@satiyap/confluence-reader-mcp 0.2.1 → 0.2.3

package/README.md

@@ -76,12 +76,13 @@ Lists the direct child pages of a Confluence page without fetching their content
 
 ### `confluence.fetch_image`
 
-Downloads an image attachment from a Confluence page by filename. Returns the image as base64-encoded data.
+Downloads an image attachment from a Confluence page by filename and saves it to a local directory.
 
 | Parameter | Type | Description |
 |-----------|------|-------------|
 | `url` | string | Confluence page URL |
 | `filename` | string | Attachment filename (e.g. `architecture.png`) |
+| `destination` | string | Local directory path to save the image to |
 
 ### `confluence.compare`
 

package/dist/confluence/transform.js

@@ -14,29 +14,40 @@ turndown.use(gfm);
  */
 function normalizeConfluenceHtml(html) {
     let out = html;
-    // Convert ac:layout-section / ac:layout-cell to divs
-    out = out.replace(/<ac:layout-section>/gi, "<div>");
+    // --- Confluence layout tags → divs ---
+    out = out.replace(/<ac:layout-section[^>]*>/gi, "<div>");
     out = out.replace(/<\/ac:layout-section>/gi, "</div>");
     out = out.replace(/<ac:layout-cell>/gi, "<div>");
     out = out.replace(/<\/ac:layout-cell>/gi, "</div>");
     out = out.replace(/<ac:layout>/gi, "<div>");
     out = out.replace(/<\/ac:layout>/gi, "</div>");
-    // Convert ac:structured-macro (panels, code blocks, etc.) to divs
-    // Preserve the macro name as a data attribute for potential future use
+    // --- Table cleanup: strip attributes and colgroup so Turndown can parse ---
+    out = out.replace(/<table[^>]*>/gi, "<table>");
+    out = out.replace(/<colgroup>[\s\S]*?<\/colgroup>/gi, "");
+    out = out.replace(/<col[^>]*\/?>/gi, "");
+    out = out.replace(/<div class="content-wrapper">/gi, "");
+    // (closing </div> for content-wrapper will be handled by generic div cleanup later)
+    // --- Jira macro → text reference (extract key) ---
+    out = out.replace(/<ac:structured-macro[^>]*ac:name="jira"[^>]*>([\s\S]*?)<\/ac:structured-macro>/gi, (_match, inner) => {
+        const keyMatch = inner.match(/<ac:parameter[^>]*ac:name="key"[^>]*>([\s\S]*?)<\/ac:parameter>/i);
+        return keyMatch ? `<code>${keyMatch[1].trim()}</code>` : "";
+    });
+    // --- TOC and other self-closing macros → remove ---
+    out = out.replace(/<ac:structured-macro[^>]*\/>/gi, "");
+    // --- Code blocks ---
     out = out.replace(/<ac:structured-macro[^>]*ac:name="code"[^>]*>([\s\S]*?)<\/ac:structured-macro>/gi, (_match, inner) => {
-        // Extract plain-text-body for code blocks
         const bodyMatch = inner.match(/<ac:plain-text-body>\s*<!\[CDATA\[([\s\S]*?)\]\]>\s*<\/ac:plain-text-body>/i);
         if (bodyMatch) {
             return `<pre><code>${bodyMatch[1]}</code></pre>`;
         }
         return `<pre><code>${inner.replace(/<[^>]+>/g, "")}</code></pre>`;
     });
-    // Convert info/note/warning/tip panels to blockquotes
+    // --- Info/note/warning/tip panels → blockquotes ---
     out = out.replace(/<ac:structured-macro[^>]*ac:name="(info|note|warning|tip|panel)"[^>]*>([\s\S]*?)<\/ac:structured-macro>/gi, (_match, _type, inner) => {
         const bodyMatch = inner.match(/<ac:rich-text-body>([\s\S]*?)<\/ac:rich-text-body>/i);
         return bodyMatch ? `<blockquote>${bodyMatch[1]}</blockquote>` : `<blockquote>${inner}</blockquote>`;
     });
-    // Generic: any remaining ac:structured-macro — unwrap to div
+    // --- Generic remaining ac:structured-macro → unwrap to div ---
     out = out.replace(/<ac:structured-macro[^>]*>/gi, "<div>");
     out = out.replace(/<\/ac:structured-macro>/gi, "</div>");
     // ac:rich-text-body → div
@@ -48,14 +59,24 @@ function normalizeConfluenceHtml(html) {
     out = out.replace(/<\/ac:plain-text-body>/gi, "</pre>");
     // ac:parameter tags — remove entirely
     out = out.replace(/<ac:parameter[^>]*>[\s\S]*?<\/ac:parameter>/gi, "");
-    // ac:image → img tag
+    // --- ac:image → img tag ---
     out = out.replace(/<ac:image[^>]*>([\s\S]*?)<\/ac:image>/gi, (_match, inner) => {
         const filenameMatch = inner.match(/ri:filename="([^"]+)"/i);
         const filename = filenameMatch ? filenameMatch[1] : "image";
         return `<img alt="${filename}" src="${filename}" />`;
     });
-    // ac:link with ri:page → anchor
+    // --- ac:link: handle attachment links, user mentions, and page links ---
     out = out.replace(/<ac:link>([\s\S]*?)<\/ac:link>/gi, (_match, inner) => {
+        // Attachment link
+        const attachMatch = inner.match(/ri:filename="([^"]+)"/i);
+        if (attachMatch) {
+            return `<a href="#">📎 ${attachMatch[1]}</a>`;
+        }
+        // User mention
+        if (/<ri:user/i.test(inner)) {
+            return `<code>@user</code>`;
+        }
+        // Page link
         const pageMatch = inner.match(/ri:content-title="([^"]+)"/i);
         const bodyMatch = inner.match(/<ac:link-body>([\s\S]*?)<\/ac:link-body>/i)
             || inner.match(/<ac:plain-text-link-body>\s*<!\[CDATA\[([\s\S]*?)\]\]>\s*<\/ac:plain-text-link-body>/i);
@@ -65,7 +86,7 @@ function normalizeConfluenceHtml(html) {
     });
     // ac:emoticon → remove
     out = out.replace(/<ac:emoticon[^>]*\/>/gi, "");
-    // ac:task-list / ac:task / ac:task-body → ul/li
+    // --- ac:task-list / ac:task → ul/li ---
     out = out.replace(/<ac:task-list>/gi, "<ul>");
     out = out.replace(/<\/ac:task-list>/gi, "</ul>");
     out = out.replace(/<ac:task>([\s\S]*?)<\/ac:task>/gi, (_match, inner) => {
@@ -75,11 +96,62 @@ function normalizeConfluenceHtml(html) {
         const body = bodyMatch ? bodyMatch[1] : inner;
         return `<li>${checked ? "[x] " : "[ ] "}${body}</li>`;
     });
-    // Remove any remaining ac:* or ri:* tags but keep their text content
+    // --- Cleanup: remove any remaining ac:*/ri:* tags, keep text content ---
     out = out.replace(/<\/?(?:ac|ri):[^>]*>/gi, "");
     // Clean up CDATA remnants
     out = out.replace(/<!\[CDATA\[/g, "");
     out = out.replace(/\]\]>/g, "");
+    // Strip attributes from th/td so Turndown sees clean cells
+    out = out.replace(/<th[^>]*>/gi, "<th>");
+    out = out.replace(/<td[^>]*>/gi, "<td>");
+    // --- Normalize table rows to uniform column count ---
+    // Turndown GFM requires every row to have the same number of cells.
+    // Confluence tables often have irregular column spans.
+    out = out.replace(/<table>([\s\S]*?)<\/table>/gi, (_match, tableInner) => {
+        // Count cells per row
+        const rows = tableInner.match(/<tr>[\s\S]*?<\/tr>/gi) ?? [];
+        const cellCounts = rows.map((row) => {
+            const cells = row.match(/<(?:th|td)>/gi);
+            return cells ? cells.length : 0;
+        });
+        const maxCols = Math.max(0, ...cellCounts);
+        if (maxCols === 0)
+            return _match;
+        // Pad short rows
+        const paddedRows = rows.map((row, i) => {
+            const deficit = maxCols - cellCounts[i];
+            if (deficit <= 0)
+                return row;
+            const pad = "<td></td>".repeat(deficit);
+            return row.replace(/<\/tr>/i, `${pad}</tr>`);
+        });
+        // Ensure first row uses <th> so Turndown generates a header row
+        const rebuilt = tableInner.replace(/<tr>[\s\S]*?<\/tr>/gi, () => paddedRows.shift());
+        return `<table>${rebuilt}</table>`;
+    });
+    // Clean up table internals so Turndown GFM can convert them
+    out = out.replace(/<table>([\s\S]*?)<\/table>/gi, (_match, inner) => {
+        let cleaned = inner;
+        // Strip block-level wrappers inside cells
+        cleaned = cleaned.replace(/<\/?p>/gi, "");
+        cleaned = cleaned.replace(/<\/?span[^>]*>/gi, "");
+        cleaned = cleaned.replace(/<\/?div[^>]*>/gi, "");
+        // Convert <br> to space (markdown tables can't have line breaks)
+        cleaned = cleaned.replace(/<br\s*\/?>/gi, " ");
+        // Strip <tbody> wrapper — Turndown wants <table><thead><tr>…</tr></thead>…
+        cleaned = cleaned.replace(/<\/?tbody>/gi, "");
+        // Ensure first row uses <th> so Turndown sees a header
+        let firstDone = false;
+        cleaned = cleaned.replace(/<tr>([\s\S]*?)<\/tr>/gi, (trMatch, trInner) => {
+            if (!firstDone) {
+                firstDone = true;
+                const promoted = trInner.replace(/<td>/gi, "<th>").replace(/<\/td>/gi, "</th>");
+                return `<thead><tr>${promoted}</tr></thead>`;
+            }
+            return `<tr>${trInner}</tr>`;
+        });
+        return `<table>${cleaned}</table>`;
+    });
     return out;
 }
 /**

package/dist/index.js

@@ -8,7 +8,7 @@ import { storageToMarkdown } from "./confluence/transform.js";
 import { generateUnifiedDiff, generateDiffStats } from "./compare/diff.js";
 const server = new McpServer({
     name: "confluence-reader-mcp",
-    version: "0.2.1"
+    version: "0.2.3"
 });
 function getEnv(name) {
     const v = process.env[name];
@@ -79,10 +79,11 @@ server.tool("confluence.list_children", "List the direct child pages of a Conflu
         : "No child pages found.";
     return { content: [{ type: "text", text }] };
 });
-server.tool("confluence.fetch_image", "Download an image attachment from a Confluence page by filename. Returns the image as base64-encoded data.", {
+server.tool("confluence.fetch_image", "Download an image attachment from a Confluence page by filename and save it to a local directory. Returns the saved file path.", {
     url: z.string().describe("Confluence page URL"),
-    filename: z.string().describe("Attachment filename (e.g. 'architecture.png')")
-}, async ({ url, filename }) => {
+    filename: z.string().describe("Attachment filename (e.g. 'architecture.png')"),
+    destination: z.string().describe("Local directory path to save the image to")
+}, async ({ url, filename, destination }) => {
     const cfg = getCfg();
     const pageId = extractConfluencePageId(url);
     const attachments = await fetchAttachments(cfg, pageId);
@@ -96,23 +97,17 @@ server.tool("confluence.fetch_image", "Download an image attachment from a Confl
                 }]
         };
     }
-    const { buffer, contentType } = await downloadAttachment(cfg, pageId, match.id);
-    const base64 = buffer.toString("base64");
-    // Return as base64 image content
-    if (contentType.startsWith("image/")) {
-        return {
-            content: [{
-                    type: "image",
-                    data: base64,
-                    mimeType: contentType,
-                }]
-        };
-    }
-    // Non-image attachment — return as base64 text
+    const { buffer } = await downloadAttachment(cfg, pageId, match.id);
+    // Ensure destination directory exists
+    const fs = await import("node:fs/promises");
+    const path = await import("node:path");
+    await fs.mkdir(destination, { recursive: true });
+    const filePath = path.join(destination, match.title);
+    await fs.writeFile(filePath, buffer);
     return {
         content: [{
                 type: "text",
-                text: `Downloaded "${filename}" (${contentType}, ${buffer.length} bytes).\nBase64: ${base64.slice(0, 200)}...`
+                text: `Saved "${match.title}" (${buffer.length} bytes) to ${filePath}`
             }]
     };
 });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@satiyap/confluence-reader-mcp",
-  "version": "0.2.1",
+  "version": "0.2.3",
   "description": "MCP server for fetching and comparing Confluence documentation with local files",
   "author": "satiyap",
   "license": "MIT",