@omnisocials/mcp-server 1.35.0 → 1.46.0

package/build/tools/media.js

@@ -1,7 +1,7 @@
 import { z } from "zod";
 import { fetchImageAsBase64, formatBytes } from "../client.js";
 export function registerMediaTools(server, getClient) {
-    server.tool("list_media", "List media files in the workspace. Returns each file's ID, name, folder, type (image/video), and size. To reuse an existing graphic, SEARCH for it by name here FIRST instead of re-uploading — re-uploading the same image creates duplicates. Media IDs from this list can be passed to create_post. Organize assets into folders with create_folder / list_folders.", {
+    server.tool("list_media", "List media files in the workspace. Returns each file's ID, name, folder, type (image/video/document — a document is a PDF kept as one item; its ID in media_ids expands into every page), and size. To reuse an existing graphic, SEARCH for it by name here FIRST instead of re-uploading — re-uploading the same image creates duplicates. Media IDs from this list can be passed to create_post. Organize assets into folders with create_folder / list_folders.", {
         limit: z.string().optional().describe("Max results to return"),
         offset: z.string().optional().describe("Offset for pagination"),
         search: z.string().optional().describe('Find a file by name or filename, e.g. "play5get50". Use this before uploading to check if the graphic already exists.'),
@@ -30,7 +30,16 @@ export function registerMediaTools(server, getClient) {
             // `type` may be "video"/"image" (normalized) or a raw MIME like "video/mp4".
             const rawType = typeof m.type === "string" ? m.type : "";
             const isVideo = rawType === "video" || rawType.startsWith("video/");
-            const type = isVideo ? "Video" : "Image";
+            const isDocument = rawType === "document";
+            // A PDF document item names its page count (pass its ONE id and the
+            // post gets every page); a slide rendered from a PDF names its file
+            // and page, so the assistant can pass the whole deck in order.
+            const pdfTag = m.pdf
+                ? isDocument
+                    ? ` (PDF, ${Array.isArray(m.pdf.pages) ? m.pdf.pages.length : m.pdf.rendered_pages} pages)`
+                    : ` (PDF ${m.pdf.name ? `"${m.pdf.name}" ` : ""}page ${m.pdf.page}/${m.pdf.total_pages})`
+                : "";
+            const type = (isVideo ? "Video" : isDocument ? "Document" : "Image") + pdfTag;
             const folder = m.folder_id ? `\`${m.folder_id}\`` : "—";
             md += `| ${i + 1} | ${type} | ${m.name || m.filename || "—"} | ${folder} | ${(m.size || "—")} | \`${m.id}\` |\n`;
         }
@@ -57,7 +66,7 @@ export function registerMediaTools(server, getClient) {
     });
     server.tool("upload_media", `Upload media to the library. Accepts EITHER a public URL OR base64-encoded image data (e.g. when the user pastes an image directly in chat). Supported: JPEG, PNG, GIF, WebP, MP4, MOV, AVI, and PDF.
 
-PDF = carousel: upload a PDF (as a 'url' or as base64_data with mime_type "application/pdf") and it is split into one image slide per page (max 20 pages). The response lists a Media ID for every slide — pass ALL of them, in order, as media_ids to create_post to post the deck as a carousel. On LinkedIn the slides post as a native swipeable DOCUMENT; on Instagram, TikTok, Threads and Pinterest as an image carousel. This is the way to let a user post an existing slide deck (Canva/PowerPoint/Figma exported to PDF) as a carousel.
+PDF = carousel: upload a PDF (as a 'url' or as base64_data with mime_type "application/pdf") and it is split into one image slide per page (max 20 pages). The response lists a Media ID for every slide — pass ALL of them, in order, as media_ids to create_post to post the deck as a carousel. Prefer pdf_mode "document" when the user wants ONE library item for the deck: the response then has a single Media ID whose media_ids entry expands into every page at post time. On LinkedIn the slides post as a native swipeable DOCUMENT made from the ORIGINAL PDF file (the file is kept: text stays sharp, in-document links work, viewers download the real file, and every page is included even past the 20-slide cap) as long as the slides are posted unchanged and in order; on Instagram, TikTok, Threads and Pinterest as an image carousel. Set linkedin.document_source to 'slides' on the post to send a document rebuilt from the slide images instead. This is the way to let a user post an existing slide deck (Canva/PowerPoint/Figma exported to PDF) as a carousel.
 
 Size limits: base64 and direct uploads are capped at 100 MB (and base64 inflates the request body, so the practical ceiling is ~75 MB). For anything larger — up to 1 GB — use ONE of: (a) pass a public 'url' and the server fetches it (handles up to 1 GB), or (b) have the user upload the file in the OmniSocials Library UI at https://app.omnisocials.com/library . IMPORTANT: if the user has a large LOCAL file (over ~100 MB) with no public URL, do NOT tell them to compress or shrink it — instead point them to the Library UI link above, which uploads files up to 1 GB directly.
 
@@ -72,6 +81,7 @@ When the user provides an image in the conversation (not a URL), use base64_data
         filename: z.string().optional().describe("Optional filename for the uploaded media"),
         name: z.string().optional().describe('Human-readable label so you can find this asset by name later instead of re-uploading it, e.g. "pp-play5get50". Strongly recommended on every upload.'),
         folder: z.string().optional().describe('Optional folder name to file this asset under (created at the top level if it does not exist), e.g. "win-graphics". See list_folders.'),
+        pdf_mode: z.enum(["slides", "document"]).optional().describe('PDF uploads only. "slides" (default): one image media item per page, one Media ID each (pass ALL of media_ids to create_post). "document": ONE media item for the whole PDF (type "document"); pass its single Media ID in media_ids and the post gets every page in order. Both keep the original file, which LinkedIn receives. Prefer "document" when the user wants one library item for the deck.'),
     }, { title: "Upload Media", readOnlyHint: false, destructiveHint: false, openWorldHint: true }, async (params) => {
         let result;
         if (params.base64_data && params.mime_type) {
@@ -81,10 +91,17 @@ When the user provides an image in the conversation (not a URL), use base64_data
                 filename: params.filename,
                 name: params.name || params.filename,
                 folder: params.folder,
+                ...(params.pdf_mode ? { pdf_mode: params.pdf_mode } : {}),
             });
         }
         else if (params.url) {
-            result = await getClient().uploadMedia({ url: params.url, filename: params.filename, name: params.name || params.filename, folder: params.folder });
+            result = await getClient().uploadMedia({
+                url: params.url,
+                filename: params.filename,
+                name: params.name || params.filename,
+                folder: params.folder,
+                ...(params.pdf_mode ? { pdf_mode: params.pdf_mode } : {}),
+            });
         }
         else {
             return {
@@ -97,6 +114,23 @@ When the user provides an image in the conversation (not a URL), use base64_data
             };
         }
         const compatibility = result.compatibility;
+        // PDF kept as ONE document item (pdf_mode "document").
+        if (result.data?.type === "document" && result.pdf) {
+            const d = result.data;
+            const pdfInfo = result.pdf;
+            const pageCount = Array.isArray(d.pdf?.pages) ? d.pdf.pages.length : pdfInfo.rendered_pages;
+            let md = `## PDF Uploaded as One Document\n\n`;
+            md += `**${d.name || pdfInfo.name || "document.pdf"}**, ${pageCount} ${pageCount === 1 ? "page" : "pages"}`;
+            if (pdfInfo.truncated)
+                md += ` (${pdfInfo.total_pages} in the file; the first ${pdfInfo.rendered_pages} became page images)`;
+            md += `.\n\n**Media ID:** \`${d.id}\`\n\n`;
+            md += `Pass this ONE id in \`create_post\` as \`media_ids: ["${d.id}"]\`; the post gets every page, in order. `;
+            md += `On LinkedIn it posts as a swipeable **document** made from the **original PDF file** (sharp text, working links, the real download); on Instagram, TikTok, Threads and Pinterest as an **image** carousel. Platforms that do not support carousels use the first page.`;
+            if (compatibility && compatibility.compatible === false && compatibility.summary) {
+                md += `\n\n⚠️ **${compatibility.summary}** It will still post to your other connected platforms. Ask the user whether to continue before posting.`;
+            }
+            return { content: [{ type: "text", text: md }] };
+        }
         // PDF upload: the server split it into image slides. Surface every slide
         // id and tell the model to pass them ALL to create_post as one carousel.
         const mediaIds = result.media_ids;
@@ -115,7 +149,7 @@ When the user provides an image in the conversation (not a URL), use base64_data
             });
             md += `\n**Pass ALL of these to \`create_post\` as \`media_ids\`, in this order**, to post the deck as a carousel:\n\n`;
             md += `\`\`\`\nmedia_ids: [${mediaIds.map((id) => `"${id}"`).join(", ")}]\n\`\`\`\n\n`;
-            md += `On LinkedIn this posts as a swipeable **document** carousel; on Instagram, TikTok, Threads and Pinterest as an **image** carousel. Platforms that do not support carousels use the first slide.`;
+            md += `On LinkedIn this posts as a swipeable **document** carousel made from the **original PDF file** (kept in storage: sharp text, working links, the real download${pdf?.truncated ? `, all ${pdf.total_pages} pages` : ""}) as long as the slides are posted unchanged and in order; on Instagram, TikTok, Threads and Pinterest as an **image** carousel. Platforms that do not support carousels use the first slide. To send a document rebuilt from the slide images instead, set linkedin.document_source to "slides" on the post.`;
             if (compatibility && compatibility.compatible === false && compatibility.summary) {
                 md += `\n\n⚠️ **${compatibility.summary}** It will still post to your other connected platforms. Ask the user whether to continue before posting.`;
             }
@@ -154,7 +188,7 @@ When the user provides an image in the conversation (not a URL), use base64_data
     });
     server.tool("check_media_compatibility", `Check whether a video/image will be accepted by the platforms CONNECTED to this workspace, BEFORE uploading or posting. Use this when the user gives you a file (URL) so you can warn them upfront — e.g. "this 995 MB video won't post to Instagram (max 300 MB), continue?" — and confirm before uploading.
 
-Provide ONE of: a public 'url' (its size/type is read via a HEAD request), an existing 'media_id', or explicit 'size_bytes' + 'mime'. Returns the connected platforms and any that would reject the file by size or format.`, {
+Provide ONE of: a public 'url' (its size/type is read via a HEAD request), an existing 'media_id', or explicit 'size_bytes' + 'mime'. Returns the connected platforms and any that would reject the file by size or format, plus IMAGE FIT rows for an image whose size is known (a library file): per connected channel, whether its feed shows the image uncropped and what a post does by default (Instagram crops a 9:16 / wide image to 4:5 / 1.91:1 around the subject; Pinterest keeps a tall Pin and reports it). When a fit row says fits: false, ask the user "crop to <target> (recommended for Instagram: Meta can boost it) or keep the original (the channel cuts it itself)?" and pass the answer as media_fit on create_post.`, {
         url: z.string().optional().describe("Public URL of the file to check. Use this OR media_id OR size_bytes+mime."),
         media_id: z.string().optional().describe("Id of an already-uploaded library item to check."),
         size_bytes: z.number().optional().describe("File size in bytes (use with mime when you already know them)."),
@@ -191,6 +225,19 @@ Provide ONE of: a public 'url' (its size/type is read via a HEAD request), an ex
             }
             md += `\nAsk the user whether to continue. It will still post to the platforms not listed above.`;
         }
+        const outside = (r.fit || []).filter((f) => !f.fits);
+        if (outside.length) {
+            const dims = r.file?.width && r.file?.height ? ` (${r.file.width}x${r.file.height})` : "";
+            md += `\n\n**Image fit**${dims}: how these channels show it in the feed\n\n| Platform | Ratio | In the feed | A post does |\n|----------|-------|-------------|-------------|\n`;
+            for (const f of outside) {
+                const does = f.default_action === "crop" ? `crops it to ${f.target} (media_fit "keep" sends the original)` : `keeps it and reports it (media_fit "crop" crops to ${f.target})`;
+                md += `| ${f.display_name} | ${f.ratio} | cut to ${f.target} | ${does} |\n`;
+            }
+            md += `\nAsk the user per channel above: crop to the target (Instagram: Meta can then boost the post) or keep the original (the channel cuts it itself)? Pass the answer as \`media_fit\` on create_post. Stories, reels and the other channels are not affected.`;
+        }
+        else if ((r.fit || []).length) {
+            md += `\n\nImage fit: this image shows uncropped in every connected feed with a rule (${(r.fit || []).map((f) => f.display_name).join(", ")}).`;
+        }
         return { content: [{ type: "text", text: md }] };
     });
     server.tool("update_media", "Rename an existing media file or move it into a folder (so it's findable later instead of re-uploading it). Only the fields you pass are changed.", {