Update conver.py

conver.py

@@ -1,5 +1,5 @@
 from dataclasses import dataclass
-from typing import List, Tuple, Dict, Optional
+from typing import List, Tuple, Dict
 import os
 import json
 import httpx
@@ -14,7 +14,7 @@ from pathlib import Path
 class ConversationConfig:
     max_words: int = 3000
     prefix_url: str = "https://r.jina.ai/"
-    model_name: str = "meta-llama/Llama-3-8b-chat-hf"
+    model_name: str = "meta-llama/Llama-3-8b-chat-hf"  # Modelo serverless
 
 class URLToAudioConverter:
     def __init__(self, config: ConversationConfig, llm_api_key: str):
@@ -34,23 +34,39 @@ class URLToAudioConverter:
             raise RuntimeError(f"Failed to fetch URL: {e}")
 
     def extract_conversation(self, text: str) -> Dict:
+        """Versión que parsea 'Host1: texto' -> JSON"""
         if not text:
             raise ValueError("Input text cannot be empty")
+        
+        prompt = (
+            f"{text}\nCreate a podcast dialogue between Host1 and Host2. "
+            "Use EXACTLY this format:\n\n"
+            "Host1: [message]\nHost2: [reply]\nHost1: [response]..."
+        )
+        
         try:
-            prompt = (
-                f"{text}\nConvert this text into a podcast conversation between two hosts. "
-                "Return ONLY JSON with this structure:\n"
-                '{"conversation": [{"speaker": "Host1", "text": "..."}, {"speaker": "Host2", "text": "..."}]}'
-            )
             response = self.llm_client.chat.completions.create(
                 messages=[{"role": "user", "content": prompt}],
                 model=self.config.model_name,
-                response_format={"type": "json_object"}
+                temperature=0.7
             )
-            json_str = response.choices[0].message.content.strip()
-            return json.loads(json_str[json_str.find('{'):json_str.rfind('}')+1])
+            raw_dialogue = response.choices[0].message.content
+            
+            # Parseo seguro del formato
+            conversation = {"conversation": []}
+            for line in raw_dialogue.split('\n'):
+                if ':' in line:
+                    speaker, _, content = line.partition(':')
+                    if speaker.strip() in ("Host1", "Host2"):
+                        conversation["conversation"].append({
+                            "speaker": speaker.strip(),
+                            "text": content.strip()
+                        })
+            
+            return conversation
+            
         except Exception as e:
-            raise RuntimeError(f"Failed to extract conversation: {str(e)}")
+            raise RuntimeError(f"Failed to parse dialogue: {str(e)}")
 
     async def text_to_speech(self, conversation_json: Dict, voice_1: str, voice_2: str) -> Tuple[List[str], str]:
         output_dir = Path(self._create_output_directory())
@@ -100,7 +116,7 @@ class URLToAudioConverter:
     ) -> AudioSegment:
         music = AudioSegment.from_file(music_path).fade_out(2000) - 25
         if len(music) < len(speech_audio):
-            music = music * ((len(speech_audio) // len(music)) + 1)
+            music = music * ((len(speech_audio) // len(music) + 1)
         music = music[:len(speech_audio)]
         mixed = speech_audio.overlay(music)
         
@@ -108,6 +124,7 @@ class URLToAudioConverter:
         tag_trans = AudioSegment.from_file(tags_paths[1]) - 10
         final_audio = tag_intro + mixed
         
+        # Insertar tags en silencios >500ms
         silent_ranges = []
         for i in range(0, len(speech_audio) - 500, 100):
             chunk = speech_audio[i:i+500]
@@ -132,7 +149,7 @@ class URLToAudioConverter:
         return await self._process_to_audio(conversation, voice_1, voice_2)
 
     async def raw_text_to_audio(self, text: str, voice_1: str, voice_2: str) -> Tuple[str, str]:
-        conversation = {"conversation": [{"speaker": "Narrator", "text": text}]}
+        conversation = {"conversation": [{"speaker": "Host1", "text": text}]}
         return await self._process_to_audio(conversation, voice_1, voice_2)
 
     async def _process_to_audio(
@@ -148,7 +165,7 @@ class URLToAudioConverter:
             "musica.mp3",
             ["tag.mp3", "tag2.mp3"]
         )
-        output_path = os.path.join(folder_name, "output.mp3")
+        output_path = os.path.join(folder_name, "podcast_final.mp3")
         final_audio.export(output_path, format="mp3")
         
         for f in audio_files: