feat(core): Support cross-service data recording and analyze (#665)

Close #659

docker/base/build_image.sh

@@ -15,7 +15,7 @@ IMAGE_NAME_ARGS=""
 PIP_INDEX_URL="https://pypi.org/simple"
 # en or zh
 LANGUAGE="en"
-BUILD_LOCAL_CODE="false"
+BUILD_LOCAL_CODE="true"
 LOAD_EXAMPLES="true"
 BUILD_NETWORK=""
 DB_GPT_INSTALL_MODEL="default"
@@ -26,7 +26,7 @@ usage () {
     echo "  [-n|--image-name image name] Current image name, default: db-gpt"
     echo "  [-i|--pip-index-url pip index url] Pip index url, default: https://pypi.org/simple"
     echo "  [--language en or zh] You language, default: en"
-    echo "  [--build-local-code true or false] Whether to use the local project code to package the image, default: false"
+    echo "  [--build-local-code true or false] Whether to use the local project code to package the image, default: true"
     echo "  [--load-examples true or false] Whether to load examples to default database default: true"
     echo "  [--network network name] The network of docker build"
     echo "  [--install-mode mode name] Installation mode name, default: default, If you completely use openai's service, you can set the mode name to 'openai'"

docs/index.rst

@@ -53,6 +53,7 @@ Getting Started
    getting_started/concepts.md
    getting_started/tutorials.md
    getting_started/faq.rst
+   getting_started/observability.md
 
 
 Modules

docs/locales/zh_CN/LC_MESSAGES/getting_started/observability.po

@@ -0,0 +1,121 @@
+# SOME DESCRIPTIVE TITLE.
+# Copyright (C) 2023, csunny
+# This file is distributed under the same license as the DB-GPT package.
+# FIRST AUTHOR <EMAIL@ADDRESS>, 2023.
+#
+#, fuzzy
+msgid ""
+msgstr ""
+"Project-Id-Version: DB-GPT 👏👏 0.3.9\n"
+"Report-Msgid-Bugs-To: \n"
+"POT-Creation-Date: 2023-10-12 11:54+0800\n"
+"PO-Revision-Date: YEAR-MO-DA HO:MI+ZONE\n"
+"Last-Translator: FULL NAME <EMAIL@ADDRESS>\n"
+"Language: zh_CN\n"
+"Language-Team: zh_CN <LL@li.org>\n"
+"Plural-Forms: nplurals=1; plural=0;\n"
+"MIME-Version: 1.0\n"
+"Content-Type: text/plain; charset=utf-8\n"
+"Content-Transfer-Encoding: 8bit\n"
+"Generated-By: Babel 2.12.1\n"
+
+#: ../../getting_started/observability.md:1 c88ce18295444597baa7355efc79ab15
+msgid "Debugging"
+msgstr ""
+
+#: ../../getting_started/observability.md:4 522a4e83c62c493381ca3a452ced8ecf
+msgid ""
+"DB-GPT provides a set of tools to help you troubleshoot and resolve some "
+"of the issues you may encounter."
+msgstr ""
+"DB-GPT 提供了一套工具来帮助你排查和解决一些遇到的问题。"
+
+#: ../../getting_started/observability.md:7 7fe5b8ab29194e42b74a8ab3e77006c7
+msgid "Trace Logs"
+msgstr "追踪日志"
+
+#: ../../getting_started/observability.md:9 8a55f7a2b5a247d49728969f179bf50d
+msgid ""
+"DB-GPT writes some critical system runtime information to trace logs. By "
+"default, these are located in `logs/dbgpt*.jsonl`."
+msgstr ""
+"DB-GPT 会将一些系统运行的关键的信息写入到追踪日志中，默认情况下，在 `logs/dbgpt*.jsonl` 中。"
+
+#: ../../getting_started/observability.md:11 5e2b847179e9427a8ae022b5338cbbd9
+msgid ""
+"DB-GPT also offers a command-line tool, `dbgpt trace`, to help you "
+"analyze these trace logs. You can see its specific usage with the command"
+" `dbgpt trace --help`."
+msgstr ""
+"DB-GPT 也提供了命令工具 `dbgpt trace` 命令来帮助你分析追踪日志，你可以使用命令 `dbgpt trace --help` 来查看具体的用法。"
+
+#: ../../getting_started/observability.md:14 6e0c9c4ba6ac4eb49d9289b0d63f77fb
+msgid "查看对话详情"
+msgstr ""
+
+#: ../../getting_started/observability.md:16 b9badbab74de47f192ff117d7d36fa72
+msgid ""
+"You can use the `dbgpt trace chat` command to view chat details. By "
+"default, it will display the latest chat message."
+msgstr ""
+"你可以使用 `dbgpt trace chat` 命令来查看对话信息，默认情况会显示你最新的一条对话信息。"
+
+#: ../../getting_started/observability.md:18 55c7466bd80d43c9a355d87daf2a2be7
+msgid "Viewing Service Runtime Information"
+msgstr "查看服务运行信息"
+
+#: ../../getting_started/observability.md:24
+#: ../../getting_started/observability.md:66
+#: ../../getting_started/observability.md:98
+#: ../../getting_started/observability.md:146
+#: ../../getting_started/observability.md:200
+#: ../../getting_started/observability.md:229
+#: ../../getting_started/observability.md:249
+#: ../../getting_started/observability.md:267 18907a58a0c3493aa24c17e367309471
+#: 387f08b6cd864a7682522b5a40863e79 5fe0baa5803d4ca5ad0e8cbd8a859c8c
+#: 7abbfdd996444999a24cbad852d2e545 957bdf6826e045608c8c3ebd06c8fe76
+#: acddf2cdf8c94bd6864d66739fa26459 f3fcaf2b47774779bad2feb3ef4318c4
+#: f99bb110024443f68cc8b7f19956eff4
+msgid "You will see an output like:"
+msgstr "你将会看到类似的输出："
+
+#: ../../getting_started/observability.md:60 5c8c213a5bac434bb3defe6611a03813
+msgid "Viewing the Latest Chat Message"
+msgstr "查看最近的一条对话信息"
+
+#: ../../getting_started/observability.md:92 ce19873d8e754173849d14eaeab963d2
+msgid "Viewing Chat Details and Call Chain"
+msgstr "查看对话信息和调用链路"
+
+#: ../../getting_started/observability.md:140 36bcc37971ce4d6682f1ea32e2e9a980
+msgid "Viewing Chat Details Based on trace_id"
+msgstr "根据 `trace_id` 查看对应的对话信息"
+
+#: ../../getting_started/observability.md:172 c74968492f7544758c9d95fa831c4fcf
+msgid "More `chat` Usage"
+msgstr "更多 `chat` 用法"
+
+#: ../../getting_started/observability.md:194 c2e5a7e7b1ee40fea15790d66b79eb11
+msgid "Viewing Call Tree Based on `trace_id`"
+msgstr "根据 `trace_id` 查看调用树"
+
+#: ../../getting_started/observability.md:220 6bd64d2ad0ce442e8e81aa1ae7dd2189
+msgid "Listing Trace Information"
+msgstr "列出追踪信息"
+
+#: ../../getting_started/observability.md:222 ce643441e8744ab09fcbd4081d2adb4a
+msgid "Listing All Trace Information"
+msgstr "列出全部追踪信息"
+
+#: ../../getting_started/observability.md:243 374376d81ed54bc2a450505abfe7dc6d
+msgid "Listing Trace Information by Trace Type"
+msgstr "根据追踪类型列出追踪信息"
+
+#: ../../getting_started/observability.md:261 8fc08a9c924d47309dc7062811c4fb62
+msgid "Searching Trace Information"
+msgstr "搜索追踪信息"
+
+#: ../../getting_started/observability.md:281 3681413d196144389431422010a7e30f
+msgid "More `list` Usage"
+msgstr "更多 `list` 用法"
+

pilot/component.py

@@ -47,6 +47,8 @@ class ComponentType(str, Enum):
     WORKER_MANAGER_FACTORY = "dbgpt_worker_manager_factory"
     MODEL_CONTROLLER = "dbgpt_model_controller"
     EXECUTOR_DEFAULT = "dbgpt_thread_pool_default"
+    TRACER = "dbgpt_tracer"
+    TRACER_SPAN_STORAGE = "dbgpt_tracer_span_storage"
 
 
 class BaseComponent(LifeCycle, ABC):
@@ -70,6 +72,8 @@ def init_app(self, system_app: SystemApp):
 
 T = TypeVar("T", bound=BaseComponent)
 
+_EMPTY_DEFAULT_COMPONENT = "_EMPTY_DEFAULT_COMPONENT"
+
 
 class SystemApp(LifeCycle):
     """Main System Application class that manages the lifecycle and registration of components."""
@@ -104,13 +108,18 @@ def register_instance(self, instance: T):
         instance.init_app(self)
 
     def get_component(
-        self, name: Union[str, ComponentType], component_type: Type[T]
+        self,
+        name: Union[str, ComponentType],
+        component_type: Type[T],
+        default_component=_EMPTY_DEFAULT_COMPONENT,
     ) -> T:
         """Retrieve a registered component by its name and type."""
         if isinstance(name, ComponentType):
             name = name.value
         component = self.components.get(name)
         if not component:
+            if default_component != _EMPTY_DEFAULT_COMPONENT:
+                return default_component
             raise ValueError(f"No component found with name {name}")
         if not isinstance(component, component_type):
             raise TypeError(f"Component {name} is not of type {component_type}")

pilot/model/adapter.py

@@ -354,7 +354,7 @@ def _parse_model_path(model_path: str) -> Tuple[bool, str]:
         if not path.is_file():
             model_paths = list(path.glob("*ggml*.gguf"))
             if not model_paths:
-                return False
+                return False, None
             model_path = str(model_paths[0])
             logger.warn(
                 f"Model path {model_path} is not single file, use first *gglm*.gguf model file: {model_path}"

pilot/model/base.py

@@ -53,6 +53,9 @@ class ModelOutput:
     error_code: int
     model_context: Dict = None
 
+    def to_dict(self) -> Dict:
+        return asdict(self)
+
 
 @dataclass
 class WorkerApplyOutput:

pilot/model/cluster/base.py

@@ -18,11 +18,13 @@ class PromptRequest(BaseModel):
     max_new_tokens: int = None
     stop: str = None
     echo: bool = True
+    span_id: str = None
 
 
 class EmbeddingsRequest(BaseModel):
     model: str
     input: List[str]
+    span_id: str = None
 
 
 class WorkerApplyRequest(BaseModel):

pilot/model/cluster/embedding/loader.py

@@ -3,6 +3,8 @@
 from typing import TYPE_CHECKING
 
 from pilot.model.parameter import BaseEmbeddingModelParameters
+from pilot.utils.parameter_utils import _get_dict_from_obj
+from pilot.utils.tracer import root_tracer, SpanType, SpanTypeRunName
 
 if TYPE_CHECKING:
     from langchain.embeddings.base import Embeddings
@@ -15,13 +17,21 @@ def __init__(self) -> None:
     def load(
         self, model_name: str, param: BaseEmbeddingModelParameters
     ) -> "Embeddings":
-        # add more models
-        if model_name in ["proxy_openai", "proxy_azure"]:
-            from langchain.embeddings import OpenAIEmbeddings
-
-            return OpenAIEmbeddings(**param.build_kwargs())
-        else:
-            from langchain.embeddings import HuggingFaceEmbeddings
-
-            kwargs = param.build_kwargs(model_name=param.model_path)
-            return HuggingFaceEmbeddings(**kwargs)
+        metadata = {
+            "model_name": model_name,
+            "run_service": SpanTypeRunName.EMBEDDING_MODEL.value,
+            "params": _get_dict_from_obj(param),
+        }
+        with root_tracer.start_span(
+            "EmbeddingLoader.load", span_type=SpanType.RUN, metadata=metadata
+        ):
+            # add more models
+            if model_name in ["proxy_openai", "proxy_azure"]:
+                from langchain.embeddings import OpenAIEmbeddings
+
+                return OpenAIEmbeddings(**param.build_kwargs())
+            else:
+                from langchain.embeddings import HuggingFaceEmbeddings
+
+                kwargs = param.build_kwargs(model_name=param.model_path)
+                return HuggingFaceEmbeddings(**kwargs)