why aren't you working

web2json/ai_extractor.py

@@ -224,7 +224,7 @@ class NvidiaLLMClient(LLMClient):
         """
         self.model_name = model_name
 
-    @retry_on_ratelimit(max_retries=6, base_delay=0.5, max_delay=5.0)
+    @retry_on_ratelimit(max_retries=20, base_delay=0.5, max_delay=5.0)
     def call_api(self, prompt: str) -> str:
         """
         Call the NVIDIA API with the given prompt (non-streaming).
@@ -241,7 +241,8 @@ class NvidiaLLMClient(LLMClient):
             messages=[{"role": "user", "content": prompt}],
             temperature=self.temperature,
             top_p=self.top_p,
-            max_tokens=self.max_tokens
+            max_tokens=self.max_tokens,
+            extra_body={"chat_template_kwargs": {"thinking":True}},
             # stream is omitted (defaults to False)
         )
         # print("DONE")
@@ -262,6 +263,7 @@ class NvidiaLLMClient(LLMClient):
                 idx = futures[fut]
                 try:
                     results[idx] = fut.result()
+                    print("DONE")
                 except RateLimitError:
                     # You could set results[idx] = None or a default string
                     results[idx] = f"<failed after retries>"
@@ -367,6 +369,7 @@ class ModalRerankerClient(RerankerClient):
         scores = np.array([p["score"] for p in ranked], dtype=float)
         # Max normalization
         max_score = scores.max() if len(scores) > 0 else 1.0
+        # max_score = 1
         if max_score == 0:
             norm_scores = np.zeros_like(scores)
         else:
@@ -381,6 +384,7 @@ class ModalRerankerClient(RerankerClient):
             for p, norm in filtered
         ]
         
+        # docs.reverse()
 
         return docs