Add validation and logging for custom prompt size

Introduces a check to raise a ValueError if a custom moderation prompt is too large to allow sufficient space for content and output. Adds detailed logging for prompt and content token counts to aid debugging and prevent exceeding model context window limits.

Author: Bentlybro

app/services/ai/ai_moderator.py

@@ -43,6 +43,9 @@ def calculate_max_content_tokens(self, custom_prompt=None):
         Calculate the maximum tokens available for content based on prompt size.
         Dynamically adjusts for custom prompts to prevent exceeding context window.
         Uses conservative estimates to account for tokenizer inaccuracies.
+
+        Raises:
+            ValueError: If custom prompt is too large to leave any space for content
         """
         # Calculate custom prompt tokens if provided
         prompt_tokens = 0
@@ -62,6 +65,22 @@ def calculate_max_content_tokens(self, custom_prompt=None):
             # Count tokens for the prompt parts (excluding content placeholder)
             prompt_tokens = self.count_tokens(system_message) + self.count_tokens(user_template)
 
+            # Check if prompt alone exceeds safe limits
+            # Reserve at least 20k tokens for content + output + safety margin
+            MIN_CONTENT_SPACE = 20000
+            max_allowed_prompt = self.model_context_window - self.max_output_tokens - MIN_CONTENT_SPACE - 1000
+
+            if prompt_tokens > max_allowed_prompt:
+                current_app.logger.error(
+                    f"Custom prompt is too large: {prompt_tokens} tokens exceeds maximum of {max_allowed_prompt}. "
+                    f"This rule cannot be processed."
+                )
+                raise ValueError(
+                    f"Custom moderation rule is too large ({prompt_tokens} tokens). "
+                    f"Maximum allowed is {max_allowed_prompt} tokens. "
+                    f"Please reduce the size of your custom rule."
+                )
+
             # Log large prompts for debugging
             if prompt_tokens > 10000:
                 current_app.logger.warning(
@@ -415,6 +434,19 @@ def _analyze_with_custom_prompt(self, content, custom_prompt):
 
 Does content violate this rule? JSON only:"""
 
+            # DEBUG: Log exact token counts before API call
+            system_tokens = self.count_tokens(system_message)
+            user_tokens = self.count_tokens(user_message)
+            total_tokens = system_tokens + user_tokens
+            current_app.logger.info(
+                f"[DEBUG] About to call OpenAI API - "
+                f"System: {system_tokens} tokens, "
+                f"User: {user_tokens} tokens, "
+                f"Total: {total_tokens} tokens, "
+                f"Custom prompt length: {len(custom_prompt)} chars, "
+                f"Content length: {len(content)} chars"
+            )
+
             # Wrap API call with retry logic
             def make_api_call():
                 client = self.client_manager.get_client()