修改获取文件名称方法

backend/app/engine/loaders/projectJson.py

@@ -55,6 +55,7 @@ class JsonTable:
     def comment(self):
         return self._comment
 
+
 class ProjectJson:
     def __init__(self,dir:str) -> None:
         self._dir = dir
@@ -75,5 +76,19 @@ class ProjectJson:
     def tables(self):
         return self._tables
 
+def getProjectName(dir:str):
+    result = dir.split('\\')
+    if len(result) > 0:
+        return result[-1]
+    return "未知工程名称"
 
+    prjJson = ProjectJson(dir)
+    prjJson.parse()
+    tb:JsonTable = prjJson.table('工程属性')
+    records = tb.records()
+    for record in records:
+        name = record.value('名称') 
+        if name == '工程名称':
+            return record.value('值') 
+    return ''
 

backend/app/engine/prompt.py

@@ -2,36 +2,29 @@ from llama_index.core import PromptTemplate
 
 text_qa_template_str = (
     "# 角色\n"
-    "你是一名博微造价工程数据查询助手，专精于电力工程文件中的信息。"
-    "你的职责是提供有关电力造价、造价编制软件、文件结构及相关数据的精准、客观的回答，"
-    "如同直接从文件中提取的内容。\n"
-    "知识库中已经导入一个工程的全部数据，请你站在当前工程的角度回答用户关于工程文件的问题。\n"
-    "例如：询问“此工程”指当前导入的工程。询问“此工程名称”指当前导入的工程的工程名称。\n"
-    
-    "## 技能\n"
-    "### 技能 1: 数据查询与提供\n"
-    "- 准确回答所有关于电力工程造价的相关问题。\n"
-    "- 提供具体数据，如成本估算、材料清单、劳动力需求等。\n"
-    "- 确保提供的信息严格基于工程文档中的记录。\n"
-    
-    "### 技能 2: 技术性解释\n"
-    "- 解释造价工程中的技术术语和概念。\n"
-    "- 为复杂的工程细节提供清晰易懂的说明。\n"
-    
-    "## 约束\n"
-    "- 仅回答与电力工程造价文件相关的具体问题。\n"
-    "- 不进行任何超出文件内容的猜测或假设。\n"
-    "- 所有回答均基于文件内容，采用客观和技术性的语言。\n"
-    "- 请基于这些信息回答问题。如果无法找到相关信息，请不要额外发散回答，不要回答多余的信息，只需要回答“我不知道这个问题的答案”。\n"
-    "以下为上下文信息\n"
+    "你是一名擅长博微造价工程数据问答的专家，可以根据电力工程文件中的内容回答用户问题。\n"
+    "\n"
+    "# 任务描述：\n"
+    "请仔细阅读所给的文档片段，并根据其内容回答问题。\n"
+    "您需要判断文档的内容是否可以回答问题，不要强行回答。如果可以回答，答案必须严格遵循文档内容，即使与事实不符。\n"
+    "如果答案与事实不符，直接给出答案，不要做解释。\n"
+    "\n"
+    "# 回答规则：\n"
+    "- 请使用与文档材料相同的语言回答问题。\n"
+    "- 评估文档是否含有足够信息回答问题。无关时不要回答。\n"
+    "- 如果问题能被回答，你的回答必须严格遵循文档内容，即使与事实不符。一定不要做多余解释。\n"
+    "- 如果问题能被回答，直接引用文档的相关信息保证答案准确、完整，并追求简洁。\n"
+    "- 当文档中只有少量信息与问题相关时，重点关注这部分信息，这种情况下一定回答。\n"
+    "- 当文档中信息与问题无关时，请不要额外发散回答，只需要回答“我不知道这个问题的答案”。\n"
+    "\n"
+    "来自多个来源的文档片段如下,请充分理解以下参考资料内容，组织出满足用户提问的条理清晰的回复。\n"
     "---------------------\n"
     "{context_str}\n"
     "---------------------\n"
-    "请根据上下文信息而非先前知识回答我的问题或回复我的指令。前面的上下文信息可能有用，也可能没用，你需要从我给出的上下文信息中选出与我的问题最相关的那些，来为你的回答提供依据。回答一定要忠于原文，简洁但不丢信息，不要胡乱编造。如果无法找到相关信息，请不要额外发散回答，不要回答多余的信息，只需要回答“我不知道这个问题的答案”。我的问题或指令是什么语种，你就用什么语种回复。\n"
+    "鉴于来自多个来源的文档片段而非先验知识，回答查询。\n"
     "如果是表结构或者是数据库的相关内容，只用于推导问题，不需要告诉用户数据库或表结构等物理信息。\n"
-
-    "问题：{query_str}\n"
-    "你的回复： "
+    "Query: {query_str}\n"
+    "Answer: "
 )
 
 text_qa_template = PromptTemplate(text_qa_template_str)
@@ -57,31 +50,26 @@ refine_template = PromptTemplate(refine_template_str)
 
 summary_template_str = (
     "# 角色\n"
-    "你是一名博微造价工程数据查询助手，专精于电力工程文件中的信息。"
-    "你的职责是提供有关电力造价、造价编制软件、文件结构及相关数据的精准、客观的回答，"
-    "如同直接从文件中提取的内容。\n"
-    
-    "## 技能\n"
-    "### 技能 1: 数据查询与提供\n"
-    "- 准确回答所有关于电力工程造价的相关问题。\n"
-    "- 提供具体数据，如成本估算、材料清单、劳动力需求等。\n"
-    "- 确保提供的信息严格基于工程文档中的记录。\n"
-    
-    "### 技能 2: 技术性解释\n"
-    "- 解释造价工程中的技术术语和概念。\n"
-    "- 为复杂的工程细节提供清晰易懂的说明。\n"
-    
-    "## 约束\n"
-    "- 仅回答与电力工程造价文件相关的具体问题。\n"
-    "- 不进行任何超出文件内容的猜测或假设。\n"
-    "- 所有回答均基于文件内容，采用客观和技术性的语言。\n"
-    "- 请基于这些信息回答问题。如果无法找到相关信息，请不要额外发散回答，不要回答多余的信息，只需要回答“我不知道这个问题的答案”。\n"
-    "来自多个来源的上下文信息如下。\n"
+    "你是一名擅长博微造价工程数据问答的专家，可以根据电力工程文件中的内容回答用户问题。\n"
+    "\n"
+    "# 任务描述：\n"
+    "请仔细阅读所给的文档片段，并根据其内容回答问题。\n"
+    "您需要判断文档的内容是否可以回答问题，不要强行回答。如果可以回答，答案必须严格遵循文档内容，即使与事实不符。\n"
+    "如果答案与事实不符，直接给出答案，不要做解释。\n"
+    "\n"
+    "# 回答规则：\n"
+    "- 请使用与文档材料相同的语言回答问题。\n"
+    "- 评估文档是否含有足够信息回答问题。无关时不要回答。\n"
+    "- 如果问题能被回答，你的回答必须严格遵循文档内容，即使与事实不符。一定不要做多余解释。\n"
+    "- 如果问题能被回答，直接引用文档的相关信息保证答案准确、完整，并追求简洁。\n"
+    "- 当文档中只有少量信息与问题相关时，重点关注这部分信息，这种情况下一定回答。\n"
+    "- 当文档中信息与问题无关时，请不要额外发散回答，只需要回答“我不知道这个问题的答案”。\n"
+    "\n"
+    "来自多个来源的文档片段如下,请充分理解以下参考资料内容，组织出满足用户提问的条理清晰的回复。\n"
     "---------------------\n"
     "{context_str}\n"
     "---------------------\n"
-    "鉴于来自多个来源的信息而非先验知识， "
-    "回答查询。\n"
+    "鉴于来自多个来源的文档片段而非先验知识，回答查询。\n"
     "如果是表结构或者是数据库的相关内容，只用于推导问题，不需要告诉用户数据库或表结构等物理信息。\n"
     "Query: {query_str}\n"
     "Answer: "
@@ -108,9 +96,6 @@ ReActChatFormatter_messages = (
     "Action Input: 输入给工具的内容，使用JSON格式表示kwargs（例如{{\"input\": \"hello world\", \"num_beams\": 5}}）\n"
     "'''\n\n"
     "请始终以Thought开始。\n\n"
-    "请始终以Thought开始。\n\n"
-    "请始终以Thought开始。\n\n"
-    "请始终以Thought开始。\n\n"
     "切勿用Markdown代码标记包围你的响应。如果需要，可以在响应中使用代码标记。\n\n"
     "请为Action Input使用有效的JSON格式。不要这样做{{\'input\': \'hello world\', \'num_beams\': 5}}。\n\n"
     "如果使用此格式，用户将以下面的格式进行回应：\n\n"

backend/app/settings.py

@@ -3,15 +3,18 @@ from typing import Dict
 from abc import abstractmethod
 from llama_index.core.constants import DEFAULT_TEMPERATURE
 from llama_index.core.settings import Settings
+from llama_index.embeddings.xinference import XinferenceEmbedding
 from llama_index.llms.xinference import Xinference
+#from llama_index.embeddings.xinference import XinferenceEmbedding
 from llama_index.llms.xinference.base import DEFAULT_XINFERENCE_TEMP
+from llama_index.postprocessor.xinference_rerank import XinferenceRerank
 
-from app.xinference.base import XinferenceEmbedding, XinferenceRerank
 from app.engine.loaders import getProjectInfos
 from app.api.routers.request.base import ProjectInfo
+from modelProvide.customDashScope import CustomDashScope
 from util.register import *
 from llama_index.core.callbacks import CallbackManager
-from modelProvide.customDashScope import CustomDashScope
+
 
 ModelPlateCategory = '模型平台'
 
@@ -103,7 +106,7 @@ class XinferencePlatform(ModelPlatform):
         embed_model_name = os.getenv("EMBEDDING_MODEL")
         dimensions = os.getenv("EMBEDDING_DIM")
         dimensions = int(dimensions) if dimensions is not None else None
-        return XinferenceEmbedding(embed_model_name, embedding_base_url, dimensions=dimensions)
+        return XinferenceEmbedding(embed_model_name, embedding_base_url)
     
     def rerank(self):
         rerank_model = os.getenv("RERANK_MODEL")

backend/app/xinference/base.py

@@ -1,272 +0,0 @@
-"""Xinference embeddings file."""
-
-import logging
-from enum import Enum
-from http import HTTPStatus
-from typing import Any, Dict, List, Optional, Union, Tuple
-
-from llama_index.core.base.embeddings.base import BaseEmbedding, Embedding, dispatcher
-from llama_index.core.bridge.pydantic import PrivateAttr
-from llama_index.core.callbacks import CBEventType, EventPayload
-from llama_index.core.embeddings.multi_modal_base import MultiModalEmbedding
-from llama_index.core.instrumentation.events.rerank import ReRankStartEvent, ReRankEndEvent
-from llama_index.core.postprocessor.types import BaseNodePostprocessor
-from llama_index.core.schema import ImageType, NodeWithScore, QueryBundle
-from pydantic import Field
-
-logger = logging.getLogger(__name__)
-
-
-EMBED_MAX_INPUT_LENGTH = 2048
-EMBED_MAX_BATCH_SIZE = 1
-
-
-class XinferenceEmbedding(BaseEmbedding):
-    """Xinference class for text embedding.
-
-     """
-    model_description: Dict[str, Any] = Field(
-        description="The model description from Xinference."
-    )
-    _generator: Any = PrivateAttr()
-    _model_uid: str = Field(description="The Xinference model to use.")
-    _endpoint: str = Field(description="The Xinference endpoint URL to use.")
-
-    def __init__(
-            self,
-            model_uid: str,
-            endpoint: str,
-            embed_batch_size: int = EMBED_MAX_BATCH_SIZE,
-            dimensions: Optional[int] = None,
-            additional_kwargs: Optional[Dict[str, Any]] = None,
-            api_key: Optional[str] = None,
-            api_base: Optional[str] = None,
-            api_version: Optional[str] = None,
-            max_retries: int = 10,
-            # timeout: float = 60.0,
-            # reuse_client: bool = True,
-            # callback_manager: Optional[CallbackManager] = None,
-            # default_headers: Optional[Dict[str, str]] = None,
-            # http_client: Optional[httpx.Client] = None,
-            # async_http_client: Optional[httpx.AsyncClient] = None,
-            # num_workers: Optional[int] = None,
-            **kwargs: Any,
-    ) -> None:
-        generator, model_description, embed_batch_size, dimensions = self.load_model(
-                model_uid, endpoint
-            )
-        self._generator = generator
-        #self._model_uid = model_uid
-        #self._endpoint = endpoint
-        super().__init__(
-            embed_batch_size=embed_batch_size,
-            dimensions=dimensions,
-            #callback_manager=callback_manager,
-            model_name=model_uid,
-            additional_kwargs=additional_kwargs,
-            api_key=api_key,
-            api_base=api_base,
-            api_version=api_version,
-            max_retries=max_retries,
-            # reuse_client=reuse_client,
-            # timeout=timeout,
-            # default_headers=default_headers,
-            # num_workers=num_workers,
-            **kwargs,
-        )
-
-    def load_model(self, model_uid: str, endpoint: str) -> Tuple[Any, int, dict]:
-        try:
-            from xinference.client import RESTfulClient
-        except ImportError:
-            raise ImportError(
-                "Could not import Xinference library."
-                'Please install Xinference with `pip install "xinference[all]"`'
-            )
-
-        client = RESTfulClient(endpoint)
-
-        try:
-            assert isinstance(client, RESTfulClient)
-        except AssertionError:
-            raise RuntimeError(
-                "Could not create RESTfulClient instance."
-                "Please make sure Xinference endpoint is running at the correct port."
-            )
-
-        generator = client.get_model(model_uid)
-        model_description = client.list_models()[model_uid]
-
-        try:
-            assert generator is not None
-            assert model_description is not None
-        except AssertionError:
-            raise RuntimeError(
-                "Could not get model from endpoint."
-                "Please make sure Xinference endpoint is running at the correct port."
-            )
-
-        model = model_description["model_name"]
-        replica = model_description['replica']
-        dimensions = model_description['dimensions']
-        max_tokens = model_description['max_tokens']
-
-        return generator, model_description, replica, dimensions
-
-    @classmethod
-    def class_name(cls) -> str:
-        return "XinferenceEmbedding"
-
-    def _get_text_embedding(self, text: str) -> Embedding:
-        """
-        Embed the input text synchronously.
-
-        Subclasses should implement this method. Reference get_text_embedding's
-        docstring for more information.
-        """
-        assert self._generator is not None
-
-        response = self._generator.create_embedding(input=text)
-        return response['data'][0]['embedding']
-
-    def _get_query_embedding(self, query: str) -> Embedding:
-        """
-        Embed the input query synchronously.
-
-        Subclasses should implement this method. Reference get_query_embedding's
-        docstring for more information.
-        """
-        return self._get_text_embedding(query)
-
-    async def _aget_query_embedding(self, query: str) -> Embedding:
-        """
-        Embed the input query asynchronously.
-
-        Subclasses should implement this method. Reference get_query_embedding's
-        docstring for more information.
-        """
-        return self._get_query_embedding(query)
-
-class XinferenceRerank(BaseNodePostprocessor):
-    """Xinference class for rerank.
-
-     """
-    model_description: Dict[str, Any] = Field(
-        description="The model description from Xinference."
-    )
-    _generator: Any = PrivateAttr()
-    _model_uid: str = Field(description="The Xinference model to use.")
-    _endpoint: str = Field(description="The Xinference endpoint URL to use.")
-    model: str = Field(description="Dashscope rerank model name.")
-    top_n: int = Field(description="Top N nodes to return.")
-    threshold: float = Field(description="threshold nodes to return.")
-
-    def __init__(
-            self,
-            model_uid: str,
-            endpoint: str,
-            top_n: int = None,
-            threshold: float = None,
-            return_documents: bool = False
-    ):
-        _model_uid = model_uid
-        _endpoint = endpoint
-        _op_n = top_n
-        threshold = threshold
-        generator, model_description = self.load_model(
-            model_uid, endpoint
-        )
-        self._generator = generator
-        super().__init__(top_n=top_n, model=model_uid, model_uid=model_uid, threshold = threshold, return_documents=return_documents)
-
-    @classmethod
-    def class_name(cls) -> str:
-        return "XinferenceRerank"
-
-    def _postprocess_nodes(
-            self,
-            nodes: List[NodeWithScore],
-            query_bundle: Optional[QueryBundle] = None,
-    ) -> List[NodeWithScore]:
-        if query_bundle is None:
-            raise ValueError("Missing query bundle in extra info.")
-        if len(nodes) == 0:
-            return []
-
-        dispatcher.event(
-            ReRankStartEvent(
-                nodes = nodes,
-                top_n = self.top_n,
-                query = query_bundle,
-                model_name = self.model
-            )
-        )
-
-        with self.callback_manager.event(
-                CBEventType.RERANKING,
-                payload={
-                    EventPayload.NODES: nodes,
-                    EventPayload.MODEL_NAME: self._model_uid,
-                    EventPayload.QUERY_STR: query_bundle.query_str,
-                    EventPayload.TOP_K: self.top_n,
-                },
-        ) as event:
-            texts = [node.node.get_content() for node in nodes]
-            response = self._generator.rerank(texts,query_bundle.query_str)
-            new_nodes = []
-            for result in response['results']:
-                new_node_with_score = NodeWithScore(
-                    node=nodes[result['index']].node, score=result['relevance_score']
-                )
-                if self.threshold is not None:
-                    if new_node_with_score.score >=self.threshold:
-                        new_nodes.append(new_node_with_score)
-
-                if self.top_n is not None:
-                    if len(new_nodes) > self.top_n:
-                        for index in new_nodes[self.top_n:-1]:
-                            new_nodes.remove(index)
-
-            event.on_end(payload={EventPayload.NODES: new_nodes})
-
-        dispatcher.event(
-            ReRankEndEvent(
-                nodes= new_nodes
-            )
-        )
-        return new_nodes
-
-    def load_model(self, model_uid: str, endpoint: str) -> Tuple[Any, int, dict]:
-        try:
-            from xinference.client import RESTfulClient
-        except ImportError:
-            raise ImportError(
-                "Could not import Xinference library."
-                'Please install Xinference with `pip install "xinference[all]"`'
-            )
-
-        client = RESTfulClient(endpoint)
-
-        try:
-            assert isinstance(client, RESTfulClient)
-        except AssertionError:
-            raise RuntimeError(
-                "Could not create RESTfulClient instance."
-                "Please make sure Xinference endpoint is running at the correct port."
-            )
-
-        generator = client.get_model(model_uid)
-        model_description = client.list_models()[model_uid]
-
-        try:
-            assert generator is not None
-            assert model_description is not None
-        except AssertionError:
-            raise RuntimeError(
-                "Could not get model from endpoint."
-                "Please make sure Xinference endpoint is running at the correct port."
-            )
-
-        model = model_description["model_name"]
-
-        return generator, model_description

backend/pyproject.toml

@@ -10,46 +10,54 @@ readme = "README.md"
 generate = "app.engine.generate:generate_datasource"
 
 [tool.poetry.dependencies]
-python = "^3.11,<3.12"
-fastapi = "^0.110.3"
-python-dotenv = "^1.0.0"
+python = "^3.11,<3.13"
+fastapi = "0.110.3"
+python-dotenv = "^1.0.1"
 aiostream = "^0.6.2"
-llama-index = "0.10.63"
-cachetools = "^5.3.3"
+cachetools = "^5.5.0"
 protobuf = "4.25.4"
 nltk = "^3.9.1"
 jieba = "^0.42.1"
 
-#arize-phoenix = "^4.12.0"
-openinference-instrumentation-llama-index="2.2.3"
-llama-index-callbacks-arize-phoenix = "^0.1.4"
-llama-index-llms-dashscope = "^0.1.2"
-llama-index-embeddings-dashscope = "^0.1.4"
-llama-index-postprocessor-dashscope-rerank-custom = "0.1.0"
-xinference  = "^0.14.1"
-xinference-client  = "^0.14.1"
-llama-index-llms-xinference = "^0.1.2"
-qdrant-client="^1.10.1"
-llama-index-vector-stores-qdrant = "^0.2.14"
-chroma="^0.2.0"
-llama-index-vector-stores-chroma = "^0.1.10"
-llama-index-readers-json = "^0.1.5"
-llama-index-retrievers-bm25 = "^0.2.2"
-llama-index-experimental = "^0.1.4"
-llama-index-llms-ollama = "^0.1.6"
-llama-index-embeddings-ollama = "^0.1.3"
+transformers = "^4.43.0"
 
-duckduckgo_search = "^6.2.6"
+#arize-phoenix = "^4.12.0"
+openinference-instrumentation-llama-index="^3.0.2"
+llama-index = "^0.11.7"
+llama-index-core = "^0.11.7"
+llama-index-callbacks-arize-phoenix = "^0.2.1"
+llama-index-llms-dashscope = "^0.2.0"
+llama-index-embeddings-dashscope = "^0.2.1"
+#llama-index-postprocessor-dashscope-rerank = "^0.2.0"
+
+llama-index-llms-ollama = "^0.3.1"
+llama-index-embeddings-ollama = "^0.3.0"
+
+xinference = "^0.15.0"
+xinference-client = "^0.15.0"
+llama-index-llms-xinference = "^0.2.1"
+llama-index-embeddings-xinference = "^0.1.0"
+llama-index-postprocessor-xinference-rerank = "^0.1.0"
+qdrant-client="^1.11.0"
+llama-index-vector-stores-qdrant = "^0.3.0"
+chroma="^0.2.0"
+llama-index-vector-stores-chroma = "^0.2.0"
+llama-index-readers-json = "^0.2.0"
+llama-index-retrievers-bm25 = "^0.3.0"
+llama-index-experimental = "^0.3.0"
+
+
+duckduckgo_search = "^6.2.10"
 
 [tool.poetry.dependencies.uvicorn]
 extras = [ "standard" ]
-version = "^0.23.2"
+version = "^0.30.6"
 
 [tool.poetry.dependencies.llama-index-readers-database]
-version = "^0.1.3"
+version = "^0.2.0"
 
 [tool.poetry.dependencies.pymysql]
-version = "^1.1.0"
+version = "^1.1.1"
 extras = [ "rsa" ]
 
 #[tool.poetry.dependencies.psycopg2]
@@ -62,7 +70,8 @@ extras = [ "rsa" ]
 version = "^0.8"
 
 [tool.poetry.dependencies.e2b_code_interpreter]
-version = "0.0.7"
+version = "^0.0.7"
+
 
 
 [[tool.poetry.source]]
@@ -82,6 +91,7 @@ name = "tsinghua"
 url = "https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple/"
 priority = "primary"
 
+
 [build-system]
 requires = [ "poetry-core" ]
 build-backend = "poetry.core.masonry.api"