fix(copilot): Address PR review comments

- Shield transcript upload and session save from generator cancellation
  (asyncio.shield) so page refresh/disconnect doesn't lose the transcript
- Return content_base64 for small image files (not just text) so
  _extract_image_block can actually work
- Add 32KB size limit to _extract_image_block to prevent oversized images
- Fix gap fill when transcript_msg_count == 0 (metadata absent)
- Add truncation to files.join in GenericTool.tsx

autogpt_platform/backend/backend/copilot/sdk/service.py

@@ -576,7 +576,7 @@ async def stream_chat_completion_sdk(
                 query_message = current_message
                 current_msg_count = len(session.messages)
 
-                if use_resume and transcript_msg_count > 0:
+                if use_resume and transcript_msg_count >= 0:
                     # Transcript covers messages[0..M-1].  Current session
                     # has N messages (last one is the new user msg).
                     # Gap = messages[M .. N-2] (everything between upload
@@ -707,11 +707,17 @@ async def stream_chat_completion_sdk(
                     raw_transcript = None
 
                 if raw_transcript:
-                    await _try_upload_transcript(
-                        user_id,
-                        session_id,
-                        raw_transcript,
-                        message_count=len(session.messages),
+                    # Shield the upload from generator cancellation so a
+                    # client disconnect / page refresh doesn't lose the
+                    # transcript.  The upload must finish even if the SSE
+                    # connection is torn down.
+                    await asyncio.shield(
+                        _try_upload_transcript(
+                            user_id,
+                            session_id,
+                            raw_transcript,
+                            message_count=len(session.messages),
+                        )
                     )
 
         except ImportError:
@@ -721,7 +727,7 @@ async def stream_chat_completion_sdk(
                 "to use the OpenAI-compatible fallback."
             )
 
-        await upsert_chat_session(session)
+        await asyncio.shield(upsert_chat_session(session))
         logger.debug(
             f"[SDK] Session {session_id} saved with {len(session.messages)} messages"
         )
@@ -731,7 +737,7 @@ async def stream_chat_completion_sdk(
     except Exception as e:
         logger.error(f"[SDK] Error: {e}", exc_info=True)
         try:
-            await upsert_chat_session(session)
+            await asyncio.shield(upsert_chat_session(session))
         except Exception as save_err:
             logger.error(f"[SDK] Failed to save session on error: {save_err}")
         yield StreamError(

autogpt_platform/backend/backend/copilot/sdk/tool_adapter.py

@@ -190,7 +190,14 @@ def _extract_image_block(text: str) -> dict[str, str] | None:
     mime_type = data.get("mime_type", "")
     base64_content = data.get("content_base64", "")
 
-    if mime_type in _SUPPORTED_IMAGE_TYPES and base64_content:
+    # Only inline small images — large ones would exceed Claude's limits.
+    # 32 KB raw ≈ ~43 KB base64.
+    _MAX_IMAGE_BASE64_BYTES = 43_000
+    if (
+        mime_type in _SUPPORTED_IMAGE_TYPES
+        and base64_content
+        and len(base64_content) <= _MAX_IMAGE_BASE64_BYTES
+    ):
         return {
             "type": "image",
             "data": base64_content,

autogpt_platform/backend/backend/copilot/tools/workspace_files.py

@@ -312,8 +312,18 @@ class ReadWorkspaceFileTool(BaseTool):
             is_small_file = file_info.size_bytes <= self.MAX_INLINE_SIZE_BYTES
             is_text_file = self._is_text_mime_type(file_info.mime_type)
 
-            # Return inline content for small text files (unless force_download_url)
-            if is_small_file and is_text_file and not force_download_url:
+            # Return inline content for small text/image files (unless force_download_url)
+            is_image_file = file_info.mime_type in {
+                "image/png",
+                "image/jpeg",
+                "image/gif",
+                "image/webp",
+            }
+            if (
+                is_small_file
+                and (is_text_file or is_image_file)
+                and not force_download_url
+            ):
                 content = await manager.read_file_by_id(target_file_id)
                 content_b64 = base64.b64encode(content).decode("utf-8")
 

autogpt_platform/frontend/src/app/(platform)/copilot/tools/GenericTool/GenericTool.tsx

@@ -501,7 +501,7 @@ function getFileAccordionData(
           <ContentCodeBlock>{truncate(content, 2000)}</ContentCodeBlock>
         )}
         {files && files.length > 0 && (
-          <ContentCodeBlock>{files.join("\n")}</ContentCodeBlock>
+          <ContentCodeBlock>{truncate(files.join("\n"), 2000)}</ContentCodeBlock>
         )}
         {!content && !files && message && (
           <ContentMessage>{message}</ContentMessage>