bdim/Canto
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

feat: add admin usage statistics and LLM configuration management

Browse Source
This commit is contained in:
bdim404
2026-03-12 16:30:24 +08:00
parent 202f2fa83b
commit 7f25dd09f6
16 changed files with 757 additions and 300 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
qwen3-tts-backend/core/llm_service.py
View File

@@ -1,7 +1,7 @@
import asyncio
import json
import logging
from typing import Any, Dict
from typing import Any, Callable, Dict, Optional
import httpx
@@ -14,7 +14,7 @@ class LLMService:
self.api_key = api_key
self.model = model
async def stream_chat(self, system_prompt: str, user_message: str, on_token=None, max_tokens: int = 8192) -> str:
async def stream_chat(self, system_prompt: str, user_message: str, on_token=None, max_tokens: int = 8192, usage_callback: Optional[Callable[[int, int], None]] = None) -> str:
url = f"{self.base_url}/chat/completions"
headers = {
"Authorization": f"Bearer {self.api_key}",
@@ -29,8 +29,10 @@ class LLMService:
"temperature": 0.3,
"max_tokens": max_tokens,
"stream": True,
"stream_options": {"include_usage": True},
}
full_text = ""
_usage = None
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:
@@ -46,6 +48,9 @@ class LLMService:
break
try:
chunk = json.loads(data)
if chunk.get("usage"):
_usage = chunk["usage"]
continue
delta = chunk["choices"][0]["delta"].get("content", "")
if delta:
full_text += delta
@@ -53,10 +58,12 @@ class LLMService:
on_token(delta)
except (json.JSONDecodeError, KeyError, IndexError):
continue
if _usage and usage_callback:
usage_callback(_usage.get("prompt_tokens", 0), _usage.get("completion_tokens", 0))
return full_text
async def stream_chat_json(self, system_prompt: str, user_message: str, on_token=None, max_tokens: int = 8192):
raw = await self.stream_chat(system_prompt, user_message, on_token, max_tokens=max_tokens)
async def stream_chat_json(self, system_prompt: str, user_message: str, on_token=None, max_tokens: int = 8192, usage_callback: Optional[Callable[[int, int], None]] = None):
raw = await self.stream_chat(system_prompt, user_message, on_token, max_tokens=max_tokens, usage_callback=usage_callback)
raw = raw.strip()
if not raw:
raise ValueError("LLM returned empty response")
@@ -74,7 +81,7 @@ class LLMService:
logger.error(f"JSON parse failed. Raw (first 500): {raw[:500]}")
raise
async def chat(self, system_prompt: str, user_message: str) -> str:
async def chat(self, system_prompt: str, user_message: str, usage_callback: Optional[Callable[[int, int], None]] = None) -> str:
url = f"{self.base_url}/chat/completions"
headers = {
"Authorization": f"Bearer {self.api_key}",
@@ -97,10 +104,13 @@ class LLMService:
logger.error(f"LLM API error {resp.status_code}: {resp.text}")
resp.raise_for_status()
data = resp.json()
usage = data.get("usage", {})
if usage and usage_callback:
usage_callback(usage.get("prompt_tokens", 0), usage.get("completion_tokens", 0))
return data["choices"][0]["message"]["content"]
async def chat_json(self, system_prompt: str, user_message: str) -> Any:
raw = await self.chat(system_prompt, user_message)
async def chat_json(self, system_prompt: str, user_message: str, usage_callback: Optional[Callable[[int, int], None]] = None) -> Any:
raw = await self.chat(system_prompt, user_message, usage_callback=usage_callback)
raw = raw.strip()
if not raw:
raise ValueError("LLM returned empty response")
@@ -118,7 +128,7 @@ class LLMService:
logger.error(f"JSON parse failed. Raw response (first 500 chars): {raw[:500]}")
raise
async def extract_characters(self, text_samples: list[str], on_token=None, on_sample=None, turbo: bool = False) -> list[Dict]:
async def extract_characters(self, text_samples: list[str], on_token=None, on_sample=None, turbo: bool = False, usage_callback: Optional[Callable[[int, int], None]] = 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"
"gender字段必须明确标注性别，只能取以下三个值之一：\"男\"、\"女\"、\"未知\"。\n"
@@ -145,7 +155,7 @@ class LLMService:
async def _extract_one(i: int, sample: str) -> list[Dict]:
user_message = f"请分析以下小说文本并提取角色：\n\n{sample}"
result = await self.stream_chat_json(system_prompt, user_message, None)
result = await self.stream_chat_json(system_prompt, user_message, None, usage_callback=usage_callback)
if on_sample:
on_sample(i, len(text_samples))
return result.get("characters", [])
@@ -160,14 +170,14 @@ class LLMService:
logger.warning(f"Character extraction failed for sample {i+1}: {r}")
else:
raw_all.extend(r)
return await self.merge_characters(raw_all)
return await self.merge_characters(raw_all, usage_callback=usage_callback)
raw_all: list[Dict] = []
for i, sample in enumerate(text_samples):
logger.info(f"Extracting characters from sample {i+1}/{len(text_samples)}")
user_message = f"请分析以下小说文本并提取角色：\n\n{sample}"
try:
result = await self.stream_chat_json(system_prompt, user_message, on_token)
result = await self.stream_chat_json(system_prompt, user_message, on_token, usage_callback=usage_callback)
raw_all.extend(result.get("characters", []))
except Exception as e:
logger.warning(f"Character extraction failed for sample {i+1}: {e}")
@@ -175,9 +185,9 @@ class LLMService:
on_sample(i, len(text_samples))
if len(text_samples) == 1:
return raw_all
return await self.merge_characters(raw_all)
return await self.merge_characters(raw_all, usage_callback=usage_callback)
async def merge_characters(self, raw_characters: list[Dict]) -> list[Dict]:
async def merge_characters(self, raw_characters: list[Dict], usage_callback: Optional[Callable[[int, int], None]] = None) -> list[Dict]:
system_prompt = (
"你是一个专业的小说角色整合助手。你收到的是从同一本书不同段落中提取的角色列表，其中可能存在重复。\n"
"请完成以下任务：\n"
@@ -191,7 +201,7 @@ class LLMService:
)
user_message = f"请整合以下角色列表：\n\n{json.dumps(raw_characters, ensure_ascii=False, indent=2)}"
try:
result = await self.chat_json(system_prompt, user_message)
result = await self.chat_json(system_prompt, user_message, usage_callback=usage_callback)
return result.get("characters", [])
except Exception as e:
logger.warning(f"Character merge failed, falling back to name-dedup: {e}")
@@ -202,7 +212,7 @@ class LLMService:
seen[name] = c
return list(seen.values())
async def parse_chapter_segments(self, chapter_text: str, character_names: list[str], on_token=None) -> list[Dict]:
async def parse_chapter_segments(self, chapter_text: str, character_names: list[str], on_token=None, usage_callback: Optional[Callable[[int, int], None]] = 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。"
@@ -217,7 +227,7 @@ class LLMService:
'{"character": "角色名", "text": "对话内容", "emo_text": "开心", "emo_alpha": 0.6}, ...]'
)
user_message = f"请解析以下章节文本：\n\n{chapter_text}"
result = await self.stream_chat_json(system_prompt, user_message, on_token, max_tokens=16384)
result = await self.stream_chat_json(system_prompt, user_message, on_token, max_tokens=16384, usage_callback=usage_callback)
if isinstance(result, list):
return result
return []