Address remaining CodeRabbit review feedback for streaming and rendering.

- Use distinct 3-tuple ('complete', text, messages) for completion events
  to avoid ambiguity with ('tool_use', [...]) 2-tuples in chat streaming.
- Pass conversation history from request into chat_with_database_stream()
  so follow-up NLQ turns retain context.
- Add re.IGNORECASE to SQL fence regex for case-insensitive matching.
- Render MarkdownContent as block element instead of span to avoid
  invalid DOM when response contains paragraphs, lists, or tables.
- Keep stop notice as a separate message instead of appending to partial
  markdown, preventing it from being swallowed by open code fences.
- Snapshot streamingIdRef before setMessages in error handler to avoid
  race condition where ref is cleared before React executes the updater.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: dpage

web/pgadmin/llm/chat.py

@@ -174,8 +174,8 @@ def chat_with_database_stream(
     Yields:
         str: Text content chunks from the final LLM response.
 
-    The last item yielded is a tuple of
-    (final_response_text, updated_conversation_history).
+    The last item yielded is a 3-tuple of
+    ('complete', final_response_text, updated_conversation_history).
 
     Raises:
         LLMClientError: If the LLM request fails.
@@ -220,8 +220,9 @@ def chat_with_database_stream(
         messages.append(response.to_message())
 
         if response.stop_reason != StopReason.TOOL_USE:
-            # Final response - yield the completion tuple
-            yield (response.content, messages)
+            # Final response - yield a 3-tuple to distinguish from
+            # the 2-tuple tool_use event
+            yield ('complete', response.content, messages)
             return
 
         # Signal that tools are being executed so the caller can

web/pgadmin/tools/sqleditor/__init__.py

@@ -2875,6 +2875,7 @@ def nlq_chat_stream(trans_id):
     data = request.get_json(silent=True) or {}
     user_message = data.get('message', '').strip()
     conversation_id = data.get('conversation_id')
+    history_data = data.get('history', [])
 
     if not user_message:
         return make_json_response(
@@ -2885,6 +2886,10 @@ def nlq_chat_stream(trans_id):
     def generate():
         """Generator for SSE events."""
         import secrets as py_secrets
+        from pgadmin.llm.compaction import (
+            deserialize_history, compact_history
+        )
+        from pgadmin.llm.utils import get_default_provider
 
         try:
             # Send thinking status
@@ -2893,13 +2898,24 @@ def generate():
                 'message': gettext('Analyzing your request...')
             })
 
+            # Deserialize and compact conversation history
+            conversation_history = None
+            if history_data:
+                conversation_history = deserialize_history(history_data)
+                provider = get_default_provider() or 'openai'
+                conversation_history = compact_history(
+                    conversation_history,
+                    provider=provider
+                )
+
             # Stream the LLM response with database tools
             response_text = ''
             for item in chat_with_database_stream(
                 user_message=user_message,
                 sid=trans_obj.sid,
                 did=trans_obj.did,
-                system_prompt=NLQ_SYSTEM_PROMPT
+                system_prompt=NLQ_SYSTEM_PROMPT,
+                conversation_history=conversation_history
             ):
                 if isinstance(item, str):
                     # Text chunk from streaming LLM response
@@ -2916,15 +2932,16 @@ def generate():
                             'Querying the database...'
                         )
                     })
-                elif isinstance(item, tuple):
-                    # Final result: (response_text, messages)
-                    response_text = item[0]
+                elif isinstance(item, tuple) and \
+                        item[0] == 'complete':
+                    # Final result: ('complete', response_text, messages)
+                    response_text = item[1]
 
             # Extract SQL from markdown code fences
             sql_blocks = re.findall(
                 r'```(?:sql|pgsql|postgresql)\s*\n(.*?)```',
                 response_text,
-                re.DOTALL
+                re.DOTALL | re.IGNORECASE
             )
             sql = ';\n\n'.join(
                 block.strip().rstrip(';') for block in sql_blocks

web/pgadmin/tools/sqleditor/static/js/components/sections/NLQChatPanel.jsx

@@ -550,9 +550,8 @@ function ChatMessage({ message, onInsertSQL, onReplaceSQL, textColors, cmKey, fo
           const content = seg.content?.trim();
           if (!content && !cursor) return null;
           return (
-            <Box key={idx} sx={{ marginTop: idx > 0 ? 1 : 0, display: 'inline' }}>
+            <Box key={idx} sx={{ marginTop: idx > 0 ? 1 : 0 }}>
               <MarkdownContent
-                component="span"
                 dangerouslySetInnerHTML={{ __html: renderMarkdownText(content || '') }}
               />
               {cursor}
@@ -882,13 +881,18 @@ export function NLQChatPanel() {
       // Check if user manually stopped
       if (stoppedRef.current) {
         const streamId = streamingIdRef.current;
-        // If we have partial streaming content, show it as-is
+        // If we have partial streaming content, show it separately
+        // from the stop notice to avoid breaking open markdown fences
         if (streamingTextRef.current) {
           setMessages((prev) => [
             ...prev.filter((m) => m.id !== thinkingId && m.id !== streamId),
             {
               type: MESSAGE_TYPES.ASSISTANT,
-              content: streamingTextRef.current + '\n\n' + gettext('(Generation stopped)'),
+              content: streamingTextRef.current,
+            },
+            {
+              type: MESSAGE_TYPES.ASSISTANT,
+              content: gettext('Generation stopped.'),
             },
           ]);
         } else {
@@ -912,13 +916,17 @@ export function NLQChatPanel() {
       if (error.name === 'AbortError') {
         // Check if this was a user-initiated stop or a timeout
         if (stoppedRef.current) {
-          // User manually stopped - show partial content if any
+          // User manually stopped - show partial content separately
           if (streamingTextRef.current) {
             setMessages((prev) => [
               ...prev.filter((m) => m.id !== thinkingId && m.id !== streamId),
               {
                 type: MESSAGE_TYPES.ASSISTANT,
-                content: streamingTextRef.current + '\n\n' + gettext('(Generation stopped)'),
+                content: streamingTextRef.current,
+              },
+              {
+                type: MESSAGE_TYPES.ASSISTANT,
+                content: gettext('Generation stopped.'),
               },
             ]);
           } else {
@@ -1043,9 +1051,10 @@ export function NLQChatPanel() {
       break;
     }
 
-    case 'error':
+    case 'error': {
+      const streamId = streamingIdRef.current;
       setMessages((prev) => [
-        ...prev.filter((m) => m.id !== thinkingId && m.id !== streamingIdRef.current),
+        ...prev.filter((m) => m.id !== thinkingId && m.id !== streamId),
         {
           type: MESSAGE_TYPES.ERROR,
           content: event.message,

web/pgadmin/tools/sqleditor/tests/test_nlq_chat.py

@@ -98,7 +98,7 @@ def runTest(self):
         if hasattr(self, 'mock_response'):
             def mock_stream_gen(*args, **kwargs):
                 yield self.mock_response
-                yield (self.mock_response, [])
+                yield ('complete', self.mock_response, [])
             mock_chat = patch(
                 'pgadmin.llm.chat.chat_with_database_stream',
                 side_effect=mock_stream_gen
@@ -254,7 +254,7 @@ def runTest(self):
         sql_blocks = re.findall(
             r'```(?:sql|pgsql|postgresql)\s*\n(.*?)```',
             response_text,
-            re.DOTALL
+            re.DOTALL | re.IGNORECASE
         )
         sql = ';\n\n'.join(
             block.strip() for block in sql_blocks
@@ -328,8 +328,8 @@ def mock_stream_gen(*args, **kwargs):
             for chunk in [self.mock_response[i:i + 10]
                           for i in range(0, len(self.mock_response), 10)]:
                 yield chunk
-            # Yield final tuple
-            yield (self.mock_response, [])
+            # Yield final 3-tuple
+            yield ('complete', self.mock_response, [])
 
         mock_chat = patch(
             'pgadmin.llm.chat.chat_with_database_stream',