feat(tier4): Add patch generation for auto-patching

- Add TIER4_PATCH_PROMPT to mma_prompts.py with unified diff format
- Add run_tier4_patch_generation function to ai_client.py
- Import mma_prompts in ai_client.py
- Add unit tests for patch generation

src/ai_client.py

@@ -28,6 +28,7 @@ from pathlib import Path
 from src import project_manager
 from src import file_cache
 from src import mcp_client
+from src import mma_prompts
 import anthropic
 from src.gemini_cli_adapter import GeminiCliAdapter as GeminiCliAdapter
 from google import genai
@@ -1960,6 +1961,33 @@ def run_tier4_analysis(stderr: str) -> str:
  except Exception as e:
   return f"[QA ANALYSIS FAILED] {e}"
 
+
+def run_tier4_patch_generation(error: str, file_context: str) -> str:
+ if not error or not error.strip():
+  return ""
+ try:
+  _ensure_gemini_client()
+  if not _gemini_client:
+   return ""
+  prompt = (
+   f"{mma_prompts.TIER4_PATCH_PROMPT}\n\n"
+   f"Error:\n```\n{error}\n```\n\n"
+   f"File Context:\n```\n{file_context}\n```\n"
+  )
+  model_name = "gemini-2.5-flash-lite"
+  resp = _gemini_client.models.generate_content(
+   model=model_name,
+   contents=prompt,
+   config=types.GenerateContentConfig(
+    temperature=0.0,
+    max_output_tokens=2048,
+   )
+  )
+  patch = resp.text.strip() if resp.text else ""
+  return patch
+ except Exception as e:
+  return f"[PATCH GENERATION FAILED] {e}"
+
 def get_token_stats(md_content: str) -> dict[str, Any]:
  global _provider, _gemini_client, _model, _CHARS_PER_TOKEN
  total_tokens = 0