fix: add diagnostic logging and increase scaffold max_tokens to 2048

The "Unterminated string" JSON parse error is likely caused by Gemini
output truncation at 1024 tokens. Increases scaffold max_tokens to 2048
and adds logging for: raw response text, finish_reason (truncation
detection), and JSON parse failures.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

backend/app/core/ai_provider.py

@@ -5,10 +5,13 @@ Supports Gemini (google-genai) and Anthropic (anthropic) as interchangeable
 backends for JSON generation used by the AI Flow Builder.
 """
 
+import logging
 from abc import ABC, abstractmethod
 
 from app.core.config import settings
 
+logger = logging.getLogger(__name__)
+
 
 class AIProvider(ABC):
     """Abstract base class for AI providers."""
@@ -74,6 +77,16 @@ class GeminiProvider(AIProvider):
             config=config,
         )
 
+        # Log finish reason to detect truncation
+        if response.candidates:
+            finish_reason = getattr(response.candidates[0], "finish_reason", None)
+            logger.info("Gemini finish_reason=%s model=%s", finish_reason, self._model)
+            if str(finish_reason) == "MAX_TOKENS":
+                logger.warning(
+                    "Gemini output truncated (MAX_TOKENS). max_output_tokens=%d",
+                    max_tokens,
+                )
+
         text = response.text or ""
         input_tokens = getattr(response.usage_metadata, "prompt_token_count", 0) or 0
         output_tokens = (