"""用于跟踪运行的基本接口。"""
from __future__ import annotations
import logging
import sys
import traceback
from abc import ABC, abstractmethod
from datetime import datetime, timezone
from typing import (
TYPE_CHECKING,
Any,
Dict,
List,
Literal,
Optional,
Sequence,
Set,
Tuple,
Union,
cast,
)
from uuid import UUID
from tenacity import RetryCallState
from langchain_core.callbacks.base import BaseCallbackHandler
from langchain_core.exceptions import TracerException
from langchain_core.load import dumpd
from langchain_core.messages import BaseMessage
from langchain_core.outputs import (
ChatGeneration,
ChatGenerationChunk,
GenerationChunk,
LLMResult,
)
from langchain_core.tracers.schemas import Run
if TYPE_CHECKING:
from langchain_core.documents import Document
logger = logging.getLogger(__name__)
[docs]class BaseTracer(BaseCallbackHandler, ABC):
"""用于跟踪器的基本接口。"""
[docs] def __init__(
self,
*,
_schema_format: Literal["original", "streaming_events"] = "original",
**kwargs: Any,
) -> None:
"""初始化跟踪器。
参数：
_schema_format: 主要用于改变输入和输出的处理方式。仅供内部使用。此API将会更改。
- 'original' 是所有当前跟踪器使用的格式。
该格式在处理输入和输出方面略有不一致。
- 'streaming_events' 用于支持流事件，仅供内部使用。将来可能会更改，或者
完全弃用，转而使用专门的异步跟踪器来支持流事件。
kwargs: 将传递给超类的额外关键字参数。
"""
super().__init__(**kwargs)
self._schema_format = _schema_format # For internal use only API will change.
self.run_map: Dict[str, Run] = {}
"""Map of run ID to run. Cleared on run end."""
self.order_map: Dict[UUID, Tuple[UUID, str]] = {}
"""Map of run ID to (trace_id, dotted_order). Cleared when tracer GCed."""
@staticmethod
def _add_child_run(
parent_run: Run,
child_run: Run,
) -> None:
"""将子运行添加到链式运行或工具运行中。"""
parent_run.child_runs.append(child_run)
@abstractmethod
def _persist_run(self, run: Run) -> None:
"""持久化一个运行。"""
@staticmethod
def _get_stacktrace(error: BaseException) -> str:
"""获取父错误的堆栈跟踪。"""
msg = repr(error)
try:
if sys.version_info < (3, 10):
tb = traceback.format_exception(
error.__class__, error, error.__traceback__
)
else:
tb = traceback.format_exception(error)
return (msg + "\n\n".join(tb)).strip()
except: # noqa: E722
return msg
def _start_trace(self, run: Run) -> None:
"""开始一个运行的跟踪。"""
current_dotted_order = run.start_time.strftime("%Y%m%dT%H%M%S%fZ") + str(run.id)
if run.parent_run_id:
if parent := self.order_map.get(run.parent_run_id):
run.trace_id, run.dotted_order = parent
run.dotted_order += "." + current_dotted_order
if parent_run := self.run_map.get(str(run.parent_run_id)):
self._add_child_run(parent_run, run)
else:
logger.debug(
f"Parent run {run.parent_run_id} not found for run {run.id}."
" Treating as a root run."
)
run.parent_run_id = None
run.trace_id = run.id
run.dotted_order = current_dotted_order
else:
run.trace_id = run.id
run.dotted_order = current_dotted_order
self.order_map[run.id] = (run.trace_id, run.dotted_order)
self.run_map[str(run.id)] = run
self._on_run_create(run)
def _end_trace(self, run: Run) -> None:
"""结束一次运行的跟踪。"""
if not run.parent_run_id:
self._persist_run(run)
self.run_map.pop(str(run.id))
self._on_run_update(run)
def _get_run(
self, run_id: UUID, run_type: Union[str, Set[str], None] = None
) -> Run:
try:
run = self.run_map[str(run_id)]
except KeyError as exc:
raise TracerException(f"No indexed run ID {run_id}.") from exc
if isinstance(run_type, str):
run_types: Union[Set[str], None] = {run_type}
else:
run_types = run_type
if run_types is not None and run.run_type not in run_types:
raise TracerException(
f"Found {run.run_type} run at ID {run_id}, "
f"but expected {run_types} run."
)
return run
[docs] def on_chat_model_start(
self,
serialized: Dict[str, Any],
messages: List[List[BaseMessage]],
*,
run_id: UUID,
tags: Optional[List[str]] = None,
parent_run_id: Optional[UUID] = None,
metadata: Optional[Dict[str, Any]] = None,
name: Optional[str] = None,
**kwargs: Any,
) -> Run:
"""开始跟踪LLM运行。"""
if self._schema_format != "streaming_events":
# Please keep this un-implemented for backwards compatibility.
# When it's unimplemented old tracers that use the "original" format
# fallback on the on_llm_start method implementation if they
# find that the on_chat_model_start method is not implemented.
# This can eventually be cleaned up by writing a "modern" tracer
# that has all the updated schema changes corresponding to
# the "streaming_events" format.
raise NotImplementedError(
f"Chat model tracing is not supported in "
f"for {self._schema_format} format."
)
start_time = datetime.now(timezone.utc)
if metadata:
kwargs.update({"metadata": metadata})
chat_model_run = Run(
id=run_id,
parent_run_id=parent_run_id,
serialized=serialized,
inputs={"messages": [[dumpd(msg) for msg in batch] for batch in messages]},
extra=kwargs,
events=[{"name": "start", "time": start_time}],
start_time=start_time,
# WARNING: This is valid ONLY for streaming_events.
# run_type="llm" is what's used by virtually all tracers.
# Changing this to "chat_model" may break triggering on_llm_start
run_type="chat_model",
tags=tags,
name=name, # type: ignore[arg-type]
)
self._start_trace(chat_model_run)
self._on_chat_model_start(chat_model_run)
return chat_model_run
[docs] def on_llm_start(
self,
serialized: Dict[str, Any],
prompts: List[str],
*,
run_id: UUID,
tags: Optional[List[str]] = None,
parent_run_id: Optional[UUID] = None,
metadata: Optional[Dict[str, Any]] = None,
name: Optional[str] = None,
**kwargs: Any,
) -> Run:
"""开始跟踪LLM运行。"""
start_time = datetime.now(timezone.utc)
if metadata:
kwargs.update({"metadata": metadata})
llm_run = Run(
id=run_id,
parent_run_id=parent_run_id,
serialized=serialized,
# TODO: Figure out how to expose kwargs here
inputs={"prompts": prompts},
extra=kwargs,
events=[{"name": "start", "time": start_time}],
start_time=start_time,
run_type="llm",
tags=tags or [],
name=name, # type: ignore[arg-type]
)
self._start_trace(llm_run)
self._on_llm_start(llm_run)
return llm_run
[docs] def on_llm_new_token(
self,
token: str,
*,
chunk: Optional[Union[GenerationChunk, ChatGenerationChunk]] = None,
run_id: UUID,
parent_run_id: Optional[UUID] = None,
**kwargs: Any,
) -> Run:
"""在新的LLM令牌上运行。仅在启用流式传输时可用。"""
# "chat_model" is only used for the experimental new streaming_events format.
# This change should not affect any existing tracers.
llm_run = self._get_run(run_id, run_type={"llm", "chat_model"})
event_kwargs: Dict[str, Any] = {"token": token}
if chunk:
event_kwargs["chunk"] = chunk
llm_run.events.append(
{
"name": "new_token",
"time": datetime.now(timezone.utc),
"kwargs": event_kwargs,
},
)
self._on_llm_new_token(llm_run, token, chunk)
return llm_run
[docs] def on_retry(
self,
retry_state: RetryCallState,
*,
run_id: UUID,
**kwargs: Any,
) -> Run:
llm_run = self._get_run(run_id)
retry_d: Dict[str, Any] = {
"slept": retry_state.idle_for,
"attempt": retry_state.attempt_number,
}
if retry_state.outcome is None:
retry_d["outcome"] = "N/A"
elif retry_state.outcome.failed:
retry_d["outcome"] = "failed"
exception = retry_state.outcome.exception()
retry_d["exception"] = str(exception)
retry_d["exception_type"] = exception.__class__.__name__
else:
retry_d["outcome"] = "success"
retry_d["result"] = str(retry_state.outcome.result())
llm_run.events.append(
{
"name": "retry",
"time": datetime.now(timezone.utc),
"kwargs": retry_d,
},
)
return llm_run
[docs] def on_llm_end(self, response: LLMResult, *, run_id: UUID, **kwargs: Any) -> Run:
"""结束LLM运行的跟踪。"""
# "chat_model" is only used for the experimental new streaming_events format.
# This change should not affect any existing tracers.
llm_run = self._get_run(run_id, run_type={"llm", "chat_model"})
llm_run.outputs = response.dict()
for i, generations in enumerate(response.generations):
for j, generation in enumerate(generations):
output_generation = llm_run.outputs["generations"][i][j]
if "message" in output_generation:
output_generation["message"] = dumpd(
cast(ChatGeneration, generation).message
)
llm_run.end_time = datetime.now(timezone.utc)
llm_run.events.append({"name": "end", "time": llm_run.end_time})
self._end_trace(llm_run)
self._on_llm_end(llm_run)
return llm_run
[docs] def on_llm_error(
self,
error: BaseException,
*,
run_id: UUID,
**kwargs: Any,
) -> Run:
"""处理LLM运行的错误。"""
# "chat_model" is only used for the experimental new streaming_events format.
# This change should not affect any existing tracers.
llm_run = self._get_run(run_id, run_type={"llm", "chat_model"})
llm_run.error = self._get_stacktrace(error)
llm_run.end_time = datetime.now(timezone.utc)
llm_run.events.append({"name": "error", "time": llm_run.end_time})
self._end_trace(llm_run)
self._on_llm_error(llm_run)
return llm_run
[docs] def on_chain_start(
self,
serialized: Dict[str, Any],
inputs: Dict[str, Any],
*,
run_id: UUID,
tags: Optional[List[str]] = None,
parent_run_id: Optional[UUID] = None,
metadata: Optional[Dict[str, Any]] = None,
run_type: Optional[str] = None,
name: Optional[str] = None,
**kwargs: Any,
) -> Run:
"""开始跟踪链运行。"""
start_time = datetime.now(timezone.utc)
if metadata:
kwargs.update({"metadata": metadata})
chain_run = Run(
id=run_id,
parent_run_id=parent_run_id,
serialized=serialized,
inputs=self._get_chain_inputs(inputs),
extra=kwargs,
events=[{"name": "start", "time": start_time}],
start_time=start_time,
child_runs=[],
run_type=run_type or "chain",
name=name, # type: ignore[arg-type]
tags=tags or [],
)
self._start_trace(chain_run)
self._on_chain_start(chain_run)
return chain_run
def _get_chain_inputs(self, inputs: Any) -> Any:
"""获取链式运行的输入。"""
if self._schema_format == "original":
return inputs if isinstance(inputs, dict) else {"input": inputs}
elif self._schema_format == "streaming_events":
return {
"input": inputs,
}
else:
raise ValueError(f"Invalid format: {self._schema_format}")
def _get_chain_outputs(self, outputs: Any) -> Any:
"""获取链式运行的输出。"""
if self._schema_format == "original":
return outputs if isinstance(outputs, dict) else {"output": outputs}
elif self._schema_format == "streaming_events":
return {
"output": outputs,
}
else:
raise ValueError(f"Invalid format: {self._schema_format}")
[docs] def on_chain_end(
self,
outputs: Dict[str, Any],
*,
run_id: UUID,
inputs: Optional[Dict[str, Any]] = None,
**kwargs: Any,
) -> Run:
"""结束链式运行的跟踪。"""
chain_run = self._get_run(run_id)
chain_run.outputs = self._get_chain_outputs(outputs)
chain_run.end_time = datetime.now(timezone.utc)
chain_run.events.append({"name": "end", "time": chain_run.end_time})
if inputs is not None:
chain_run.inputs = self._get_chain_inputs(inputs)
self._end_trace(chain_run)
self._on_chain_end(chain_run)
return chain_run
[docs] def on_chain_error(
self,
error: BaseException,
*,
inputs: Optional[Dict[str, Any]] = None,
run_id: UUID,
**kwargs: Any,
) -> Run:
"""处理链式运行的错误。"""
chain_run = self._get_run(run_id)
chain_run.error = self._get_stacktrace(error)
chain_run.end_time = datetime.now(timezone.utc)
chain_run.events.append({"name": "error", "time": chain_run.end_time})
if inputs is not None:
chain_run.inputs = self._get_chain_inputs(inputs)
self._end_trace(chain_run)
self._on_chain_error(chain_run)
return chain_run
[docs] def on_retriever_start(
self,
serialized: Dict[str, Any],
query: str,
*,
run_id: UUID,
parent_run_id: Optional[UUID] = None,
tags: Optional[List[str]] = None,
metadata: Optional[Dict[str, Any]] = None,
name: Optional[str] = None,
**kwargs: Any,
) -> Run:
"""当Retriever开始运行时运行。"""
start_time = datetime.now(timezone.utc)
if metadata:
kwargs.update({"metadata": metadata})
retrieval_run = Run(
id=run_id,
name=name or "Retriever",
parent_run_id=parent_run_id,
serialized=serialized,
inputs={"query": query},
extra=kwargs,
events=[{"name": "start", "time": start_time}],
start_time=start_time,
tags=tags,
child_runs=[],
run_type="retriever",
)
self._start_trace(retrieval_run)
self._on_retriever_start(retrieval_run)
return retrieval_run
[docs] def on_retriever_error(
self,
error: BaseException,
*,
run_id: UUID,
**kwargs: Any,
) -> Run:
"""当Retriever发生错误时运行。"""
retrieval_run = self._get_run(run_id, run_type="retriever")
retrieval_run.error = self._get_stacktrace(error)
retrieval_run.end_time = datetime.now(timezone.utc)
retrieval_run.events.append({"name": "error", "time": retrieval_run.end_time})
self._end_trace(retrieval_run)
self._on_retriever_error(retrieval_run)
return retrieval_run
[docs] def on_retriever_end(
self, documents: Sequence[Document], *, run_id: UUID, **kwargs: Any
) -> Run:
"""当Retriever运行结束时运行。"""
retrieval_run = self._get_run(run_id, run_type="retriever")
retrieval_run.outputs = {"documents": documents}
retrieval_run.end_time = datetime.now(timezone.utc)
retrieval_run.events.append({"name": "end", "time": retrieval_run.end_time})
self._end_trace(retrieval_run)
self._on_retriever_end(retrieval_run)
return retrieval_run
def __deepcopy__(self, memo: dict) -> BaseTracer:
"""深度复制追踪器。"""
return self
def __copy__(self) -> BaseTracer:
"""复制跟踪器。"""
return self
def _on_run_create(self, run: Run) -> None:
"""创建时处理一个运行。"""
def _on_run_update(self, run: Run) -> None:
"""处理更新时的运行。"""
def _on_llm_start(self, run: Run) -> None:
"""处理LLM运行时的启动。"""
def _on_llm_new_token(
self,
run: Run,
token: str,
chunk: Optional[Union[GenerationChunk, ChatGenerationChunk]],
) -> None:
"""处理新的LLM令牌。"""
def _on_llm_end(self, run: Run) -> None:
"""处理LLM运行。"""
def _on_llm_error(self, run: Run) -> None:
"""处理LLM运行中的错误。"""
def _on_chain_start(self, run: Run) -> None:
"""处理链式运行的开始。"""
def _on_chain_end(self, run: Run) -> None:
"""处理链式运行。"""
def _on_chain_error(self, run: Run) -> None:
"""处理链式运行中的错误。"""
def _on_tool_start(self, run: Run) -> None:
"""处理工具启动时的运行。"""
def _on_tool_end(self, run: Run) -> None:
"""处理工具运行。"""
def _on_tool_error(self, run: Run) -> None:
"""处理工具运行时的错误。"""
def _on_chat_model_start(self, run: Run) -> None:
"""处理启动时的聊天模型运行。"""
def _on_retriever_start(self, run: Run) -> None:
"""处理检索器运行时的启动。"""
def _on_retriever_end(self, run: Run) -> None:
"""处理Retriever运行。"""
def _on_retriever_error(self, run: Run) -> None:
"""处理检索器运行中的错误。"""