bloby-bot 0.36.0 → 0.37.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bloby-bot",
-  "version": "0.36.0",
+  "version": "0.37.0",
   "releaseNotes": [
     "1. # voice note (PTT bubble)",
     "2. # audio file + caption",

package/supervisor/channels/manager.ts

@@ -27,11 +27,25 @@ import { startBlobyAgentQuery, startConversation, pushMessage, hasConversation,
 import { WhatsAppChannel } from './whatsapp.js';
 import type { ChannelConfig, ChannelProvider, ChannelStatus, ChannelType, InboundMessage, InboundMessageAttachment, SenderRole } from './types.js';
 import type { AgentAttachment } from '../bloby-agent.js';
+import { saveAttachment, type SavedFile } from '../file-saver.js';
 
 const MAX_CONCURRENT_AGENTS = 5;
 const MAX_BUFFER_MESSAGES = 30;
 const DEBOUNCE_MS = 4000; // 4s — wait for the user to finish typing
 
+/** Persist channel-inbound attachments to disk so harnesses that consume file
+ *  paths (Codex's `localImage`) can see them. Failures are logged and the
+ *  attachment is dropped — text-only delivery is still useful. */
+function saveInboundAttachments(attachments?: AgentAttachment[]): SavedFile[] {
+  if (!attachments?.length) return [];
+  const saved: SavedFile[] = [];
+  for (const att of attachments) {
+    try { saved.push(saveAttachment(att)); }
+    catch (err: any) { log.warn(`[channels] Failed to save inbound attachment: ${err.message}`); }
+  }
+  return saved;
+}
+
 interface ChannelManagerOpts {
   broadcastBloby: (type: string, data: any) => void;
   workerApi: (path: string, method?: string, body?: any) => Promise<any>;
@@ -120,9 +134,9 @@ export class ChannelManager {
     let provider = this.providers.get('whatsapp');
     if (!provider) {
       const whatsapp = new WhatsAppChannel(
-        (sender, senderName, text, fromMe, isSelfChat, images) => {
+        (sender, senderName, text, fromMe, isSelfChat, chatJid, isGroup, 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);
+          this.handleInboundMessage('whatsapp', sender, senderName, text, fromMe, isSelfChat, chatJid, isGroup, attachments);
         },
         (status) => this.handleStatusChange(status),
         (audioBase64) => this.transcribeAudio(audioBase64),
@@ -693,6 +707,10 @@ export class ChannelManager {
       mediaType: att.mediaType,
       data: att.data,
     }));
+    // Save to disk so providers that consume file paths (Codex → localImage)
+    // can see the attachment. Claude consumes raw base64 from `agentAttachments`
+    // directly, but the on-disk copy is still useful for the path mention.
+    const savedFiles = saveInboundAttachments(agentAttachments);
 
     // Show "typing..." in the correct chat
     this.startTyping(msg.channel, msg.rawSender);
@@ -740,7 +758,7 @@ export class ChannelManager {
 
     // Push the message into the live conversation
     const channelContent = channelContext + msg.text;
-    pushMessage(convId, channelContent, agentAttachments);
+    pushMessage(convId, channelContent, agentAttachments, savedFiles);
   }
 
   /** Handle message from a customer — runs support agent in parallel with conversation context */
@@ -809,6 +827,7 @@ export class ChannelManager {
       mediaType: att.mediaType,
       data: att.data,
     }));
+    const savedFiles = saveInboundAttachments(agentAttachments);
 
     // Stable convId per customer (not per message)
     const convId = `channel-${agentKey}`;
@@ -869,7 +888,7 @@ export class ChannelManager {
         }
       },
       agentAttachments,
-      undefined,
+      savedFiles,
       { botName, humanName },
       recentMessages,
       enrichedScript,