Upload tool

tool.py

@@ -1,8 +1,8 @@
 from smolagents.tools import Tool
+import string
 import json
-import pronouncing
 import difflib
-import string
+import pronouncing
 
 class ParodyWordSuggestionTool(Tool):
     name = "parody_word_suggester"
@@ -98,7 +98,24 @@ class ParodyWordSuggestionTool(Tool):
         target_phones = pronouncing.phones_for_word(target)
         if not target_phones:
             return json.dumps({
-                "error": f"'{target}' not found in CMU dictionary",
+                "error": f"Target word '{target}' not found in CMU dictionary",
+                "suggestions": []
+            }, indent=2)
+    
+        # Filter word list to only words in CMU dictionary
+        valid_words = []
+        invalid_words = []
+        for word in words:
+            word = word.lower().strip(string.punctuation)
+            if pronouncing.phones_for_word(word):
+                valid_words.append(word)
+            else:
+                invalid_words.append(word)
+    
+        if not valid_words:
+            return json.dumps({
+                "error": "No valid words found in CMU dictionary",
+                "invalid_words": invalid_words,
                 "suggestions": []
             }, indent=2)
     
@@ -107,8 +124,7 @@ class ParodyWordSuggestionTool(Tool):
         target_vowel, target_end = self._get_last_syllable(target_phone_list)
     
         # Check each word
-        for word in words:
-            word = word.lower().strip(string.punctuation)
+        for word in valid_words:
             phones = pronouncing.phones_for_word(word)
             if phones:
                 word_phones = phones[0]
@@ -126,8 +142,8 @@ class ParodyWordSuggestionTool(Tool):
                     
                         if word_end_clean == target_end_clean:
                             rhyme_score = 1.0
-                            # Extra boost for exact match
-                            if len(word) == len(target):
+                            # Extra boost for exact match minus first letter
+                            if len(word) == len(target) and word[1:] == target[1:]:
                                 rhyme_score = 1.2
                         else:
                             rhyme_score = 0.6
@@ -137,10 +153,15 @@ class ParodyWordSuggestionTool(Tool):
                 word_syl = pronouncing.syllable_count(word_phones)
                 syllable_score = 1.0 if target_syl == word_syl else 0.0
             
-                # 3. Overall similarity (15%)
-                string_similarity = SequenceMatcher(None, target, word).ratio()
+                # 3. String similarity (15%)
+                # Higher weight for end of word similarity
+                if len(word) > 1 and len(target) > 1:
+                    end_similarity = SequenceMatcher(None, word[1:], target[1:]).ratio()
+                    string_similarity = end_similarity
+                else:
+                    string_similarity = SequenceMatcher(None, target, word).ratio()
             
-                # Combined score with phonetic similarity bonus
+                # Combined score
                 similarity = (rhyme_score * 0.6) + (syllable_score * 0.25) + (string_similarity * 0.15)
             
                 if similarity >= min_similarity:
@@ -171,6 +192,7 @@ class ParodyWordSuggestionTool(Tool):
             "target_phones": target_phones,
             "target_last_vowel": target_vowel,
             "target_ending": " ".join(target_end) if target_end else "",
+            "invalid_words": invalid_words,  # List of words not in CMU
             "suggestions": suggestions
         }