commit code

scheduler/job_story_portal.py

@@ -0,0 +1,80 @@
+import json
+from utils.logger import logger
+import datetime
+import os
+import asyncio
+from models.script import Script
+from config.database import SessionLocal
+from llm.generate_daily_article import generate_daily_article
+
+project_name = "故事任意门"
+
+# for daily article generation
+def job_generate_daily_article():
+    """定时任务：生成每日文章并保存至数据库。"""
+
+    # 1. 调用 LLM 生成每日文章
+    content = generate_daily_article()
+    if not content:
+        logger.warning("No daily article generated.")
+        return
+    
+    # 2. 保存至数据库
+    # subject 以当前日期为准，格式 YYYY-MM-DD
+    today_str = datetime.datetime.now().strftime("%Y-%m-%d")
+    article_title= content["阶段4_今日文章"]["文章标题"]
+    db = SessionLocal()
+    try:
+        # 查询是否已存在 project+subject 唯一记录
+        script = db.query(Script).filter_by(project=project_name, subject=today_str).first()
+        if script:
+            # 存在则更新内容
+            script.content = json.dumps(content, ensure_ascii=False, separators=(",", ":"))
+            db.commit()
+            logger.info(f"Updated script for {today_str} with {article_title}.")
+        else:
+            # 不存在则新建
+            script = Script(
+                project=project_name,
+                subject=today_str,
+                content=json.dumps(content, ensure_ascii=False, separators=(",", ":"))
+            )
+            db.add(script)
+            db.commit()
+            logger.info(f"Saved script for {today_str} with {article_title}.")
+    except Exception as e:
+        db.rollback()
+        logger.error(f"Failed to save/update script for {today_str}: {e}")
+
+    # 3. 生成音频
+    try:
+        from tts.service import TTSService
+
+        article_text = content["阶段4_今日文章"]["文章正文"]
+        logger.debug(f"Synthesizing daily article audio for '{article_title}'")
+        article_audio = asyncio.run(TTSService.synthesize(
+            text=article_text,
+            voice="yanglan",
+            language="zh-CN"
+        ))
+
+        if not article_audio:
+            logger.warning("No audio synthesized for daily article.")
+            return
+
+        # 保存音频文件
+        out_dir = os.path.join("output", project_name)
+        os.makedirs(out_dir, exist_ok=True)
+        safe_title = "_".join(article_title.split())
+        audio_filename = f"{safe_title}_{today_str}.wav"
+        audio_path = os.path.join(out_dir, audio_filename)
+        with open(audio_path, "wb") as fw:
+            fw.write(article_audio.getvalue())
+        logger.info(f"Saved daily article audio to {audio_path}")
+    except Exception as e:
+        logger.error(f"Failed to synthesize/save daily article audio: {e}")
+
+# For manual testing
+if __name__ == "__main__":
+    # 每日文章生成
+    job_generate_daily_article()

scheduler/jobs.py

@@ -31,7 +31,7 @@ def job_generate_topics():
         script = db.query(Script).filter_by(project="梗文化研究所", subject=today_str).first()
         if script:
             # 存在则更新内容
-            script.content = json.dumps(content, ensure_ascii=False, indent=2)
+            script.content = json.dumps(content, ensure_ascii=False, separators=(",", ":"))
             db.commit()
             logger.info(f"Updated script for {today_str} with {len(topics)} topics.")
         else:
@@ -39,7 +39,7 @@ def job_generate_topics():
             script = Script(
                 project="梗文化研究所",
                 subject=today_str,
-                content=json.dumps(content, ensure_ascii=False, indent=2)
+                content=json.dumps(content, ensure_ascii=False, separators=(",", ":"))
             )
             db.add(script)
             db.commit()
@@ -96,7 +96,7 @@ def job_generate_bits():
         logger.debug(f"Generated bits for meme '{meme_name}': {bit}")
         bits.append(bit)
         content = {"topics": topics, "bits": bits}
-        script.content = json.dumps(content, ensure_ascii=False, indent=2)
+        script.content = json.dumps(content, ensure_ascii=False, separators=(",", ":"))
         db.commit()
         logger.info(f"Saved bits for meme '{meme_name}' with {len(bits)} segments.")
     except Exception as e:
@@ -141,13 +141,13 @@ def job_generate_script():
             parts.append("角度：" + "; ".join(top.get("angles", [])))
         
         research_text = "\n".join(parts)
-        materials_text = research_text + "\n\n" + json.dumps(bits, ensure_ascii=False, indent=2)
+        materials_text = research_text + "\n\n" + json.dumps(bits, ensure_ascii=False, separators=(",", ":"))
         
         # 调用 LLM 生成完整脚本
         from llm.generate_podcast import generate_script
         full_script = generate_script(meme_name, materials_text)
         content = {"topics": topics, "bits": bits, "script": full_script}
-        script.content = json.dumps(content, ensure_ascii=False, indent=2)
+        script.content = json.dumps(content, ensure_ascii=False, separators=(",", ":"))
         db.commit()
         logger.info(f"Saved full script for meme '{meme_name}'.")
     except Exception as e:
@@ -183,10 +183,12 @@ def job_synthesize_podcast_audio():
 
         # 角色到声音的映射（可按需扩展或放到配置中）
         role_voice_map = {
-            "host": settings.TTS_VOICE or "zh-CN-XiaoxiaoNeural",
-            "guest": "zh-CN-YunxiNeural",
+            "host": settings.TTS_VOICE or "yanglan",
+            "guest_a": "zhisheng",
+            "guest_b": "trump",
+            "guest_c": "tangseng",
             # fallback for other roles
-            "default": settings.TTS_VOICE or "zh-CN-XiaoxiaoNeural",
+            "default": settings.TTS_VOICE or "yanglan",
         }
 
         segment_audio_bytes = []
@@ -210,46 +212,66 @@ def job_synthesize_podcast_audio():
         if not segment_audio_bytes:
             logger.warning("No audio segments synthesized; aborting podcast save.")
             return
+        
+        # 保存每个分段为独立文件，并记录它们
+        segment_out_dir = os.path.join("output", "segments")
+        if os.path.exists(segment_out_dir):
+            # 删除旧文件
+            for f in os.listdir(segment_out_dir):
+                os.remove(os.path.join(segment_out_dir, f))
+        else:
+            os.makedirs(segment_out_dir, exist_ok=True)
+        segment_paths = []
+        safe_title = "_".join(title.split())
+        for idx, role, seg in sorted(segment_audio_bytes, key=lambda x: x[0]):
+            seg.seek(0)
+            seg_filename = f"{safe_title}_{script.subject}_{script.id}_seg{idx}_{role}.wav"
+            seg_path = os.path.join(segment_out_dir, seg_filename)
+            with open(seg_path, "wb") as fw:
+                fw.write(seg.getvalue())
+            segment_paths.append(seg_path)
+
+        logger.info(f"Saved {len(segment_paths)} segment files to {segment_out_dir}; combined file not created")
+
+        # 从segment_out_dir读取音频并合并
+        segment_audio_to_combined_bytes = []
+        for seg_path in segment_paths:
+            with open(seg_path, "rb") as fr:
+                audio_data = fr.read()
+                from io import BytesIO
+                segment_audio_to_combined_bytes.append((0, "segment", BytesIO(audio_data)))
+                # seg = segment_audio_to_combined_bytes[0][2]
+                # data = seg.getvalue()
+                # print("LEN:", len(data))
+                # print("HEAD (hex):", data[:32].hex())
+                # print("HEAD (ascii):", data[:32])
 
         # 保存或合并音频：优先使用 pydub (ffmpeg)，否则保存为独立段文件
         out_dir = os.path.join("output", "podcasts")
         os.makedirs(out_dir, exist_ok=True)
-        safe_title = "_".join(title.split())
-        final_filename = f"{safe_title}_{script.subject}_{script.id}.mp3"
+        final_filename = f"{safe_title}_{script.subject}_{script.id}.wav"
         final_path = os.path.join(out_dir, final_filename)
 
         try:
             from pydub import AudioSegment
 
             combined = None
-            for idx, role, seg in sorted(segment_audio_bytes, key=lambda x: x[0]):
+            for idx, role, seg in sorted(segment_audio_to_combined_bytes, key=lambda x: x[0]):
                 seg.seek(0)
-                audio_seg = AudioSegment.from_file(seg, format="mp3")
+                audio_seg = AudioSegment.from_file(seg, format="wav")
                 if combined is None:
                     combined = audio_seg
                 else:
                     combined = combined + audio_seg
 
             if combined is not None:
-                combined.export(final_path, format="mp3")
+                combined.export(final_path, format="wav")
                 logger.info(f"Saved combined podcast audio to {final_path}")
                 return
 
         except Exception as e:
             logger.warning(f"pydub/ffmpeg not available or merge failed: {e}; falling back to per-segment files")
 
-        # 回退：保存每个分段为独立文件，并记录它们
-        segment_paths = []
-        for idx, role, seg in sorted(segment_audio_bytes, key=lambda x: x[0]):
-            seg.seek(0)
-            seg_filename = f"{safe_title}_{script.subject}_{script.id}_seg{idx}_{role}.mp3"
-            seg_path = os.path.join(out_dir, seg_filename)
-            with open(seg_path, "wb") as fw:
-                fw.write(seg.getvalue())
-            segment_paths.append(seg_path)
-
-        logger.info(f"Saved {len(segment_paths)} segment files to {out_dir}; combined file not created")
-
     except Exception as e:
         logger.error(f"Failed to synthesize/save podcast audio: {e}")
     finally:
@@ -257,7 +279,14 @@ def job_synthesize_podcast_audio():
 
 # For manual testing
 if __name__ == "__main__":
+    # 选题策划和背景素材搜集
     # job_generate_topics()
+
+    # 脱口秀段子创作
     # job_generate_bits()
+
+    # 完整播客脚本生成
     # job_generate_script()
-    job_synthesize_podcast_audio()
+
+    # 播客音频合成
+    # job_synthesize_podcast_audio()