Added setting frontend. Prepare for v0.2.0 release.

Author: wehos

README.MD

@@ -44,33 +44,27 @@ Lanlan（兰兰）是一个新手友好、开箱即用的，具有听觉、视
 
 1. **体验桌宠模式**。如果网页版可以正常使用，可以考虑继续通过`启动App版.bat`实现桌面穿透。注意，**请不要同时使用网页版和App版。请确认exe文件没有被系统或杀毒软件隔离。** *使用结束后请在桌面右下角找到小八图标，右键退出。*
 
->  *对于开发者，请在克隆本项目后，(1)新建pyhon3.12环境。(2)执行`pip install -r requirements.txt`安装依赖。(3)复制`config/api_template.py`到`config/api.py`并进行必要配置。(4)执行`python memory_server.py`和`python main_server.py`。(5)通过main server中指定的端口（默认为`localhost:48911`）访问网页版。*
+>  *对于开发者，请在克隆本项目后，(1)新建pyhon3.12环境。(2)执行`pip install -r requirements.txt`安装依赖。(3)复制`config/api_template.py`到`config/api.py`并进行必要配置。(4)执行`python memory_server.py`和`python main_server.py`。(5)通过main server中指定的端口（默认为`http://localhost:48911`）访问网页版。*
 
 # 进阶内容
 
-## A. 修改人设
+## 修改人设
 
-基本人设位于`config/__init__.py`内，请用文本编辑器打开。请将`MASTER_NAME`修改为自己的名字，`her_name`修改为~~猫娘~~伙伴的名字 *（注意，这只是一个临时措施，本项目支持多个角色并发）*。`master_basic_config`和`lanlan_basic_config`中以json格式填入基本信息。如果对json格式有疑问，请查询豆包等AI工具。
+- 网页版访问`http://localhost:48911/chara_manager`即可进入人设编辑页面。初始 ~~猫娘~~ 伙伴的预设名称为`test`，建议直接修改名字，并一项一项添加或修改基础人设，但尽量控制数量。
 
-进阶人设位于`config/prompts_chara.py`内，请谨慎修改。冗长的人设会降低系统的运行效率和稳定性。开发者由衷希望对猫娘进行设定时请遵循奥卡姆剃刀原则，"如无必要，勿增设定"。
+- 进阶人设主要包括**Live2D模型设置(live2d)**和**声音设置(voice_id)**。如果你想要更改**Live2D模型**，请先将模型目录复制到本项目中的`static`文件夹下。从进阶设置中可以进入Live2D模型管理界面，可以更换模型，并通过拖拽和鼠标滚轮调整模型的位置和大小。如果你想要更改**角色声音**，请准备一段15秒左右的连贯、干净的语音录音。通过进阶设置进入语音设置页面，上传录音即可完成自定义语音。
 
-## B. 修改Live2D模型
+- 进阶人设中还有一个`system_prompt`，可以对系统指令进行完全自定义，但不建议修改。
 
-Live2D模型的路径参数目前硬编码在了`main_server.py`里的`"model_path"`部分，可以自行修改（有两个，先改第一个）。更换Live2D模型后如果想要调整大小和位置，还需要修改`templates/index.html`里的`model.scale`和`model.anchor`参数。表情控制目前还没有准备好release，待UI完善后才会考虑正式支持自定义Live2D模型。
-
-## C. 修改声音
-
-本项目已经内置了基于CosyVoice API的语音克隆功能，代码中已经包含该功能并且已经通过测试。请根据阿里百炼大模型平台[官网的教程](https://help.aliyun.com/zh/model-studio/cosyvoice-clone-api)，进行语音克隆。克隆后，将`VOICE_ID`填写在`config/api.py`中，并将`USE_TTS`设置为`True`。
-
-## D. 参与开发
+## 参与开发
 
 本项目环境依赖非常简单，请在`python3.12`环境中执行`pip install -r requirements.txt`即可。请注意将`config/api_template.py`复制为`config/api.py`.开发者建议加入企鹅群1048307485，猫娘名称见项目标题。
 
 # TODO List（开发计划）
 
 ## A. 高优先级
 
-1. 添加前端UI用于人设管理（L2D模型/语音/性格等）、记忆管理（记忆检索和纠错）。
+1. 添加前端UI用于记忆管理（记忆检索和纠错）。
 
 1. 支持L2D表情控制和动作控制。
 

config/__init__.py

@@ -1,39 +1,46 @@
 from config.api import *
 from config.prompts_chara import *
+import json
+import os
+
+# 读取角色配置
+CHARACTER_JSON_PATH = os.path.join(os.path.dirname(__file__), 'characters.json')
+# 默认值
+_default_master = {"档案名": "哥哥", "性别": "男", "昵称": "哥哥"}
+_default_lanlan = {"test": {"性别": "女", "年龄": 15, "昵称": "T酱, 小T", "live2d": "mao_pro", "voice_id": "", "system_prompt": lanlan_prompt}}
+
+def get_character_data():
+    try:
+        with open(CHARACTER_JSON_PATH, 'r', encoding='utf-8') as f:
+            character_data = json.load(f)
+    except FileNotFoundError:
+        print(f"⚠️ 未找到猫娘配置文件: {CHARACTER_JSON_PATH}，请检查文件是否存在。使用默认人设。")
+        character_data = {"主人": _default_master, "猫娘": _default_lanlan}
+    except Exception as e:
+        print(f"💥 读取猫娘配置文件出错: {e}，使用默认人设。")
+        character_data = {"主人": _default_master, "猫娘": _default_lanlan}
+
+    # MASTER_NAME 必须始终存在，取档案名
+    MASTER_NAME = character_data.get('主人', {}).get('档案名', _default_master['档案名'])
+    # 获取所有猫娘名
+    catgirl_names = list(character_data['猫娘'].keys()) if character_data['猫娘'] and len(character_data['猫娘']) > 0 else list(_default_lanlan.keys())
+    her_name = catgirl_names[0] if catgirl_names else ''
+    master_basic_config = character_data.get('主人', _default_master)
+    lanlan_basic_config = character_data['猫娘'] if catgirl_names else _default_lanlan
+
+    NAME_MAPPING = {'human': MASTER_NAME, 'system': "SYSTEM_MESSAGE"}
+    # 生成以猫娘名为key的各类store
+    LANLAN_PROMPT = {name: character_data['猫娘'][name].get('system_prompt', lanlan_prompt) for name in catgirl_names}
+    SEMANTIC_STORE = {name: f'memory/store/semantic_memory_{name}' for name in catgirl_names}
+    TIME_STORE = {name: f'memory/store/time_indexed_{name}' for name in catgirl_names}
+    SETTING_STORE = {name: f'memory/store/settings_{name}.json' for name in catgirl_names}
+    RECENT_LOG = {name: f'memory/store/recent_{name}.json' for name in catgirl_names}
+
+    return MASTER_NAME, her_name, master_basic_config, lanlan_basic_config, NAME_MAPPING, LANLAN_PROMPT, SEMANTIC_STORE, TIME_STORE, SETTING_STORE, RECENT_LOG
 
 TIME_ORIGINAL_TABLE_NAME = "time_indexed_original"
 TIME_COMPRESSED_TABLE_NAME = "time_indexed_compressed"
 
-
-'''
-↓↓↓ 核心人设在这里 ↓↓↓
-'''
-MASTER_NAME = '哥哥'
-her_name = "test" 
-master_basic_config = {'性别': '男', '昵称': MASTER_NAME}
-
-lanlan_basic_config = {her_name: {'性别': '女',
-                                '年龄': 15,
-                                '昵称': ["T酱", "小T"],
-                                }}
-'''
-↑↑↑ 核心人设在这里 ↑↑↑
-'''
-
-
-"""
-本项目支持多个角色，但是为了方便新手用户进行配置，临时增加了一个her_name变量来帮助批量设置初始角色的信息。
-请将her_name后的字符串修改为角色名称。
-"""
-NAME_MAPPING = {'human': MASTER_NAME, 'system': "SYSTEM_MESSAGE"}
-LANLAN_PROMPT = {her_name: lanlan_prompt}
-SEMANTIC_STORE = {her_name: f'memory/store/semantic_memory_{her_name}'}
-TIME_STORE = {her_name: f'memory/store/time_indexed_{her_name}'}
-SETTING_STORE = {her_name: f'memory/store/settings_{her_name}.json'}
-RECENT_LOG = {her_name: f'memory/store/recent_{her_name}.json'}
-
-
-import json
 try:
     with open('core_config.txt', 'r') as f:
         core_cfg = json.load(f)

config/prompts_sys.py

@@ -1,4 +1,5 @@
-from config import MASTER_NAME
+from config import get_character_data
+MASTER_NAME, _, _, _, _, _, _, _, _, _ = get_character_data()
 
 gpt4_1_system = """## PERSISTENCE
 You are an agent - please keep going until the user's query is completely 

main_helper/core.py

@@ -22,7 +22,7 @@
 import base64
 from io import BytesIO
 from PIL import Image
-from config import MASTER_NAME, MEMORY_SERVER_PORT, CORE_API_KEY, CORE_URL, CORE_MODEL, USE_TTS
+from config import get_character_data, CORE_URL, CORE_MODEL, CORE_API_KEY, MEMORY_SERVER_PORT, AUDIO_API_KEY
 from multiprocessing import Process, Queue as MPQueue
 from uuid import uuid4
 import numpy as np
@@ -62,7 +62,27 @@ def __init__(self, sync_message_queue, lanlan_name, lanlan_prompt):
 
         self.lanlan_prompt = lanlan_prompt
         self.lanlan_name = lanlan_name
-        self.MODEL = CORE_MODEL
+        # 获取角色相关配置
+        (
+            self.master_name,
+            self.her_name,
+            self.master_basic_config,
+            self.lanlan_basic_config,
+            self.name_mapping,
+            self.lanlan_prompt_map,
+            self.semantic_store,
+            self.time_store,
+            self.setting_store,
+            self.recent_log
+        ) = get_character_data()
+        # 获取API相关配置
+        self.model = CORE_MODEL
+        self.core_url = CORE_URL
+        self.core_api_key = CORE_API_KEY
+        self.memory_server_port = MEMORY_SERVER_PORT
+        self.audio_api_key = AUDIO_API_KEY
+        self.voice_id = self.lanlan_basic_config[self.lanlan_name].get('voice_id', '')
+        self.use_tts = False if not self.voice_id else True
         self.generation_config = {}  # Qwen暂时不用
         self.message_cache_for_new_session = []
         self.is_preparing_new_session = False
@@ -75,9 +95,6 @@ def __init__(self, sync_message_queue, lanlan_name, lanlan_prompt):
         self.pending_session = None
         self.is_hot_swap_imminent = False
         self.tts_handler_task = None
-        self.use_tts = USE_TTS
-        # 将TTS相关的导入移到外部，确保始终可用
-        
         # 热切换相关变量
         self.background_preparation_task = None
         self.final_swap_task = None
@@ -86,9 +103,9 @@ def __init__(self, sync_message_queue, lanlan_name, lanlan_prompt):
 
         # 注册回调
         self.session = OmniRealtimeClient(
-            base_url=CORE_URL,
-            api_key=CORE_API_KEY,
-            model=self.MODEL,
+            base_url=self.core_url,
+            api_key=self.core_api_key,
+            model=self.model,
             voice="Chelsie",
             on_text_delta=self.handle_text_data,
             on_audio_delta=self.handle_audio_data,
@@ -188,8 +205,8 @@ async def handle_input_transcript(self, transcript: str):
             if not hasattr(self, 'message_cache_for_new_session'):
                 self.message_cache_for_new_session = []
             if len(self.message_cache_for_new_session) == 0 or self.message_cache_for_new_session[-1]['role'] == self.lanlan_name:
-                self.message_cache_for_new_session.append({"role": MASTER_NAME, "text": transcript.strip()})
-            elif self.message_cache_for_new_session[-1]['role'] == MASTER_NAME:
+                self.message_cache_for_new_session.append({"role": self.master_name, "text": transcript.strip()})
+            elif self.message_cache_for_new_session[-1]['role'] == self.master_name:
                 self.message_cache_for_new_session[-1]['text'] += transcript.strip()
         # 可选：推送用户活动
         with self.lock:
@@ -215,7 +232,7 @@ async def send_lanlan_response(self, text: str, is_first_chunk: bool = False):
                 if hasattr(self, 'is_preparing_new_session') and self.is_preparing_new_session:
                     if not hasattr(self, 'message_cache_for_new_session'):
                         self.message_cache_for_new_session = []
-                    if len(self.message_cache_for_new_session) == 0 or self.message_cache_for_new_session[-1]['role']==MASTER_NAME:
+                    if len(self.message_cache_for_new_session) == 0 or self.message_cache_for_new_session[-1]['role']==self.master_name:
                         self.message_cache_for_new_session.append(
                             {"role": self.lanlan_name, "text": text})
                     elif self.message_cache_for_new_session[-1]['role'] == self.lanlan_name:
@@ -289,10 +306,9 @@ async def start_session(self, websocket: WebSocket, new=False):
         if self.use_tts:
             # 启动TTS子进程
             if self.tts_process is None or not self.tts_process.is_alive():
-                from config import AUDIO_API_KEY, VOICE_ID
                 self.tts_process = Process(
                     target=speech_synthesis_worker,
-                    args=(self.tts_request_queue, self.tts_response_queue, AUDIO_API_KEY, VOICE_ID)
+                    args=(self.tts_request_queue, self.tts_response_queue, self.audio_api_key, self.voice_id)
                 )
                 self.tts_process.daemon = True
                 self.tts_process.start()
@@ -309,9 +325,9 @@ async def start_session(self, websocket: WebSocket, new=False):
         try:
             # 获取初始 prompt
             initial_prompt = self.lanlan_prompt
-            initial_prompt += requests.get(f"http://localhost:{MEMORY_SERVER_PORT}/new_dialog/{self.lanlan_name}").text
-            logger.info("====Initial Prompt=====")
-            logger.info(initial_prompt)
+            initial_prompt += requests.get(f"http://localhost:{self.memory_server_port}/new_dialog/{self.lanlan_name}").text
+            # logger.info("====Initial Prompt=====")
+            # logger.info(initial_prompt)
 
             # 标记 session 激活
             if self.session:
@@ -361,9 +377,9 @@ async def _background_prepare_pending_session(self):
         try:
             # 创建新的pending session
             self.pending_session = OmniRealtimeClient(
-                base_url=CORE_URL,
-                api_key=CORE_API_KEY,
-                model=self.MODEL,
+                base_url=self.core_url,
+                api_key=self.core_api_key,
+                model=self.model,
                 voice="Chelsie",
                 on_text_delta=self.handle_text_data,
                 on_audio_delta=self.handle_audio_data,
@@ -377,7 +393,7 @@ async def _background_prepare_pending_session(self):
             initial_prompt = self.lanlan_prompt
             self.initial_cache_snapshot_len = len(self.message_cache_for_new_session)
             async with httpx.AsyncClient() as client:
-                resp = await client.get(f"http://localhost:{MEMORY_SERVER_PORT}/new_dialog/{self.lanlan_name}")
+                resp = await client.get(f"http://localhost:{self.memory_server_port}/new_dialog/{self.lanlan_name}")
                 initial_prompt += resp.text + self._convert_cache_to_str(self.message_cache_for_new_session)
             # print(initial_prompt)
             await self.pending_session.connect(initial_prompt, native_audio = not self.use_tts)
@@ -415,7 +431,7 @@ async def _perform_final_swap_sequence(self):
             # 1. Send incremental cache (or a heartbeat) to PENDING session for its *second* ignored response
             if incremental_cache:
                 final_prime_text = f"SYSTEM_MESSAGE | " + self._convert_cache_to_str(incremental_cache) + \
-                    f'=======以上为前情概要。现在请{self.lanlan_name}准备，即将开始用语音与{MASTER_NAME}继续对话。\n'
+                    f'=======以上为前情概要。现在请{self.lanlan_name}准备，即将开始用语音与{self.master_name}继续对话。\n'
             else:  # Ensure session cycles a turn even if no incremental cache
                 logger.error(f"💥 Unexpected: No incremental cache found. {len(self.message_cache_for_new_session)}, {self.initial_cache_snapshot_len}")
                 final_prime_text = f"SYSTEM_MESSAGE | 系统自动报时，当前时间： " + str(
@@ -717,14 +733,14 @@ async def tts_response_handler(self):
 
 # TTS多进程worker函数，供主进程Process(target=...)调用
 
-def speech_synthesis_worker(request_queue, response_queue, AUDIO_API_KEY, VOICE_ID):
+def speech_synthesis_worker(request_queue, response_queue, audio_api_key, voice_id):
     import dashscope
     from dashscope.audio.tts_v2 import ResultCallback, SpeechSynthesizer, AudioFormat
     import numpy as np
     from librosa import resample
     import re
     import time
-    dashscope.api_key = AUDIO_API_KEY
+    dashscope.api_key = audio_api_key
     class Callback(ResultCallback):
         def __init__(self, response_queue):
             self.response_queue = response_queue
@@ -766,7 +782,7 @@ def on_data(self, data: bytes) -> None:
                         pass
                 synthesizer = SpeechSynthesizer(
                     model="cosyvoice-v2",
-                    voice=VOICE_ID,
+                    voice=voice_id,
                     speech_rate=1.1,
                     format=AudioFormat.PCM_24000HZ_MONO_16BIT,
                     callback=callback,