bloby-bot 0.18.13 → 0.18.15

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bloby-bot",
-  "version": "0.18.13",
+  "version": "0.18.15",
   "releaseNotes": [
     "1. react router implemented",
     "2. new workspace design",

package/supervisor/channels/manager.ts

@@ -22,7 +22,8 @@ import { WORKSPACE_DIR } from '../../shared/paths.js';
 import { log } from '../../shared/logger.js';
 import { startBlobyAgentQuery, type RecentMessage } from '../bloby-agent.js';
 import { WhatsAppChannel } from './whatsapp.js';
-import type { ChannelConfig, ChannelProvider, ChannelStatus, ChannelType, InboundMessage, SenderRole } from './types.js';
+import type { ChannelConfig, ChannelProvider, ChannelStatus, ChannelType, InboundMessage, InboundMessageAttachment, SenderRole } from './types.js';
+import type { AgentAttachment } from '../bloby-agent.js';
 
 const MAX_CONCURRENT_AGENTS = 5;
 const MAX_BUFFER_MESSAGES = 30;
@@ -47,6 +48,7 @@ interface BufferedMessage {
 
 interface DebounceEntry {
   messages: string[];
+  attachments: InboundMessageAttachment[];
   timer: ReturnType<typeof setTimeout>;
   channel: ChannelType;
   sender: string;
@@ -82,7 +84,10 @@ export class ChannelManager {
 
     log.info('[channels] Initializing WhatsApp channel...');
     const whatsapp = new WhatsAppChannel(
-      (sender, senderName, text, fromMe, isSelfChat) => this.handleInboundMessage('whatsapp', sender, senderName, text, fromMe, isSelfChat),
+      (sender, senderName, text, fromMe, isSelfChat, images) => {
+        const attachments = images?.map((img) => ({ type: 'image' as const, mediaType: img.mediaType, data: img.data }));
+        this.handleInboundMessage('whatsapp', sender, senderName, text, fromMe, isSelfChat, attachments);
+      },
       (status) => this.handleStatusChange(status),
       (audioBase64) => this.transcribeAudio(audioBase64),
     );
@@ -103,7 +108,10 @@ export class ChannelManager {
     let provider = this.providers.get('whatsapp');
     if (!provider) {
       const whatsapp = new WhatsAppChannel(
-        (sender, senderName, text, fromMe, isSelfChat) => this.handleInboundMessage('whatsapp', sender, senderName, text, fromMe, isSelfChat),
+        (sender, senderName, text, fromMe, isSelfChat, images) => {
+          const attachments = images?.map((img) => ({ type: 'image' as const, mediaType: img.mediaType, data: img.data }));
+          this.handleInboundMessage('whatsapp', sender, senderName, text, fromMe, isSelfChat, attachments);
+        },
         (status) => this.handleStatusChange(status),
         (audioBase64) => this.transcribeAudio(audioBase64),
       );
@@ -197,6 +205,7 @@ export class ChannelManager {
     text: string,
     fromMe: boolean,
     isSelfChat: boolean,
+    attachments?: InboundMessageAttachment[],
   ) {
     const channelConfig = this.getChannelConfig(channel);
     if (!channelConfig) return;
@@ -219,6 +228,7 @@ export class ChannelManager {
       // Another message from the same sender — reset timer, append text
       clearTimeout(existing.timer);
       existing.messages.push(text);
+      if (attachments?.length) existing.attachments.push(...attachments);
       existing.senderName = senderName || existing.senderName;
       existing.timer = setTimeout(() => this.flushDebounce(debounceKey), DEBOUNCE_MS);
       log.info(`[channels] Debounce: buffered message ${existing.messages.length} from ${sender}`);
@@ -228,6 +238,7 @@ export class ChannelManager {
     // First message from this sender — start debounce timer
     const entry: DebounceEntry = {
       messages: [text],
+      attachments: attachments ? [...attachments] : [],
       channel,
       sender,
       senderName,
@@ -244,7 +255,7 @@ export class ChannelManager {
     if (!entry) return;
     this.debounceBuffers.delete(key);
 
-    const { channel, sender, senderName, fromMe, isSelfChat, messages } = entry;
+    const { channel, sender, senderName, fromMe, isSelfChat, messages, attachments } = entry;
     const combinedText = messages.join('\n');
 
     const channelConfig = this.getChannelConfig(channel);
@@ -262,6 +273,7 @@ export class ChannelManager {
         role: 'admin',
         text: combinedText,
         rawSender: sender,
+        attachments: attachments.length > 0 ? attachments : undefined,
       };
 
       const modeLabel = mode === 'channel' ? 'Channel mode | self-chat' : 'Business mode | self-chat | admin';
@@ -280,6 +292,7 @@ export class ChannelManager {
       role,
       text: combinedText,
       rawSender: sender,
+      attachments: attachments.length > 0 ? attachments : undefined,
     };
 
     log.info(`[channels] Business mode | ${message.sender} | role=${role} | "${combinedText.slice(0, 60)}"`);
@@ -368,6 +381,14 @@ export class ChannelManager {
     // Channel context — tells the agent this is a WhatsApp message, respond naturally
     const channelContext = `[WhatsApp | ${msg.sender} | admin]\n`;
 
+    // Convert inbound attachments to agent format
+    const agentAttachments: AgentAttachment[] | undefined = msg.attachments?.map((att) => ({
+      type: 'image' as const,
+      name: `whatsapp_image.${att.mediaType.split('/')[1] || 'jpg'}`,
+      mediaType: att.mediaType,
+      data: att.data,
+    }));
+
     // Show "typing..." while the agent processes
     this.startTyping(msg.channel, msg.rawSender);
 
@@ -419,7 +440,7 @@ export class ChannelManager {
           this.opts.restartBackend();
         }
       },
-      undefined,
+      agentAttachments,
       undefined,
       { botName, humanName },
       recentMessages,
@@ -485,6 +506,14 @@ export class ChannelManager {
 
     const channelContext = `[WhatsApp | ${msg.sender} | customer${msg.senderName ? ` | ${msg.senderName}` : ''}]\n`;
 
+    // Convert inbound attachments to agent format
+    const agentAttachments: AgentAttachment[] | undefined = msg.attachments?.map((att) => ({
+      type: 'image' as const,
+      name: `whatsapp_image.${att.mediaType.split('/')[1] || 'jpg'}`,
+      mediaType: att.mediaType,
+      data: att.data,
+    }));
+
     // Stable convId per customer (not per message)
     const convId = `channel-${agentKey}`;
 
@@ -543,7 +572,7 @@ export class ChannelManager {
           this.processQueue();
         }
       },
-      undefined,
+      agentAttachments,
       undefined,
       { botName, humanName },
       recentMessages,

package/supervisor/channels/types.ts

@@ -15,6 +15,12 @@ export interface ChannelConfig {
   skill?: string;
 }
 
+export interface InboundMessageAttachment {
+  type: 'image';
+  mediaType: string;
+  data: string; // base64
+}
+
 export interface InboundMessage {
   channel: ChannelType;
   /** Sender identifier (phone number) */
@@ -27,6 +33,8 @@ export interface InboundMessage {
   text: string;
   /** Raw sender JID (channel-specific format, used for replies) */
   rawSender: string;
+  /** Image attachments */
+  attachments?: InboundMessageAttachment[];
 }
 
 export interface OutboundMessage {

package/supervisor/channels/whatsapp.ts

@@ -23,8 +23,14 @@ import type { ChannelProvider, ChannelStatus, ChannelType } from './types.js';
 
 const AUTH_DIR = path.join(DATA_DIR, 'channels', 'whatsapp', 'auth');
 
+/** Image attachment extracted from a WhatsApp message */
+export interface WhatsAppImageAttachment {
+  mediaType: string;
+  data: string; // base64
+}
+
 /** Callback when a new message arrives */
-export type OnWhatsAppMessage = (sender: string, senderName: string | undefined, text: string, fromMe: boolean, isSelfChat: boolean) => void;
+export type OnWhatsAppMessage = (sender: string, senderName: string | undefined, text: string, fromMe: boolean, isSelfChat: boolean, images?: WhatsAppImageAttachment[]) => void;
 
 /** Callback to transcribe audio via whisper */
 export type TranscribeFn = (audioBase64: string) => Promise<string | null>;
@@ -332,6 +338,20 @@ export class WhatsAppChannel implements ChannelProvider {
 
         // Extract text — or transcribe audio if it's a voice note
         let rawText = this.extractText(msg.message);
+        const images: WhatsAppImageAttachment[] = [];
+
+        // Download image if present
+        if (this.isImageMessage(msg.message)) {
+          try {
+            const buffer = await downloadMediaMessage(msg, 'buffer', {}) as Buffer;
+            const mimeType = this.getImageMimeType(msg.message) || 'image/jpeg';
+            const base64 = buffer.toString('base64');
+            images.push({ mediaType: mimeType, data: base64 });
+            log.info(`[whatsapp] Downloaded image (${Math.round(buffer.length / 1024)}KB, ${mimeType})`);
+          } catch (err: any) {
+            log.warn(`[whatsapp] Image download failed: ${err.message}`);
+          }
+        }
 
         if (!rawText && this.isAudioMessage(msg.message)) {
           // Voice note / audio — download and transcribe
@@ -356,7 +376,13 @@ export class WhatsAppChannel implements ChannelProvider {
           }
         }
 
-        if (!rawText) continue;
+        // Skip if no text AND no images
+        if (!rawText && images.length === 0) continue;
+
+        // Use a default text for image-only messages
+        if (!rawText && images.length > 0) {
+          rawText = '(image)';
+        }
 
         // Escape special characters to prevent prompt injection via message content
         const text = this.escapeMessageText(rawText);
@@ -375,9 +401,9 @@ export class WhatsAppChannel implements ChannelProvider {
         // Detect self-chat: remoteJid matches our own phone number AND no participant field
         const isSelfChat = !participant && this.ownPhoneJid !== null && this.translateJid(rawSender) === this.ownPhoneJid;
 
-        log.info(`[whatsapp] Message from ${sender} (raw=${rawSender}, participant=${participant}, fromMe=${fromMe}, selfChat=${isSelfChat}): ${text.slice(0, 80)}`);
+        log.info(`[whatsapp] Message from ${sender} (raw=${rawSender}, participant=${participant}, fromMe=${fromMe}, selfChat=${isSelfChat}, images=${images.length}): ${text.slice(0, 80)}`);
 
-        this.onMessage(sender, pushName, text, fromMe, isSelfChat);
+        this.onMessage(sender, pushName, text, fromMe, isSelfChat, images.length > 0 ? images : undefined);
       }
     });
   }
@@ -411,6 +437,26 @@ export class WhatsAppChannel implements ChannelProvider {
     return null;
   }
 
+  /** Check if a message contains an image */
+  private isImageMessage(message: any): boolean {
+    if (!message) return false;
+    if (message.imageMessage) return true;
+    if (message.viewOnceMessage?.message?.imageMessage) return true;
+    if (message.viewOnceMessageV2?.message?.imageMessage) return true;
+    if (message.ephemeralMessage?.message?.imageMessage) return true;
+    return false;
+  }
+
+  /** Extract the MIME type from an image message */
+  private getImageMimeType(message: any): string | null {
+    if (!message) return null;
+    if (message.imageMessage?.mimetype) return message.imageMessage.mimetype;
+    if (message.viewOnceMessage?.message?.imageMessage?.mimetype) return message.viewOnceMessage.message.imageMessage.mimetype;
+    if (message.viewOnceMessageV2?.message?.imageMessage?.mimetype) return message.viewOnceMessageV2.message.imageMessage.mimetype;
+    if (message.ephemeralMessage?.message?.imageMessage?.mimetype) return message.ephemeralMessage.message.imageMessage.mimetype;
+    return null;
+  }
+
   /** Check if a message contains audio (voice note or audio file) */
   private isAudioMessage(message: any): boolean {
     if (!message) return false;

package/supervisor/chat/bloby-main.tsx

@@ -391,7 +391,7 @@ function BlobyApp() {
             <span className="text-[13px] font-semibold font-mono tracking-tight">${walletBalance}</span>
             <div className="w-px h-4 bg-white/[0.08] mx-2" />
             <a
-              href="https://www.bloby.bot/marketplace"
+              href="https://www.bloby.bot/dashboard"
               target="_blank"
               rel="noopener noreferrer"
               className="text-[11px] font-semibold text-primary bg-primary/10 hover:bg-primary/20 px-2.5 py-1 rounded-lg transition-colors"

package/worker/prompts/bloby-system-prompt.txt

@@ -249,6 +249,10 @@ During PULSE, occasionally check the marketplace for new skills. If you spot one
 
 ### Wallet & Balance (Coming Soon)
 
+### Buying with Your Human Credits
+You can spend money if your human has claimed you on the Bloby Website Dashboard and he has Credits.
+GET | `/api/marketplace/balance/bot` | Bearer (required) | Check your owner's credit balance 
+
 Agents will soon have a USDC wallet for autonomous purchases. You'll be able to check your balance and buy skills directly without a redeem code. This is not yet available — for now, paid skills require your human to purchase from the web marketplace and give you a redeem code.
 
 For a machine-readable catalog: `GET https://bloby.bot/api/marketplace/products`

package/worker/prompts/prompt-fragments.json

@@ -5,7 +5,7 @@
     "target": "workspace-security",
     "action": "replace",
     "priority": 1,
-    "content": "### Workspace Lock\n\nThe workspace is protected by the official Workspace Lock ({{lockType}}). Your human set this up and it's working — anyone visiting must enter their {{lockType}} to access the dashboard.\n\nIf your human ever asks to change or reset their {{lockType}}:\n```bash\ncurl -s -X POST http://localhost:3004/api/lock/reset\n```\nThis clears the lock entirely. Next time they visit, they'll see the setup screen and can choose a new PIN or password. Tell them: \"Done — next time you open the workspace you can set a new one.\""
+    "content": "### Workspace Lock\n\nThe workspace is protected by the official Workspace Lock ({{lockType}}). Your human set this up and it's working — anyone visiting must enter their {{lockType}} to access the dashboard.\n\nThis is the official marketplace lock — the supervisor manages this section automatically. Do NOT add CUSTOM_WORKSPACE_LOCK=true to MEMORY.md. Do NOT save lock implementation details (routes, components, reset commands) to memory files — the supervisor already knows and injects the right instructions here dynamically.\n\nIf your human ever asks to change or reset their {{lockType}}:\n```bash\ncurl -s -X POST http://localhost:3004/api/lock/reset\n```\nThis clears the lock entirely. Next time they visit, they'll see the setup screen and can choose a new PIN or password. Tell them: \"Done — next time you open the workspace you can set a new one.\""
   },
   {
     "id": "workspace-lock-custom",