- chatchat init 可以直接指定 Xinference API 地址、LLM 模型、Embedding 模型，一步完成知识库初…

…始化 (chatchat-space#4425)

- 修复一些配置项错误

README.md

@@ -182,6 +182,10 @@ set CHATCHAT_ROOT=/path/to/chatchat_data
 2. 执行初始化
 ```shell
 chatchat init
+# 如果你已经启动了 Xinference 服务，可以直接指定 Xinference API 地址、LLM 模型、Embedding 模型，可以跳过第3、4步，直接按第5步启动服务：
+# chatchat init -x http://127.0.0.1:9999/v1 -l qwen2-instruct -e bce -r
+# chatchat start -a
+# 具体查看 chatchat init --help
 ```
 该命令会执行以下操作：
 - 创建所有需要的数据目录

libs/chatchat-server/chatchat/cli.py

@@ -4,7 +4,7 @@
 import typing as t
 
 from chatchat.startup import main as startup_main
-from chatchat.init_database import main as kb_main, create_tables
+from chatchat.init_database import main as kb_main, create_tables, folder2db
 from chatchat.settings import Settings
 from chatchat.utils import build_logger
 
@@ -18,20 +18,60 @@ def main():
 
 
 @main.command("init", help="项目初始化")
-# @click.option("-k", "--recreate-kb", "kb_names",
-#               show_default=True,
-#               default="samples",
-#               help="同时重建知识库。可以指定多个知识库名称，以 , 分隔。")
-def init():
+@click.option("-x", "--xinference-endpoint", "xf_endpoint",
+              help="指定Xinference API 服务地址。默认为 http://127.0.0.1:9997/v1")
+@click.option("-l", "--llm-model",
+              help="指定默认 LLM 模型。默认为 glm4-chat")
+@click.option("-e", "--embed-model",
+              help="指定默认 Embedding 模型。默认为 bge-large-zh-v1.5")
+@click.option("-r", "--recreate-kb",
+              is_flag=True,
+              show_default=True,
+              default=False,
+              help="同时重建知识库（必须确保指定的 embed model 可用）。")
+@click.option("-k", "--kb-names", "kb_names",
+              show_default=True,
+              default="samples",
+              help="要重建知识库的名称。可以指定多个知识库名称，以 , 分隔。")
+def init(
+    xf_endpoint: str = "",
+    llm_model: str = "",
+    embed_model: str = "",
+    recreate_kb: bool = False,
+    kb_names: str = "",
+):
+    Settings.set_auto_reload(False)
     bs = Settings.basic_settings
+    kb_names = [x.strip() for x in kb_names.split(",")]
     logger.info(f"开始初始化项目数据目录：{Settings.CHATCHAT_ROOT}")
     Settings.basic_settings.make_dirs()
     logger.info("创建所有数据目录：成功。")
     shutil.copytree(bs.PACKAGE_ROOT / "data/knowledge_base/samples", Path(bs.KB_ROOT_PATH) / "samples", dirs_exist_ok=True)
-    logger.info("复制 samples 知识库：成功。")
+    logger.info("复制 samples 知识库文件：成功。")
+    create_tables()
+    logger.info("初始化知识库数据库：成功。")
+
+    if xf_endpoint:
+        Settings.model_settings.MODEL_PLATFORMS[0].api_base_url = xf_endpoint
+    if llm_model:
+        Settings.model_settings.DEFAULT_LLM_MODEL = llm_model
+    if embed_model:
+        Settings.model_settings.DEFAULT_EMBEDDING_MODEL = embed_model
+
     Settings.createl_all_templates()
+    Settings.set_auto_reload(True)
+
     logger.info("生成默认配置文件：成功。")
-    logger.warning("<red>请先修改 model_settings.yaml 配置正确的模型平台、LLM模型和Embed模型，然后执行 chatchat kb -r 初始化知识库。</red>")
+    logger.warning("<red>请先检查 model_settings.yaml 里模型平台、LLM模型和Embed模型信息正确</red>")
+
+    if recreate_kb:
+        folder2db(kb_names=kb_names,
+                  mode="recreate_vs",
+                  vs_type=Settings.kb_settings.DEFAULT_VS_TYPE,
+                  embed_model=Settings.model_settings.DEFAULT_EMBEDDING_MODEL)
+        logger.success("所有初始化已完成，执行 chatchat start -a 启动服务。")
+    else:
+        logger.warning("执行 chatchat kb -r 初始化知识库，然后 chatchat start -a 启动服务。")
 
 
 main.add_command(startup_main, "start")

libs/chatchat-server/chatchat/pydantic_settings_file.py

@@ -180,6 +180,18 @@ class BaseFileSettings(BaseSettings):
         env_file_encoding="utf-8",
     )
 
+    def model_post_init(self, __context: os.Any) -> None:
+        self._auto_reload = True
+        return super().model_post_init(__context)
+
+    @property
+    def auto_reload(self) -> bool:
+        return self._auto_reload
+
+    @auto_reload.setter
+    def auto_reload(self, val: bool):
+        self._auto_reload = val
+
     @classmethod
     def settings_customise_sources(
         cls,
@@ -191,23 +203,24 @@ def settings_customise_sources(
     ) -> tuple[PydanticBaseSettingsSource, ...]:
         return init_settings, env_settings, dotenv_settings, YamlConfigSettingsSource(settings_cls)
 
-    @classmethod
     def create_template_file(
-        cls,
-        init_kwds: t.Dict = {},
+        self,
+        model_obj: BaseFileSettings=None,
         dump_kwds: t.Dict={},
         sub_comments: t.Dict[str, SubModelComment]={},
         write_file: bool | str | Path = False,
         file_format: t.Literal["yaml", "json"] = "yaml",
     ) -> str:
+        if model_obj is None:
+            model_obj = self
         if file_format == "yaml":
-            template = YamlTemplate(cls(**init_kwds), dump_kwds=dump_kwds, sub_comments=sub_comments)
+            template = YamlTemplate(model_obj=model_obj, dump_kwds=dump_kwds, sub_comments=sub_comments)
             return template.create_yaml_template(write_to=write_file)
         else:
             dump_kwds.setdefault("indent", 4)
-            data = cls(**init_kwds).model_dump_json(**dump_kwds)
+            data = model_obj.model_dump_json(**dump_kwds)
             if write_file:
-                write_file = cls.model_config.get("json_file")
+                write_file = self.model_config.get("json_file")
                 with open(write_file, "w", encoding="utf-8") as fp:
                     fp.write(data)
             return data
@@ -218,20 +231,21 @@ def _lazy_load_key(settings: BaseSettings):
     for n in ["env_file", "json_file", "yaml_file", "toml_file"]:
         key = None
         if file := settings.model_config.get(n):
-            if os.path.isfile(file):
+            if os.path.isfile(file) and os.path.getsize(file) > 0:
                 key = int(os.path.getmtime(file))
         keys.append(key)
     return tuple(keys)
 
 
-_T = t.TypeVar("_T", bound=BaseSettings)
+_T = t.TypeVar("_T", bound=BaseFileSettings)
 
 @cached(max_size=1, algorithm=CachingAlgorithmFlag.LRU, thread_safe=True, custom_key_maker=_lazy_load_key)
 def _cached_settings(settings: _T) -> _T:
     '''
     the sesstings is cached, and refreshed when configuration files changed
     '''
-    settings.__init__()
+    if settings.auto_reload:
+        settings.__init__()
     return settings
 
 

libs/chatchat-server/chatchat/server/chat/chat.py

@@ -37,23 +37,23 @@ def create_models_from_config(configs, callbacks, stream, max_tokens):
     configs = configs or Settings.model_settings.LLM_MODEL_CONFIG
     models = {}
     prompts = {}
-    for model_type, model_configs in configs.items():
-        for model_name, params in model_configs.items():
-            callbacks = callbacks if params.get("callbacks", False) else None
-            # 判断是否传入 max_tokens 的值, 如果传入就按传入的赋值(api 调用且赋值), 如果没有传入则按照初始化配置赋值(ui 调用或 api 调用未赋值)
-            max_tokens_value = max_tokens if max_tokens is not None else params.get("max_tokens", 1000)
-            model_instance = get_ChatOpenAI(
-                model_name=model_name,
-                temperature=params.get("temperature", 0.5),
-                max_tokens=max_tokens_value,
-                callbacks=callbacks,
-                streaming=stream,
-                local_wrap=True,
-            )
-            models[model_type] = model_instance
-            prompt_name = params.get("prompt_name", "default")
-            prompt_template = get_prompt_template(type=model_type, name=prompt_name)
-            prompts[model_type] = prompt_template
+    for model_type, params in configs.items():
+        model_name = params.get("model", "").strip() or Settings.model_settings.DEFAULT_LLM_MODEL
+        callbacks = callbacks if params.get("callbacks", False) else None
+        # 判断是否传入 max_tokens 的值, 如果传入就按传入的赋值(api 调用且赋值), 如果没有传入则按照初始化配置赋值(ui 调用或 api 调用未赋值)
+        max_tokens_value = max_tokens if max_tokens is not None else params.get("max_tokens", 1000)
+        model_instance = get_ChatOpenAI(
+            model_name=model_name,
+            temperature=params.get("temperature", 0.5),
+            max_tokens=max_tokens_value,
+            callbacks=callbacks,
+            streaming=stream,
+            local_wrap=True,
+        )
+        models[model_type] = model_instance
+        prompt_name = params.get("prompt_name", "default")
+        prompt_template = get_prompt_template(type=model_type, name=prompt_name)
+        prompts[model_type] = prompt_template
     return models, prompts
 
 

libs/chatchat-server/chatchat/server/chat/file_chat.py

@@ -76,7 +76,7 @@ def parse_file(file: UploadFile) -> dict:
 def upload_temp_docs(
     files: List[UploadFile] = File(..., description="上传文件，支持多文件"),
     prev_id: str = Form(None, description="前知识库ID"),
-    chunk_size: int = Form(Settings.model_settings.LLM_MODEL_CONFIG, description="知识库中单段文本最大长度"),
+    chunk_size: int = Form(Settings.kb_settings.CHUNK_SIZE, description="知识库中单段文本最大长度"),
     chunk_overlap: int = Form(Settings.kb_settings.OVERLAP_SIZE, description="知识库中相邻文本重合长度"),
     zh_title_enhance: bool = Form(Settings.kb_settings.ZH_TITLE_ENHANCE, description="是否开启中文标题加强"),
 ) -> BaseResponse:

libs/chatchat-server/chatchat/server/knowledge_base/kb_service/es_kb_service.py

@@ -25,11 +25,12 @@ class ESKBService(KBService):
     def do_init(self):
         self.kb_path = self.get_kb_path(self.kb_name)
         self.index_name = os.path.split(self.kb_path)[-1]
-        self.IP = Settings.kb_settings.kbs_config[self.vs_type()]["host"]
-        self.PORT = Settings.kb_settings.kbs_config[self.vs_type()]["port"]
-        self.user = Settings.kb_settings.kbs_config[self.vs_type()].get("user", "")
-        self.password = Settings.kb_settings.kbs_config[self.vs_type()].get("password", "")
-        self.dims_length = Settings.kb_settings.kbs_config[self.vs_type()].get("dims_length", None)
+        kb_config = Settings.kb_settings.kbs_config[self.vs_type()]
+        self.IP = kb_config["host"]
+        self.PORT = kb_config["port"]
+        self.user = kb_config.get("user", "")
+        self.password = kb_config.get("password", "")
+        self.dims_length = kb_config.get("dims_length", None)
         self.embeddings_model = get_Embeddings(self.embed_model)
         try:
             # ES python客户端连接（仅连接）

libs/chatchat-server/chatchat/server/knowledge_base/utils.py

@@ -350,7 +350,7 @@ def docs2texts(
         docs: List[Document] = None,
         zh_title_enhance: bool = Settings.kb_settings.ZH_TITLE_ENHANCE,
         refresh: bool = False,
-        chunk_size: int = Settings.model_settings.DEFAULT_EMBEDDING_MODEL,
+        chunk_size: int = Settings.kb_settings.CHUNK_SIZE,
         chunk_overlap: int = Settings.kb_settings.OVERLAP_SIZE,
         text_splitter: TextSplitter = None,
     ):
@@ -382,7 +382,7 @@ def file2text(
         self,
         zh_title_enhance: bool = Settings.kb_settings.ZH_TITLE_ENHANCE,
         refresh: bool = False,
-        chunk_size: int = Settings.model_settings.DEFAULT_EMBEDDING_MODEL,
+        chunk_size: int = Settings.kb_settings.CHUNK_SIZE,
         chunk_overlap: int = Settings.kb_settings.OVERLAP_SIZE,
         text_splitter: TextSplitter = None,
     ):
@@ -421,7 +421,7 @@ def files2docs_in_thread_file2docs(
 
 def files2docs_in_thread(
     files: List[Union[KnowledgeFile, Tuple[str, str], Dict]],
-    chunk_size: int = Settings.model_settings.DEFAULT_EMBEDDING_MODEL,
+    chunk_size: int = Settings.kb_settings.CHUNK_SIZE,
     chunk_overlap: int = Settings.kb_settings.OVERLAP_SIZE,
     zh_title_enhance: bool = Settings.kb_settings.ZH_TITLE_ENHANCE,
 ) -> Generator: