Upload app.py

app.py

@@ -266,48 +266,58 @@ class PodcastGenerator:
             base_url="https://api.sambanova.ai/v1",
         )
 
-        # 嘗試生成內容
-        try:
-            # Calculate the available tokens for generation
-            prompt_tokens = len(user_prompt.split())  # This is a rough estimate
-            system_tokens = len(system_prompt.split())  # This is a rough estimate
-            max_tokens = 4096 - prompt_tokens - system_tokens - 100  # 100 is a safety margin
-
-            if max_tokens <= 0:
-                return {"error": "Input prompt is too long. Please provide a shorter prompt."}
-
-            logger.info(f"Sending request to SambaNova API with prompt: {user_prompt[:100]}...")
-            response = client.chat.completions.create(
-                model='Meta-Llama-3.1-405B-Instruct',
-                messages=[
-                    {"role": "system", "content": system_prompt},
-                    {"role": "user", "content": user_prompt}
-                ],
-                temperature=1,
-                max_tokens=max_tokens
-            )
-            logger.info(f"Received response from API: {response}")
-            
-            if hasattr(response, 'error'):
-                logger.error(f"API returned an error: {response.error}")
-                return {"error": f"API error: {response.error.get('message', 'Unknown error')}"}
-            
-            if response.choices and len(response.choices) > 0:
-                generated_text = response.choices[0].message.content
-                logger.info(f"Generated text: {generated_text[:100]}...")
-            else:
-                logger.warning("No content generated from the API")
-                return {"error": "No content generated from the API"}
-
-        except Exception as e:
-            logger.error(f"Error generating script: {str(e)}")
-            # 處理可能的錯誤
-            if "API key not valid" in str(e):
-                raise gr.Error("Invalid API key. Please provide a valid SambaNova API key.")
-            elif "rate limit" in str(e).lower():
-                raise gr.Error("Rate limit exceeded for the API key. Please try again later or provide your own SambaNova API key.")
-            else:
-                raise gr.Error(f"Failed to generate podcast script: {str(e)}")
+        async def generate_chunk(chunk: str) -> str:
+            try:
+                # Calculate the available tokens for generation
+                prompt_tokens = len(chunk.split())  # This is a rough estimate
+                system_tokens = len(system_prompt.split())  # This is a rough estimate
+                max_tokens = 4096 - prompt_tokens - system_tokens - 100  # 100 is a safety margin
+
+                if max_tokens <= 0:
+                    return {"error": "Input chunk is too long. Please provide a shorter prompt."}
+
+                logger.info(f"Sending request to SambaNova API with prompt chunk: {chunk[:100]}...")
+                response = client.chat.completions.create(
+                    model='Meta-Llama-3.1-405B-Instruct',
+                    messages=[
+                        {"role": "system", "content": system_prompt},
+                        {"role": "user", "content": chunk}
+                    ],
+                    temperature=1,
+                    max_tokens=max_tokens
+                )
+                logger.info(f"Received response from API: {response}")
+                
+                if hasattr(response, 'error'):
+                    logger.error(f"API returned an error: {response.error}")
+                    return {"error": f"API error: {response.error.get('message', 'Unknown error')}"}
+                
+                if response.choices and len(response.choices) > 0:
+                    generated_text = response.choices[0].message.content
+                    logger.info(f"Generated text: {generated_text[:100]}...")
+                    return generated_text
+                else:
+                    logger.warning("No content generated from the API")
+                    return {"error": "No content generated from the API"}
+
+            except Exception as e:
+                logger.error(f"Error generating script chunk: {str(e)}")
+                return {"error": f"Failed to generate podcast script chunk: {str(e)}"}
+
+        # Split the prompt into chunks
+        chunk_size = 1000  # Adjust this value as needed
+        chunks = [prompt[i:i+chunk_size] for i in range(0, len(prompt), chunk_size)]
+
+        # Generate script for each chunk
+        generated_chunks = []
+        for chunk in chunks:
+            result = await generate_chunk(chunk)
+            if isinstance(result, dict) and "error" in result:
+                return result
+            generated_chunks.append(result)
+
+        # Combine generated chunks
+        generated_text = " ".join(generated_chunks)
 
         # 嘗試解析JSON，如果失敗則嘗試從原始文本中提取對話
         try:
@@ -520,12 +530,6 @@ async def process_input(input_text: str, input_file, language: str, speaker1: st
             gr.Error(f"Selected voices may not be compatible with the chosen language: {language}")
             return None
 
-    # Check input text length
-    max_input_length = 1000  # Adjust this value as needed
-    if len(input_text) > max_input_length:
-        gr.Error(f"Input text is too long. Please limit your input to {max_input_length} characters.")
-        return None
-
     # 如果提供了輸入檔案，則從檔案中提取文字
     if input_file:
         input_text = await TextExtractor.extract_text(input_file.name)