feat(audiobook): implement log streaming for project status updates and enhance progress tracking

qwen3-tts-backend/core/llm_service.py

@@ -13,6 +13,65 @@ class LLMService:
         self.api_key = api_key
         self.model = model
 
+    async def stream_chat(self, system_prompt: str, user_message: str, on_token=None) -> str:
+        url = f"{self.base_url}/chat/completions"
+        headers = {
+            "Authorization": f"Bearer {self.api_key}",
+            "Content-Type": "application/json",
+        }
+        payload = {
+            "model": self.model,
+            "messages": [
+                {"role": "system", "content": system_prompt},
+                {"role": "user", "content": user_message},
+            ],
+            "temperature": 0.3,
+            "stream": True,
+        }
+        full_text = ""
+        timeout = httpx.Timeout(connect=10.0, read=90.0, write=10.0, pool=5.0)
+        async with httpx.AsyncClient(timeout=timeout) as client:
+            async with client.stream("POST", url, json=payload, headers=headers) as resp:
+                if resp.status_code != 200:
+                    body = await resp.aread()
+                    logger.error(f"LLM streaming error {resp.status_code}: {body}")
+                    resp.raise_for_status()
+                async for line in resp.aiter_lines():
+                    if not line.startswith("data: "):
+                        continue
+                    data = line[6:]
+                    if data.strip() == "[DONE]":
+                        break
+                    try:
+                        chunk = json.loads(data)
+                        delta = chunk["choices"][0]["delta"].get("content", "")
+                        if delta:
+                            full_text += delta
+                            if on_token:
+                                on_token(delta)
+                    except (json.JSONDecodeError, KeyError, IndexError):
+                        continue
+        return full_text
+
+    async def stream_chat_json(self, system_prompt: str, user_message: str, on_token=None):
+        raw = await self.stream_chat(system_prompt, user_message, on_token)
+        raw = raw.strip()
+        if not raw:
+            raise ValueError("LLM returned empty response")
+        if raw.startswith("```"):
+            lines = raw.split("\n")
+            inner = lines[1:]
+            if inner and inner[-1].strip().startswith("```"):
+                inner = inner[:-1]
+            raw = "\n".join(inner).strip()
+        if not raw:
+            raise ValueError("LLM returned empty JSON after stripping markdown")
+        try:
+            return json.loads(raw)
+        except json.JSONDecodeError:
+            logger.error(f"JSON parse failed. Raw (first 500): {raw[:500]}")
+            raise
+
     async def chat(self, system_prompt: str, user_message: str) -> str:
         url = f"{self.base_url}/chat/completions"
         headers = {
@@ -56,7 +115,7 @@ class LLMService:
             logger.error(f"JSON parse failed. Raw response (first 500 chars): {raw[:500]}")
             raise
 
-    async def extract_characters(self, text: str) -> list[Dict]:
+    async def extract_characters(self, text: str, on_token=None) -> list[Dict]:
         system_prompt = (
             "你是一个专业的小说分析助手兼声音导演。请分析给定的小说文本，提取所有出现的角色（包括旁白narrator）。\n"
             "对每个角色，instruct字段必须是详细的声音导演说明，需覆盖以下六个维度，每个维度单独一句，用换行分隔：\n"
@@ -70,10 +129,10 @@ class LLMService:
             '{"characters": [{"name": "narrator", "description": "第三人称叙述者", "instruct": "音色信息：...\\n身份背景：...\\n年龄设定：...\\n外貌特征：...\\n性格特质：...\\n叙事风格：..."}, ...]}'
         )
         user_message = f"请分析以下小说文本并提取角色：\n\n{text[:30000]}"
-        result = await self.chat_json(system_prompt, user_message)
+        result = await self.stream_chat_json(system_prompt, user_message, on_token)
         return result.get("characters", [])
 
-    async def parse_chapter_segments(self, chapter_text: str, character_names: list[str]) -> list[Dict]:
+    async def parse_chapter_segments(self, chapter_text: str, character_names: list[str], on_token=None) -> list[Dict]:
         names_str = "、".join(character_names)
         system_prompt = (
             "你是一个专业的有声书制作助手。请将给定的章节文本解析为对话片段列表。"
@@ -83,7 +142,7 @@ class LLMService:
             '[{"character": "narrator", "text": "叙述文字"}, {"character": "角色名", "text": "对话内容"}, ...]'
         )
         user_message = f"请解析以下章节文本：\n\n{chapter_text}"
-        result = await self.chat_json(system_prompt, user_message)
+        result = await self.stream_chat_json(system_prompt, user_message, on_token)
         if isinstance(result, list):
             return result
         return []