from __future__ import annotations import os from pathlib import Path import sys import typing as t import nltk from chatchat import __version__ from chatchat.pydantic_settings_file import * # chatchat 数据目录，必须通过环境变量设置。如未设置则自动使用当前目录。 CHATCHAT_ROOT = Path(os.environ.get("CHATCHAT_ROOT", ".")).resolve() XF_MODELS_TYPES = { "text2image": {"model_family": ["stable_diffusion"]}, "image2image": {"model_family": ["stable_diffusion"]}, "speech2text": {"model_family": ["whisper"]}, "text2speech": {"model_family": ["ChatTTS"]}, } class BasicSettings(BaseFileSettings): """ 服务器基本配置信息 除 log_verbose/HTTPX_DEFAULT_TIMEOUT 修改后即时生效 其它配置项修改后都需要重启服务器才能生效，服务运行期间请勿修改 """ model_config = SettingsConfigDict(yaml_file=CHATCHAT_ROOT / "basic_settings.yaml") version: str = __version__ """生成该配置模板的项目代码版本，如这里的值与程序实际版本不一致，建议重建配置文件模板""" log_verbose: bool = False """是否开启日志详细信息""" HTTPX_DEFAULT_TIMEOUT: float = 300 """httpx 请求默认超时时间（秒）。如果加载模型或对话较慢，出现超时错误，可以适当加大该值。""" # @computed_field @cached_property def PACKAGE_ROOT(self) -> Path: """代码根目录""" return Path(__file__).parent # @computed_field @cached_property def DATA_PATH(self) -> Path: """用户数据根目录""" p = CHATCHAT_ROOT / "data" return p # @computed_field @cached_property def IMG_DIR(self) -> Path: """项目相关图片目录""" p = self.PACKAGE_ROOT / "img" return p # @computed_field @cached_property def NLTK_DATA_PATH(self) -> Path: """nltk 模型存储路径""" p = self.PACKAGE_ROOT / "data/nltk_data" return p # @computed_field @cached_property def LOG_PATH(self) -> Path: """日志存储路径""" p = self.DATA_PATH / "logs" return p # @computed_field @cached_property def MEDIA_PATH(self) -> Path: """模型生成内容（图片、视频、音频等）保存位置""" p = self.DATA_PATH / "media" return p # @computed_field @cached_property def BASE_TEMP_DIR(self) -> Path: """临时文件目录，主要用于文件对话""" p = self.DATA_PATH / "temp" (p / "openai_files").mkdir(parents=True, exist_ok=True) return p KB_ROOT_PATH: str = str(CHATCHAT_ROOT / "data/knowledge_base") """知识库默认存储路径""" DB_ROOT_PATH: str = str(CHATCHAT_ROOT / "data/knowledge_base/info.db") """数据库默认存储路径。如果使用sqlite，可以直接修改DB_ROOT_PATH；如果使用其它数据库，请直接修改SQLALCHEMY_DATABASE_URI。""" SQLALCHEMY_DATABASE_URI:str = "sqlite:///" + str(CHATCHAT_ROOT / "data/knowledge_base/info.db") """知识库信息数据库连接URI""" OPEN_CROSS_DOMAIN: bool = False """API 是否开启跨域""" DEFAULT_BIND_HOST: str = "0.0.0.0" if sys.platform != "win32" else "127.0.0.1" """ 各服务器默认绑定host。如改为"0.0.0.0"需要修改下方所有XX_SERVER的host Windows 下 WEBUI 自动弹出浏览器时，如果地址为 "0.0.0.0" 是无法访问的，需要手动修改地址栏 """ API_SERVER: dict = {"host": DEFAULT_BIND_HOST, "port": 7861, "public_host": "127.0.0.1", "public_port": 7861} """API 服务器地址。其中 public_host 用于生成云服务公网访问链接（如知识库文档链接）""" WEBUI_SERVER: dict = {"host": DEFAULT_BIND_HOST, "port": 8501} """WEBUI 服务器地址""" def make_dirs(self): '''创建所有数据目录''' for p in [ self.DATA_PATH, self.MEDIA_PATH, self.LOG_PATH, self.BASE_TEMP_DIR, ]: p.mkdir(parents=True, exist_ok=True) for n in ["image", "audio", "video"]: (self.MEDIA_PATH / n).mkdir(parents=True, exist_ok=True) Path(self.KB_ROOT_PATH).mkdir(parents=True, exist_ok=True) class KBSettings(BaseFileSettings): """知识库相关配置""" model_config = SettingsConfigDict(yaml_file=CHATCHAT_ROOT / "kb_settings.yaml") DEFAULT_KNOWLEDGE_BASE: str = "samples" """默认使用的知识库""" DEFAULT_VS_TYPE: t.Literal["faiss", "milvus", "zilliz", "pg", "es", "relyt", "chromadb"] = "faiss" """默认向量库/全文检索引擎类型""" CACHED_VS_NUM: int = 1 """缓存向量库数量（针对FAISS）""" CACHED_MEMO_VS_NUM: int = 10 """缓存临时向量库数量（针对FAISS），用于文件对话""" CHUNK_SIZE: int = 750 """知识库中单段文本长度(不适用MarkdownHeaderTextSplitter)""" OVERLAP_SIZE: int = 150 """知识库中相邻文本重合长度(不适用MarkdownHeaderTextSplitter)""" VECTOR_SEARCH_TOP_K: int = 3 # TODO: 与 tool 配置项重复 """知识库匹配向量数量""" SCORE_THRESHOLD: float = 2.0 """知识库匹配相关度阈值，取值范围在0-2之间，SCORE越小，相关度越高，取到2相当于不筛选，建议设置在0.5左右""" DEFAULT_SEARCH_ENGINE: t.Literal["bing", "duckduckgo", "metaphor", "searx"] = "duckduckgo" """默认搜索引擎""" SEARCH_ENGINE_TOP_K: int = 3 """搜索引擎匹配结题数量""" ZH_TITLE_ENHANCE: bool = False """是否开启中文标题加强，以及标题增强的相关配置""" PDF_OCR_THRESHOLD: t.Tuple[float, float] = (0.6, 0.6) """ PDF OCR 控制：只对宽高超过页面一定比例（图片宽/页面宽，图片高/页面高）的图片进行 OCR。 这样可以避免 PDF 中一些小图片的干扰，提高非扫描版 PDF 处理速度 """ KB_INFO: t.Dict[str, str] = {"samples": "关于本项目issue的解答"} # TODO: 都存在数据库了，这个配置项还有必要吗？ """每个知识库的初始化介绍，用于在初始化知识库时显示和Agent调用，没写则没有介绍，不会被Agent调用。""" kbs_config: t.Dict[str, t.Dict] = { "faiss": {}, "milvus": { "host": "127.0.0.1", "port": "19530", "user": "", "password": "", "secure": False }, "zilliz": { "host": "in01-a7ce524e41e3935.ali-cn-hangzhou.vectordb.zilliz.com.cn", "port": "19530", "user": "", "password": "", "secure": True }, "pg": { "connection_uri": "postgresql://postgres:postgres@127.0.0.1:5432/langchain_chatchat" }, "relyt": { "connection_uri": "postgresql+psycopg2://postgres:postgres@127.0.0.1:7000/langchain_chatchat" }, "es": { "scheme": "http", "host": "127.0.0.1", "port": "9200", "index_name": "test_index", "user": "", "password": "", "verify_certs": True, "ca_certs": None, "client_cert": None, "client_key": None }, "milvus_kwargs": { "search_params": { "metric_type": "L2" }, "index_params": { "metric_type": "L2", "index_type": "HNSW" } }, "chromadb": {} } """可选向量库类型及对应配置""" text_splitter_dict: t.Dict[str, t.Dict[str, t.Any]] = { "ChineseRecursiveTextSplitter": { "source": "", "tokenizer_name_or_path": "", }, "SpacyTextSplitter": { "source": "huggingface", "tokenizer_name_or_path": "gpt2", }, "RecursiveCharacterTextSplitter": { "source": "tiktoken", "tokenizer_name_or_path": "cl100k_base", }, "MarkdownHeaderTextSplitter": { "headers_to_split_on": [ ("#", "head1"), ("##", "head2"), ("###", "head3"), ("####", "head4"), ] }, } """ TextSplitter配置项，如果你不明白其中的含义，就不要修改。 source 如果选择tiktoken则使用openai的方法 "huggingface" """ TEXT_SPLITTER_NAME: str = "ChineseRecursiveTextSplitter" """TEXT_SPLITTER 名称""" EMBEDDING_KEYWORD_FILE: str = "embedding_keywords.txt" """Embedding模型定制词语的词表文件""" class PlatformConfig(MyBaseModel): """模型加载平台配置""" platform_name: str = "xinference" """平台名称""" platform_type: t.Literal["xinference", "ollama", "oneapi", "fastchat", "openai", "custom openai"] = "xinference" """平台类型""" api_base_url: str = "http://127.0.0.1:9997/v1" """openai api url""" api_key: str = "EMPTY" """api key if available""" api_proxy: str = "" """API 代理""" api_concurrencies: int = 5 """该平台单模型最大并发数""" auto_detect_model: bool = False """是否自动获取平台可用模型列表。设为 True 时下方不同模型类型可自动检测""" llm_models: t.Union[t.Literal["auto"], t.List[str]] = [] """该平台支持的大语言模型列表，auto_detect_model 设为 True 时自动检测""" embed_models: t.Union[t.Literal["auto"], t.List[str]] = [] """该平台支持的嵌入模型列表，auto_detect_model 设为 True 时自动检测""" text2image_models: t.Union[t.Literal["auto"], t.List[str]] = [] """该平台支持的图像生成模型列表，auto_detect_model 设为 True 时自动检测""" image2text_models: t.Union[t.Literal["auto"], t.List[str]] = [] """该平台支持的多模态模型列表，auto_detect_model 设为 True 时自动检测""" rerank_models: t.Union[t.Literal["auto"], t.List[str]] = [] """该平台支持的重排模型列表，auto_detect_model 设为 True 时自动检测""" speech2text_models: t.Union[t.Literal["auto"], t.List[str]] = [] """该平台支持的 STT 模型列表，auto_detect_model 设为 True 时自动检测""" text2speech_models: t.Union[t.Literal["auto"], t.List[str]] = [] """该平台支持的 TTS 模型列表，auto_detect_model 设为 True 时自动检测""" class ApiModelSettings(BaseFileSettings): """模型配置项""" model_config = SettingsConfigDict(yaml_file=CHATCHAT_ROOT / "model_settings.yaml") DEFAULT_LLM_MODEL: str = "glm4-chat" """默认选用的 LLM 名称""" DEFAULT_EMBEDDING_MODEL: str = "bge-m3" """默认选用的 Embedding 名称""" Agent_MODEL: str = "" # TODO: 似乎与 LLM_MODEL_CONFIG 重复了 """AgentLM模型的名称 (可以不指定，指定之后就锁定进入Agent之后的Chain的模型，不指定就是 DEFAULT_LLM_MODEL)""" HISTORY_LEN: int = 3 """默认历史对话轮数""" MAX_TOKENS: t.Optional[int] = None # TODO: 似乎与 LLM_MODEL_CONFIG 重复了 """大模型最长支持的长度，如果不填写，则使用模型默认的最大长度，如果填写，则为用户设定的最大长度""" TEMPERATURE: float = 0.7 """LLM通用对话参数""" SUPPORT_AGENT_MODELS: t.List[str] = [ "chatglm3-6b", "glm-4", "openai-api", "Qwen-2", "qwen2-instruct", "gpt-3.5-turbo", "gpt-4o", ] """支持的Agent模型""" LLM_MODEL_CONFIG: t.Dict[str, t.Dict] = { # 意图识别不需要输出，模型后台知道就行 "preprocess_model": { "model": "", "temperature": 0.05, "max_tokens": 4096, "history_len": 10, "prompt_name": "default", "callbacks": False, }, "llm_model": { "model": "", "temperature": 0.9, "max_tokens": 4096, "history_len": 10, "prompt_name": "default", "callbacks": True, }, "action_model": { "model": "", "temperature": 0.01, "max_tokens": 4096, "history_len": 10, "prompt_name": "ChatGLM3", "callbacks": True, }, "postprocess_model": { "model": "", "temperature": 0.01, "max_tokens": 4096, "history_len": 10, "prompt_name": "default", "callbacks": True, }, "image_model": { "model": "sd-turbo", "size": "256*256", }, } """ LLM模型配置，包括了不同模态初始化参数。 `model` 如果留空则自动使用 DEFAULT_LLM_MODEL """ MODEL_PLATFORMS: t.List[PlatformConfig] = [ PlatformConfig(**{ "platform_name": "xinference", "platform_type": "xinference", "api_base_url": "http://127.0.0.1:9997/v1", "api_key": "EMPTY", "api_concurrencies": 5, "auto_detect_model": True, "llm_models": [], "embed_models": [], "text2image_models": [], "image2text_models": [], "rerank_models": [], "speech2text_models": [], "text2speech_models": [], }), PlatformConfig(**{ "platform_name": "ollama", "platform_type": "ollama", "api_base_url": "http://127.0.0.1:11434/v1", "api_key": "EMPTY", "api_concurrencies": 5, "llm_models": [ "qwen:7b", "qwen2:7b", ], "embed_models": [ "quentinz/bge-large-zh-v1.5", ], }), PlatformConfig(**{ "platform_name": "oneapi", "platform_type": "oneapi", "api_base_url": "http://127.0.0.1:3000/v1", "api_key": "sk-", "api_concurrencies": 5, "llm_models": [ # 智谱 API "chatglm_pro", "chatglm_turbo", "chatglm_std", "chatglm_lite", # 千问 API "qwen-turbo", "qwen-plus", "qwen-max", "qwen-max-longcontext", # 千帆 API "ERNIE-Bot", "ERNIE-Bot-turbo", "ERNIE-Bot-4", # 星火 API "SparkDesk", ], "embed_models": [ # 千问 API "text-embedding-v1", # 千帆 API "Embedding-V1", ], "text2image_models": [], "image2text_models": [], "rerank_models": [], "speech2text_models": [], "text2speech_models": [], }), PlatformConfig(**{ "platform_name": "openai", "platform_type": "openai", "api_base_url": "https://api.openai.com/v1", "api_key": "sk-proj-", "api_concurrencies": 5, "llm_models": [ "gpt-4o", "gpt-3.5-turbo", ], "embed_models": [ "text-embedding-3-small", "text-embedding-3-large", ], }), ] """模型平台配置""" class ToolSettings(BaseFileSettings): """Agent 工具配置项""" model_config = SettingsConfigDict(yaml_file=CHATCHAT_ROOT / "tool_settings.yaml", json_file=CHATCHAT_ROOT / "tool_settings.json", extra="allow") search_local_knowledgebase: dict = { "use": False, "top_k": 3, "score_threshold": 2.0, "conclude_prompt": { "with_result": '<指令>根据已知信息，简洁和专业的来回答问题。如果无法从中得到答案，请说 "根据已知信息无法回答该问题"，' "不允许在答案中添加编造成分，答案请使用中文。 \n" "<已知信息>{{ context }}\n" "<问题>{{ question }}\n", "without_result": "请你根据我的提问回答我的问题:\n" "{{ question }}\n" "请注意，你必须在回答结束后强调，你的回答是根据你的经验回答而不是参考资料回答的。\n", }, } '''本地知识库工具配置项''' search_internet: dict = { "use": False, "search_engine_name": "tavily", "search_engine_config": { "bing": { "bing_search_url": "https://api.bing.microsoft.com/v7.0/search", "bing_key": "", }, "metaphor": { "metaphor_api_key": "", "split_result": False, "chunk_size": 500, "chunk_overlap": 0, }, "duckduckgo": {}, "searx": { "host": "https://metasearx.com", "engines": [], "categories": [], "language": "zh-CN", }, "tavily":{ "tavily_api_key": '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', "include_answer": True, "search_depth": "advanced", "include_raw_content": True, "max_results": 1 } }, "top_k": 5, "verbose": "Origin", "conclude_prompt": "<指令>这是搜索到的互联网信息，请你根据这些信息进行提取并有调理，简洁的回答问题,不得包含有重复的词汇或句子。如果无法从中得到答案，请说 “无法搜索到能回答问题的内容”。 " "\n<已知信息>{{ context }}\n" "<问题>\n" "{{ question }}\n" "\n", } '''搜索引擎工具配置项。推荐自己部署 searx 搜索引擎，国内使用最方便。''' arxiv: dict = { "use": False, } weather_check: dict = { "use": False, "api_key": "", } '''心知天气（https://www.seniverse.com/）工具配置项''' search_youtube: dict = { "use": False, } wolfram: dict = { "use": False, "appid": "", } calculate: dict = { "use": False, } '''numexpr 数学计算工具配置项''' text2images: dict = { "use": False, "model": "sd-turbo", "size": "256*256", } '''图片生成工具配置项。model 必须是在 model_settings.yaml/MODEL_PLATFORMS 中配置过的。''' text2sql: dict = { # 该工具需单独指定使用的大模型，与用户前端选择使用的模型无关 "model_name": "qwen-plus", "use": False, # SQLAlchemy连接字符串，支持的数据库有： # crate、duckdb、googlesql、mssql、mysql、mariadb、oracle、postgresql、sqlite、clickhouse、prestodb # 不同的数据库请查阅SQLAlchemy用法，修改sqlalchemy_connect_str，配置对应的数据库连接，如sqlite为sqlite:///数据库文件路径，下面示例为mysql # 如提示缺少对应数据库的驱动，请自行通过poetry安装 "sqlalchemy_connect_str": "mysql+pymysql://用户名:密码@主机地址/数据库名称", # 务必评估是否需要开启read_only,开启后会对sql语句进行检查，请确认text2sql.py中的intercept_sql拦截器是否满足你使用的数据库只读要求 # 优先推荐从数据库层面对用户权限进行限制 "read_only": False, # 限定返回的行数 "top_k": 50, # 是否返回中间步骤 "return_intermediate_steps": True, # 如果想指定特定表，请填写表名称，如["sys_user","sys_dept"]，不填写走智能判断应该使用哪些表 "table_names": [], # 对表名进行额外说明，辅助大模型更好的判断应该使用哪些表，尤其是SQLDatabaseSequentialChain模式下,是根据表名做的预测，很容易误判。 "table_comments": { # 如果出现大模型选错表的情况，可尝试根据实际情况填写表名和说明 # "tableA":"这是一个用户表，存储了用户的基本信息", # "tableB":"角色表", }, } ''' text2sql使用建议 1、因大模型生成的sql可能与预期有偏差，请务必在测试环境中进行充分测试、评估； 2、生产环境中，对于查询操作，由于不确定查询效率，推荐数据库采用主从数据库架构，让text2sql连接从数据库，防止可能的慢查询影响主业务； 3、对于写操作应保持谨慎，如不需要写操作，设置read_only为True,最好再从数据库层面收回数据库用户的写权限，防止用户通过自然语言对数据库进行修改操作； 4、text2sql与大模型在意图理解、sql转换等方面的能力有关，可切换不同大模型进行测试； 5、数据库表名、字段名应与其实际作用保持一致、容易理解，且应对数据库表名、字段进行详细的备注说明，帮助大模型更好理解数据库结构； 6、若现有数据库表名难于让大模型理解，可配置下面table_comments字段，补充说明某些表的作用。 ''' amap: dict = { "use": False, "api_key": "高德地图 API KEY", } '''高德地图、天气相关工具配置项。''' text2promql: dict = { "use": False, # : "prometheus_endpoint": "http://127.0.0.1:9090", # "username": "", # "password": "", } ''' text2promql 使用建议 1、因大模型生成的 promql 可能与预期有偏差, 请务必在测试环境中进行充分测试、评估; 2、text2promql 与大模型在意图理解、metric 选择、promql 转换等方面的能力有关, 可切换不同大模型进行测试; 3、当前仅支持 单prometheus 查询, 后续考虑支持 多prometheus 查询. ''' url_reader: dict = { "use": False, "timeout": "10000", } '''URL内容阅读（https://r.jina.ai/）工具配置项 请确保部署的网络环境良好，以免造成超时等问题''' class PromptSettings(BaseFileSettings): """Prompt 模板.除 Agent 模板使用 f-string 外，其它均使用 jinja2 格式""" model_config = SettingsConfigDict(yaml_file=CHATCHAT_ROOT / "prompt_settings.yaml", json_file=CHATCHAT_ROOT / "prompt_settings.json", extra="allow") preprocess_model: dict = { "default": ( "你只要回复0 和 1 ，代表不需要使用工具。以下几种问题不需要使用工具:\n" "1. 需要联网查询的内容\n" "2. 需要计算的内容\n" "3. 需要查询实时性的内容\n" "如果我的输入满足这几种情况，返回1。其他输入，请你回复0，你只要返回一个数字\n" "这是我的问题:" ), } """意图识别用模板""" llm_model: dict = { "default": "{{input}}", "with_history": ( "The following is a friendly conversation between a human and an AI.\n" "The AI is talkative and provides lots of specific details from its context.\n" "If the AI does not know the answer to a question, it truthfully says it does not know.\n\n" "Current conversation:\n" "{{history}}\n" "Human: {{input}}\n" "AI:" ), } '''普通 LLM 用模板''' rag: dict = { "default": ( "【指令】根据已知信息，简洁和专业的来回答问题,不得包含有重复的词汇或句子。" "如果无法从中得到答案，请说 “根据已知信息无法回答该问题”，不允许在答案中添加编造成分，答案请使用中文。\n\n" "【已知信息】{{context}}\n\n" "【问题】{{question}}\n" ), "empty": ( "请你回答我的问题:\n" "{{question}}" ), } '''RAG 用模板，可用于知识库问答、文件对话、搜索引擎对话''' action_model: dict = { "GPT-4": ( "Answer the following questions as best you can. You have access to the following tools:\n" "The way you use the tools is by specifying a json blob.\n" "Specifically, this json should have a `action` key (with the name of the tool to use) and a `action_input` key (with the input to the tool going here).\n" 'The only values that should be in the "action" field are: {tool_names}\n' "The $JSON_BLOB should only contain a SINGLE action, do NOT return a list of multiple actions. Here is an example of a valid $JSON_BLOB:\n" "```\n\n" "{{{{\n" ' "action": $TOOL_NAME,\n' ' "action_input": $INPUT\n' "}}}}\n" "```\n\n" "ALWAYS use the following format:\n" "Question: the input question you must answer\n" "Thought: you should always think about what to do\n" "Action:\n" "```\n\n" "$JSON_BLOB" "```\n\n" "Observation: the result of the action\n" "... (this Thought/Action/Observation can repeat N times)\n" "Thought: I now know the final answer\n" "Final Answer: the final answer to the original input question\n" "Begin! Reminder to always use the exact characters `Final Answer` when responding.\n" "Question:{input}\n" "Thought:{agent_scratchpad}\n" ), "ChatGLM3": ( "You can answer using the tools.Respond to the human as helpfully and accurately as possible.\n" "You have access to the following tools:\n" "{tools}\n" "Use a json blob to specify a tool by providing an action key (tool name)\n" "and an action_input key (tool input).\n" 'Valid "action" values: "Final Answer" or [{tool_names}]\n' "Provide only ONE action per $JSON_BLOB, as shown:\n\n" "```\n" "{{{{\n" ' "action": $TOOL_NAME,\n' ' "action_input": $INPUT\n' "}}}}\n" "```\n\n" "Follow this format:\n\n" "Question: input question to answer\n" "Thought: consider previous and subsequent steps\n" "Action:\n" "```\n" "$JSON_BLOB\n" "```\n" "Observation: action result\n" "... (repeat Thought/Action/Observation N times)\n" "Thought: I know what to respond\n" "Action:\n" "```\n" "{{{{\n" ' "action": "Final Answer",\n' ' "action_input": "Final response to human"\n' "}}}}\n" "Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary.\n" "Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation:.\n" "Question: {input}\n\n" "{agent_scratchpad}\n" ), "qwen": ( "Answer the following questions as best you can. You have access to the following APIs:\n\n" "{tools}\n\n" "Use the following format:\n\n" "Question: the input question you must answer\n" "Thought: you should always think about what to do\n" "Action: the action to take, should be one of [{tool_names}]\n" "Action Input: the input to the action\n" "Observation: the result of the action\n" "... (this Thought/Action/Action Input/Observation can be repeated zero or more times)\n" "Thought: I now know the final answer\n" "Final Answer: the final answer to the original input question\n\n" "Format the Action Input as a JSON object.\n\n" "Begin!\n\n" "Question: {input}\n\n" "{agent_scratchpad}\n\n" ), "structured-chat-agent": ( "Respond to the human as helpfully and accurately as possible. You have access to the following tools:\n\n" "{tools}\n\n" "Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).\n\n" 'Valid "action" values: "Final Answer" or {tool_names}\n\n' "Provide only ONE action per $JSON_BLOB, as shown:\n\n" '```\n{{\n "action": $TOOL_NAME,\n "action_input": $INPUT\n}}\n```\n\n' "Follow this format:\n\n" "Question: input question to answer\n" "Thought: consider previous and subsequent steps\n" "Action:\n```\n$JSON_BLOB\n```\n" "Observation: action result\n" "... (repeat Thought/Action/Observation N times)\n" "Thought: I know what to respond\n" 'Action:\n```\n{{\n "action": "Final Answer",\n "action_input": "Final response to human"\n}}\n\n' "Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation\n" "{input}\n\n" "{agent_scratchpad}\n\n" # '(reminder to respond in a JSON blob no matter what)') ), } """Agent 模板""" postprocess_model: dict = { "default": "{{input}}", } """后处理模板""" class SettingsContainer: CHATCHAT_ROOT = CHATCHAT_ROOT basic_settings: BasicSettings = settings_property(BasicSettings()) kb_settings: KBSettings = settings_property(KBSettings()) model_settings: ApiModelSettings = settings_property(ApiModelSettings()) tool_settings: ToolSettings = settings_property(ToolSettings()) prompt_settings: PromptSettings = settings_property(PromptSettings()) def createl_all_templates(self): self.basic_settings.create_template_file(write_file=True) self.kb_settings.create_template_file(write_file=True) self.model_settings.create_template_file(sub_comments={ "MODEL_PLATFORMS": {"model_obj": PlatformConfig(), "is_entire_comment": True}}, write_file=True) self.tool_settings.create_template_file(write_file=True, file_format="yaml", model_obj=ToolSettings()) self.prompt_settings.create_template_file(write_file=True, file_format="yaml") def set_auto_reload(self, flag: bool=True): self.basic_settings.auto_reload = flag self.kb_settings.auto_reload = flag self.model_settings.auto_reload = flag self.tool_settings.auto_reload = flag self.prompt_settings.auto_reload = flag Settings = SettingsContainer() nltk.data.path.append(str(Settings.basic_settings.NLTK_DATA_PATH)) if __name__ == "__main__": Settings.createl_all_templates()