add --json and capture the returned media_id.

Resources

• references/notion-search.json: JSON template for Notion title search.
• references/cli-commands.md: Canonical CLI command examples for notion-cli and node-wxcli.
• assets/default.css: Default CSS theme for --format markdown (--css-path).

Notes:

• Use --page-id if the title search is ambiguous.
• Use --thumb-media-id only if cover and image fallback are unavailable.