@moxn/kb-migrate 0.4.12 → 0.4.13

package/dist/targets/notion.js

@@ -62,14 +62,68 @@ function stripInvalidLinks(text) {
         return displayText;
     });
 }
-function sectionsToMarkdown(sections, options) {
-    const parts = [];
+function notionImageBlock(url, alt) {
+    return {
+        object: 'block',
+        type: 'image',
+        image: {
+            type: 'external',
+            external: { url },
+            ...(alt ? { caption: [{ type: 'text', text: { content: alt } }] } : {}),
+        },
+    };
+}
+function notionPdfBlock(url, caption) {
+    return {
+        object: 'block',
+        type: 'pdf',
+        pdf: {
+            type: 'external',
+            external: { url },
+            ...(caption ? { caption: [{ type: 'text', text: { content: caption } }] } : {}),
+        },
+    };
+}
+function notionFileBlock(url, caption) {
+    return {
+        object: 'block',
+        type: 'file',
+        file: {
+            type: 'external',
+            external: { url },
+            caption: [{ type: 'text', text: { content: caption || 'file' } }],
+        },
+    };
+}
+/**
+ * Convert KB document sections to Notion blocks.
+ *
+ * Uses a block-by-block approach: text blocks go through martian for
+ * rich formatting, while media blocks (images, PDFs, files, CSVs) are
+ * converted directly to native Notion block types. This avoids the
+ * fragility of placeholder-based post-processing.
+ *
+ * Section names become H2 headings (mirrors the import convention).
+ */
+function sectionsToNotionBlocks(sections, options) {
+    const allBlocks = [];
     const allReferences = [];
     const databaseIds = [];
-    const media = [];
     const extractRefs = options?.extractReferences ?? false;
+    // Accumulate contiguous text blocks into markdown, then flush through martian
+    let pendingMarkdown = [];
+    function flushText() {
+        if (pendingMarkdown.length === 0)
+            return;
+        const md = pendingMarkdown.join('\n').trim();
+        if (md) {
+            allBlocks.push(...markdownToBlocks(md));
+        }
+        pendingMarkdown = [];
+    }
     for (const section of sections) {
-        parts.push(`## ${section.name}\n`);
+        // Section heading — accumulate as markdown so martian handles it
+        pendingMarkdown.push(`## ${section.name}\n`);
         for (const block of section.content) {
             if (block.blockType === 'text' && block.text) {
                 let text = stripCommentTags(block.text);
@@ -78,93 +132,40 @@ function sectionsToMarkdown(sections, options) {
                     text = cleanedText;
                     allReferences.push(...references);
                 }
-                // Strip relative/internal links that aren't valid URLs
-                // (Notion rejects links without a protocol)
                 text = stripInvalidLinks(text);
-                parts.push(text);
-                parts.push('');
+                pendingMarkdown.push(text);
+                pendingMarkdown.push('');
             }
             else if (block.blockType === 'image' && block.url) {
-                const token = `MOXNMEDIA${media.length}PLACEHOLDER`;
-                media.push({ token, type: 'image', url: block.url, alt: block.alt });
-                parts.push(token);
-                parts.push('');
+                // Flush any pending text, then add image directly
+                flushText();
+                allBlocks.push(notionImageBlock(block.url, block.alt));
             }
             else if (block.blockType === 'document' && block.url) {
-                const token = `MOXNMEDIA${media.length}PLACEHOLDER`;
-                media.push({ token, type: 'file', url: block.url, filename: block.filename });
-                parts.push(token);
-                parts.push('');
+                flushText();
+                // Use native pdf block for PDFs (inline viewer), file block for others
+                if (block.mimeType === 'application/pdf') {
+                    allBlocks.push(notionPdfBlock(block.url, block.filename));
+                }
+                else {
+                    allBlocks.push(notionFileBlock(block.url, block.filename));
+                }
             }
             else if (block.blockType === 'csv' && block.url) {
-                const token = `MOXNMEDIA${media.length}PLACEHOLDER`;
-                media.push({ token, type: 'embed', url: block.url, filename: block.filename || 'data.csv' });
-                parts.push(token);
-                parts.push('');
+                flushText();
+                allBlocks.push(notionFileBlock(block.url, block.filename || 'data.csv'));
             }
             else if (block.blockType === 'database_embed' && block.databaseId) {
-                // Collect database ID for Pass 1.5 export
                 databaseIds.push(block.databaseId);
-                // Add a placeholder in the markdown
-                parts.push(`> **[Database]** *(exported as inline database)*`);
-                parts.push('');
+                // Keep database placeholder as text (handled in Pass 1.5)
+                pendingMarkdown.push(`> **[Database]** *(exported as inline database)*`);
+                pendingMarkdown.push('');
             }
         }
     }
-    return { markdown: parts.join('\n').trim(), references: allReferences, databaseIds, media };
-}
-/**
- * Replace media placeholder paragraphs in Notion blocks with proper
- * image/file/embed blocks. Martian doesn't support images, so we
- * post-process the converted blocks.
- */
-function injectMediaBlocks(blocks, media) {
-    if (media.length === 0)
-        return blocks;
-    // Build a lookup from token to media info
-    const tokenMap = new Map(media.map((m) => [m.token, m]));
-    return blocks.map((block) => {
-        // Check if this is a paragraph containing a media placeholder
-        const b = block;
-        if (b.type !== 'paragraph' || !b.paragraph?.rich_text)
-            return block;
-        const text = b.paragraph.rich_text.map((rt) => rt.text?.content ?? '').join('').trim();
-        const mediaInfo = tokenMap.get(text);
-        if (!mediaInfo)
-            return block;
-        // Replace with proper Notion block
-        if (mediaInfo.type === 'image') {
-            return {
-                object: 'block',
-                type: 'image',
-                image: {
-                    type: 'external',
-                    external: { url: mediaInfo.url },
-                    ...(mediaInfo.alt ? { caption: [{ type: 'text', text: { content: mediaInfo.alt } }] } : {}),
-                },
-            };
-        }
-        if (mediaInfo.type === 'file') {
-            return {
-                object: 'block',
-                type: 'file',
-                file: {
-                    type: 'external',
-                    external: { url: mediaInfo.url },
-                    caption: [{ type: 'text', text: { content: mediaInfo.filename || 'document' } }],
-                },
-            };
-        }
-        // For CSV/embeds, use a bookmark block (Notion doesn't have native CSV embed)
-        return {
-            object: 'block',
-            type: 'bookmark',
-            bookmark: {
-                url: mediaInfo.url,
-                caption: [{ type: 'text', text: { content: mediaInfo.filename || 'file' } }],
-            },
-        };
-    });
+    // Flush remaining text
+    flushText();
+    return { blocks: allBlocks, references: allReferences, databaseIds };
 }
 // Max 100 blocks per API call
 const MAX_BLOCKS_PER_APPEND = 100;
@@ -243,7 +244,7 @@ export class NotionExportTarget extends ExportTarget {
      */
     async resolveAndAppendReferences(doc, notionPageId) {
         // Extract references from section content
-        const { references } = sectionsToMarkdown(doc.sections, { extractReferences: true });
+        const { references } = sectionsToNotionBlocks(doc.sections, { extractReferences: true });
         if (references.length === 0) {
             return { resolved: 0, unresolved: 0 };
         }
@@ -413,8 +414,7 @@ export class NotionExportTarget extends ExportTarget {
     // Notion page creation / update
     // ============================================
     async createNotionPage(doc) {
-        const { markdown, media } = sectionsToMarkdown(doc.sections, { extractReferences: true });
-        const blocks = injectMediaBlocks(markdownToBlocks(markdown), media);
+        const { blocks } = sectionsToNotionBlocks(doc.sections, { extractReferences: true });
         // First batch: up to 100 blocks as children of the new page
         const firstBatch = blocks.slice(0, MAX_BLOCKS_PER_APPEND);
         const remainingBlocks = blocks.slice(MAX_BLOCKS_PER_APPEND);
@@ -446,8 +446,7 @@ export class NotionExportTarget extends ExportTarget {
             },
         });
         await this.clearPageContent(notionPageId);
-        const { markdown, media } = sectionsToMarkdown(doc.sections, { extractReferences: true });
-        const blocks = injectMediaBlocks(markdownToBlocks(markdown), media);
+        const { blocks } = sectionsToNotionBlocks(doc.sections, { extractReferences: true });
         await this.appendRemainingBlocks(notionPageId, blocks);
     }
     async clearPageContent(pageId) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@moxn/kb-migrate",
-  "version": "0.4.12",
+  "version": "0.4.13",
   "description": "Migration tool for importing documents into Moxn Knowledge Base from local files, Notion, Google Docs, and more",
   "type": "module",
   "main": "dist/index.js",