diff --git a/README.md b/README.md index 67aa6ef..8d41b73 100644 --- a/README.md +++ b/README.md @@ -59,7 +59,7 @@ docker run -d --gpus all -p 80:8501 registry.cn-beijing.aliyuncs.com/chatchat/ch 该项目是一个可以实现 __完全本地化__推理的知识库增强方案, 重点解决数据安全保护，私域化部署的企业痛点。 本开源方案采用```Apache License```，可以免费商用，无需付费。 -我们支持市面上主流的本地大预言模型和Embedding模型，支持开源的本地向量数据库。 +我们支持市面上主流的本地大语言模型和Embedding模型，支持开源的本地向量数据库。 支持列表详见[Wiki](https://github.com/chatchat-space/Langchain-Chatchat/wiki/) @@ -92,14 +92,14 @@ $ pip install -r requirements_webui.txt 如需在本地或离线环境下运行本项目，需要首先将项目所需的模型下载至本地，通常开源 LLM 与 Embedding 模型可以从 [HuggingFace](https://huggingface.co/models) 下载。 -以本项目中默认使用的 LLM 模型 [THUDM/ChatGLM2-6B](https://huggingface.co/THUDM/chatglm2-6b) 与 Embedding 模型 [moka-ai/m3e-base](https://huggingface.co/moka-ai/m3e-base) 为例： +以本项目中默认使用的 LLM 模型 [THUDM/ChatGLM3-6B](https://huggingface.co/THUDM/chatglm3-6b) 与 Embedding 模型 [BAAI/bge-large-zh](https://huggingface.co/BAAI/bge-large-zh) 为例： 下载模型需要先[安装 Git LFS](https://docs.github.com/zh/repositories/working-with-files/managing-large-files/installing-git-large-file-storage)，然后运行 ```Shell $ git lfs install -$ git clone https://huggingface.co/THUDM/chatglm2-6b -$ git clone https://huggingface.co/moka-ai/m3e-base +$ git clone https://huggingface.co/THUDM/chatglm3-6b +$ git clone https://huggingface.co/BAAI/bge-large-zh ``` ### 3. 初始化知识库和配置文件 @@ -148,7 +148,7 @@ $ python startup.py -a [![Telegram](https://img.shields.io/badge/Telegram-2CA5E0?style=for-the-badge&logo=telegram&logoColor=white "langchain-chatglm")](https://t.me/+RjliQ3jnJ1YyN2E9) ### 项目交流群 -二维码 +二维码 🎉 Langchain-Chatchat 项目微信交流群，如果你也对本项目感兴趣，欢迎加入群聊参与讨论交流。 diff --git a/configs/model_config.py.example b/configs/model_config.py.example index d07b960..be2f412 100644 --- a/configs/model_config.py.example +++ b/configs/model_config.py.example @@ -105,6 +105,7 @@ ONLINE_LLM_MODEL = { "version": "qwen-turbo", # 可选包括 "qwen-turbo", "qwen-plus" "api_key": "", # 请在阿里云控制台模型服务灵积API-KEY管理页面创建 "provider": "QwenWorker", + "embed_model": "text-embedding-v1" # embedding 模型名称 }, # 百川 API，申请方式请参考 https://www.baichuan-ai.com/home#api-enter @@ -241,7 +242,7 @@ MODEL_PATH = { "agentlm-13b": "THUDM/agentlm-13b", "agentlm-70b": "THUDM/agentlm-70b", - "Yi-34B-Chat": "https://huggingface.co/01-ai/Yi-34B-Chat", + "Yi-34B-Chat": "01-ai/Yi-34B-Chat", }, "reranker":{ "bge-reranker-large":"BAAI/bge-reranker-large", diff --git a/img/qr_code_77.jpg b/img/qr_code_77.jpg new file mode 100644 index 0000000..6a2ae7a Binary files /dev/null and b/img/qr_code_77.jpg differ diff --git a/img/qr_code_78.jpg b/img/qr_code_78.jpg new file mode 100644 index 0000000..a85e890 Binary files /dev/null and b/img/qr_code_78.jpg differ diff --git a/img/qr_code_79.jpg b/img/qr_code_79.jpg new file mode 100644 index 0000000..d784160 Binary files /dev/null and b/img/qr_code_79.jpg differ diff --git a/img/qr_code_80.jpg b/img/qr_code_80.jpg new file mode 100644 index 0000000..ffb0e0e Binary files /dev/null and b/img/qr_code_80.jpg differ diff --git a/img/qr_code_81.jpg b/img/qr_code_81.jpg new file mode 100644 index 0000000..07a4980 Binary files /dev/null and b/img/qr_code_81.jpg differ diff --git a/img/qr_code_82.jpg b/img/qr_code_82.jpg new file mode 100644 index 0000000..c3e0cfe Binary files /dev/null and b/img/qr_code_82.jpg differ diff --git a/knowledge_base/samples/content/wiki b/knowledge_base/samples/content/wiki index 4c6f1a9..2f24adb 160000 --- a/knowledge_base/samples/content/wiki +++ b/knowledge_base/samples/content/wiki @@ -1 +1 @@ -Subproject commit 4c6f1a9748563b5d3ae81a628ff8a71f3e5686d1 +Subproject commit 2f24adb218f23eab00d7fcd7ccf5072f2f35cb3c diff --git a/requirements.txt b/requirements.txt index 830552c..5d09a6b 100644 --- a/requirements.txt +++ b/requirements.txt @@ -1,18 +1,18 @@ # API requirements # On Windows system, install the cuda version manually from https://pytorch.org/ -torch>=2.1.2 -torchvision>=0.16.2 -torchaudio>=2.1.2 +torch~=2.1.2 +torchvision~=0.16.2 +torchaudio~=2.1.2 -langchain>=0.0.352 -langchain-experimental>=0.0.47 +langchain==0.0.352 +langchain-experimental==0.0.47 pydantic==1.10.13 -fschat>=0.2.33 -xformers>=0.0.23.post1 -openai>=1.6.0 -sentence_transformers>=2.2.2 -transformers>=4.36.2 +fschat==0.2.34 +xformers==0.0.23.post1 +openai~=1.6.0 +sentence_transformers==2.2.2 +transformers==4.36.2 fastapi>=0.105 nltk>=3.8.1 uvicorn>=0.24.0.post1 @@ -20,11 +20,11 @@ starlette~=0.27.0 unstructured[all-docs]==0.11.0 python-magic-bin; sys_platform == 'win32' SQLAlchemy==2.0.19 -faiss-cpu>=1.7.4 -accelerate>=0.24.1 -spacy>=3.7.2 -PyMuPDF>=1.23.8 -rapidocr_onnxruntime>=1.3.8 +faiss-cpu~=1.7.4 +accelerate==0.24.1 +spacy~=3.7.2 +PyMuPDF~=1.23.8 +rapidocr_onnxruntime~=1.3.8 requests>=2.31.0 pathlib>=1.0.1 pytest>=7.4.3 @@ -37,20 +37,20 @@ websockets>=12.0 numpy~=1.26.2 pandas~=2.1.4 einops>=0.7.0 -transformers_stream_generator>=0.0.4 -vllm>=0.2.6; sys_platform == "linux" +transformers_stream_generator==0.0.4 +vllm==0.2.6; sys_platform == "linux" # optional document loaders rapidocr_paddle[gpu]>=1.3.0.post5 jq>=1.6.0 html2text -beautifulsoup4>=4.12.2 -pysrt>=1.1.2 +beautifulsoup4~=4.12.2 +pysrt~=1.1.2 # Online api libs dependencies -zhipuai>=1.0.7 +zhipuai>=1.0.7，<=2.0.0 dashscope>=1.13.6 volcengine>=1.0.119 pymilvus>=2.3.4 @@ -61,17 +61,15 @@ pymilvus>=2.3.4 arxiv>=2.0.0 youtube-search>=2.1.2 -duckduckgo-search>=4.1.0 +duckduckgo-search>=3.9.9 metaphor-python>=0.1.23 # WebUI requirements -streamlit>=1.29.0 +streamlit~=1.29.0 streamlit-option-menu>=0.3.6 -streamlit-antd-components>=0.2.4 -streamlit-chatbox>=1.1.11 +streamlit-chatbox==1.1.11 streamlit-modal>=0.1.0 streamlit-aggrid>=0.3.4.post3 httpx[brotli,http2,socks]>=0.25.2 watchdog>=3.0.0 - diff --git a/requirements_api.txt b/requirements_api.txt index 4f975cf..237eb4e 100644 --- a/requirements_api.txt +++ b/requirements_api.txt @@ -1,17 +1,18 @@ # API requirements -torch>=2.1.2 -torchvision>=0.16.2 -torchaudio>=2.1.2 +# On Windows system, install the cuda version manually from https://pytorch.org/ +torch~=2.1.2 +torchvision~=0.16.2 +torchaudio~=2.1.2 -langchain>=0.0.352 -langchain-experimental>=0.0.47 +langchain==0.0.352 +langchain-experimental==0.0.47 pydantic==1.10.13 -fschat>=0.2.34 -xformers>=0.0.23.post1 -openai>=1.6.0 -sentence_transformers>=2.2.2 -transformers>=4.36.2 +fschat==0.2.34 +xformers==0.0.23.post1 +openai~=1.6.0 +sentence_transformers==2.2.2 +transformers==4.36.2 fastapi>=0.105 nltk>=3.8.1 uvicorn>=0.24.0.post1 @@ -19,11 +20,11 @@ starlette~=0.27.0 unstructured[all-docs]==0.11.0 python-magic-bin; sys_platform == 'win32' SQLAlchemy==2.0.19 -faiss-cpu>=1.7.4 -accelerate>=0.24.1 -spacy>=3.7.2 -PyMuPDF>=1.23.8 -rapidocr_onnxruntime>=1.3.8 +faiss-cpu~=1.7.4 +accelerate==0.24.1 +spacy~=3.7.2 +PyMuPDF~=1.23.8 +rapidocr_onnxruntime~=1.3.8 requests>=2.31.0 pathlib>=1.0.1 pytest>=7.4.3 @@ -36,16 +37,16 @@ websockets>=12.0 numpy~=1.26.2 pandas~=2.1.4 einops>=0.7.0 -transformers_stream_generator>=0.0.4 -vllm>=0.2.6; sys_platform == "linux" +transformers_stream_generator==0.0.4 +vllm==0.2.6; sys_platform == "linux" # optional document loaders rapidocr_paddle[gpu]>=1.3.0.post5 jq>=1.6.0 html2text -beautifulsoup4>=4.12.2 -pysrt>=1.1.2 +beautifulsoup4~=4.12.2 +pysrt~=1.1.2 # Online api libs dependencies @@ -60,5 +61,5 @@ pymilvus>=2.3.4 arxiv>=2.0.0 youtube-search>=2.1.2 -duckduckgo-search>=4.1.0 +duckduckgo-search>=3.9.9 metaphor-python>=0.1.23 \ No newline at end of file diff --git a/requirements_webui.txt b/requirements_webui.txt index 3113053..aa3eca7 100644 --- a/requirements_webui.txt +++ b/requirements_webui.txt @@ -1,9 +1,9 @@ # WebUI requirements -streamlit>=1.29.0 # do remember to add streamlit to environment variables if you use windows +streamlit~=1.29.0 streamlit-option-menu>=0.3.6 -streamlit-chatbox>=1.1.11 +streamlit-chatbox==1.1.11 streamlit-modal>=0.1.0 streamlit-aggrid>=0.3.4.post3 +httpx[brotli,http2,socks]>=0.25.2 watchdog>=3.0.0 - diff --git a/server/chat/file_chat.py b/server/chat/file_chat.py index d9ba5b6..a3bbdfc 100644 --- a/server/chat/file_chat.py +++ b/server/chat/file_chat.py @@ -132,7 +132,7 @@ async def file_chat(query: str = Body(..., description="用户输入", examples= context = "\n".join([doc.page_content for doc in docs]) if len(docs) == 0: ## 如果没有找到相关文档，使用Empty模板 - prompt_template = get_prompt_template("knowledge_base_chat", "Empty") + prompt_template = get_prompt_template("knowledge_base_chat", "empty") else: prompt_template = get_prompt_template("knowledge_base_chat", prompt_name) input_msg = History(role="user", content=prompt_template).to_msg_template(False) diff --git a/server/embeddings_api.py b/server/embeddings_api.py index 93555a3..440bb77 100644 --- a/server/embeddings_api.py +++ b/server/embeddings_api.py @@ -6,32 +6,32 @@ from fastapi import Body from fastapi.concurrency import run_in_threadpool from typing import Dict, List - online_embed_models = list_online_embed_models() def embed_texts( - texts: List[str], - embed_model: str = EMBEDDING_MODEL, - to_query: bool = False, + texts: List[str], + embed_model: str = EMBEDDING_MODEL, + to_query: bool = False, ) -> BaseResponse: ''' 对文本进行向量化。返回数据格式：BaseResponse(data=List[List[float]]) TODO: 也许需要加入缓存机制，减少 token 消耗 ''' try: - if embed_model in list_embed_models(): # 使用本地Embeddings模型 + if embed_model in list_embed_models(): # 使用本地Embeddings模型 from server.utils import load_local_embeddings embeddings = load_local_embeddings(model=embed_model) return BaseResponse(data=embeddings.embed_documents(texts)) - if embed_model in list_online_embed_models(): # 使用在线API + if embed_model in list_online_embed_models(): # 使用在线API config = get_model_worker_config(embed_model) worker_class = config.get("worker_class") + embed_model = config.get("embed_model") worker = worker_class() if worker_class.can_embedding(): - params = ApiEmbeddingsParams(texts=texts, to_query=to_query) + params = ApiEmbeddingsParams(texts=texts, to_query=to_query, embed_model=embed_model) resp = worker.do_embeddings(params) return BaseResponse(**resp) @@ -67,9 +67,10 @@ async def aembed_texts( def embed_texts_endpoint( - texts: List[str] = Body(..., description="要嵌入的文本列表", examples=[["hello", "world"]]), - embed_model: str = Body(EMBEDDING_MODEL, description=f"使用的嵌入模型，除了本地部署的Embedding模型，也支持在线API({online_embed_models})提供的嵌入服务。"), - to_query: bool = Body(False, description="向量是否用于查询。有些模型如Minimax对存储/查询的向量进行了区分优化。"), + texts: List[str] = Body(..., description="要嵌入的文本列表", examples=[["hello", "world"]]), + embed_model: str = Body(EMBEDDING_MODEL, + description=f"使用的嵌入模型，除了本地部署的Embedding模型，也支持在线API({online_embed_models})提供的嵌入服务。"), + to_query: bool = Body(False, description="向量是否用于查询。有些模型如Minimax对存储/查询的向量进行了区分优化。"), ) -> BaseResponse: ''' 对文本进行向量化，返回 BaseResponse(data=List[List[float]]) @@ -78,9 +79,9 @@ def embed_texts_endpoint( def embed_documents( - docs: List[Document], - embed_model: str = EMBEDDING_MODEL, - to_query: bool = False, + docs: List[Document], + embed_model: str = EMBEDDING_MODEL, + to_query: bool = False, ) -> Dict: """ 将 List[Document] 向量化，转化为 VectorStore.add_embeddings 可以接受的参数 diff --git a/server/model_workers/base.py b/server/model_workers/base.py index 7b456a9..88affb4 100644 --- a/server/model_workers/base.py +++ b/server/model_workers/base.py @@ -113,6 +113,9 @@ class ApiModelWorker(BaseModelWorker): sys.stdout = sys.__stdout__ sys.stderr = sys.__stderr__ + new_loop = asyncio.new_event_loop() + asyncio.set_event_loop(new_loop) + self.context_len = context_len self.semaphore = asyncio.Semaphore(self.limit_worker_concurrency) self.version = None