Add logging for context window configuration and improve max_tokens calculation debugging

src/llm_router.py

@@ -54,7 +54,8 @@ class LLMRouter:
             logger.info("Novita AI API client initialized")
             logger.info(f"Base URL: {self.settings.novita_base_url}")
             logger.info(f"Model: {self.settings.novita_model}")
-        except Exception as e:
+            logger.info(f"Context Window: {self.settings.novita_model_context_window} tokens")
+            except Exception as e:
             logger.error(f"Failed to initialize Novita AI client: {e}")
             raise RuntimeError(f"Could not initialize Novita AI API client: {e}") from e
     
@@ -149,7 +150,7 @@ class LLMRouter:
                 response_text = self._clean_reasoning_tags(response_text)
                 logger.info(f"Novita AI API generated response (length: {len(response_text)})")
                 return response_text
-            else:
+                    else:
                 # Handle non-streaming response
                 response = self.novita_client.chat.completions.create(**request_params)
                 
@@ -159,14 +160,14 @@ class LLMRouter:
                     result = self._clean_reasoning_tags(result)
                     logger.info(f"Novita AI API generated response (length: {len(result)})")
                     return result
-                else:
+                    else:
                     logger.error("Novita AI API returned empty response")
                     return None
                 
         except Exception as e:
             logger.error(f"Error calling Novita AI API: {e}", exc_info=True)
-            raise
-    
+                        raise
+            
     def _calculate_safe_max_tokens(self, prompt: str, requested_max_tokens: int) -> int:
         """
         Calculate safe max_tokens based on input token count and model context window.
@@ -182,9 +183,14 @@ class LLMRouter:
         # For more accuracy, you could use tiktoken if available
         input_tokens = len(prompt) // 4
         
-        # Get model context window
+        # Get model context window from settings
         context_window = self.settings.novita_model_context_window
         
+        logger.debug(
+            f"Calculating safe max_tokens: input ~{input_tokens} tokens, "
+            f"context_window={context_window}, requested={requested_max_tokens}"
+        )
+        
         # Reserve minimum 100 tokens for safety margin
         available_tokens = context_window - input_tokens - 100
         
@@ -197,7 +203,8 @@ class LLMRouter:
         if safe_max_tokens < requested_max_tokens:
             logger.warning(
                 f"Reduced max_tokens from {requested_max_tokens} to {safe_max_tokens} "
-                f"(input: ~{input_tokens} tokens, context window: {context_window} tokens)"
+                f"(input: ~{input_tokens} tokens, context window: {context_window} tokens, "
+                f"available: {available_tokens} tokens)"
             )
         
         return safe_max_tokens
@@ -375,7 +382,7 @@ class LLMRouter:
     def _truncate_to_tokens(self, content: str, max_tokens: int) -> str:
         """Truncate content to fit within token limit"""
         # Simple character-based truncation (1 token ≈ 4 chars)
-        max_chars = max_tokens * 4
-        if len(content) <= max_chars:
-            return content
+            max_chars = max_tokens * 4
+            if len(content) <= max_chars:
+                return content
         return content[:max_chars - 3] + "..."