-
-
Notifications
You must be signed in to change notification settings - Fork 473
/
langchain.py
180 lines (147 loc) · 6.46 KB
/
langchain.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
"""The langchain module integrates Langchain support with Panel."""
from __future__ import annotations
from typing import Any, Union
try:
from langchain.callbacks.base import BaseCallbackHandler
from langchain.schema import AgentAction, AgentFinish, LLMResult
except ImportError:
BaseCallbackHandler = object
AgentAction = None
AgentFinish = None
LLMResult = None
from ..chat.feed import ChatFeed
from ..chat.interface import ChatInterface
from ..chat.message import DEFAULT_AVATARS
from ..layout import Accordion
class PanelCallbackHandler(BaseCallbackHandler):
"""
The Langchain `PanelCallbackHandler` itself is not a widget or pane, but is useful for rendering
and streaming the *chain of thought* from Langchain Tools, Agents, and Chains
as `ChatMessage` objects.
Reference: https://panel.holoviz.org/reference/chat/PanelCallbackHandler.html
:Example:
>>> chat_interface = pn.widgets.ChatInterface(callback=callback, callback_user="Langchain")
>>> callback_handler = pn.widgets.langchain.PanelCallbackHandler(instance=chat_interface)
>>> llm = ChatOpenAI(streaming=True, callbacks=[callback_handler])
>>> chain = ConversationChain(llm=llm)
"""
def __init__(
self,
instance: ChatFeed | ChatInterface,
user: str = "LangChain",
avatar: str = DEFAULT_AVATARS["langchain"],
):
if BaseCallbackHandler is object:
raise ImportError(
"LangChainCallbackHandler requires `langchain` to be installed."
)
self.instance = instance
self._message = None
self._active_user = user
self._active_avatar = avatar
self._disabled_state = self.instance.disabled
self._is_streaming = None
self._input_user = user # original user
self._input_avatar = avatar
def _update_active(self, avatar: str, label: str):
"""
Prevent duplicate labels from being appended to the same user.
"""
# not a typo; Langchain passes a string :/
if label == "None":
return
self._active_avatar = avatar
if f"- {label}" not in self._active_user:
self._active_user = f"{self._active_user} - {label}"
def _reset_active(self):
self._active_user = self._input_user
self._active_avatar = self._input_avatar
self._message = None
def _on_start(self, serialized, kwargs):
model = kwargs.get("invocation_params", {}).get("model_name", "")
self._is_streaming = serialized.get("kwargs", {}).get("streaming")
messages = self.instance.objects
if messages[-1].user != self._active_user:
self._message = None
if self._active_user and model not in self._active_user:
self._active_user = f"{self._active_user} ({model})"
def _stream(self, message: str):
if message:
return self.instance.stream(
message,
user=self._active_user,
avatar=self._active_avatar,
message=self._message,
)
return self._message
def on_llm_start(self, serialized: dict[str, Any], *args, **kwargs):
self._on_start(serialized, kwargs)
return super().on_llm_start(serialized, *args, **kwargs)
def on_llm_new_token(self, token: str, **kwargs) -> None:
self._message = self._stream(token)
return super().on_llm_new_token(token, **kwargs)
def on_llm_end(self, response: LLMResult, *args, **kwargs):
if not self._is_streaming:
# on_llm_new_token does not get called if not streaming
self._stream(response.generations[0][0].text)
self._reset_active()
return super().on_llm_end(response, *args, **kwargs)
def on_llm_error(self, error: Union[Exception, KeyboardInterrupt], *args, **kwargs):
return super().on_llm_error(error, *args, **kwargs)
def on_agent_action(self, action: AgentAction, *args, **kwargs: Any) -> Any:
return super().on_agent_action(action, *args, **kwargs)
def on_agent_finish(self, finish: AgentFinish, *args, **kwargs: Any) -> Any:
return super().on_agent_finish(finish, *args, **kwargs)
def on_tool_start(
self, serialized: dict[str, Any], input_str: str, *args, **kwargs
):
self._update_active(DEFAULT_AVATARS["tool"], serialized["name"])
self._stream(f"Tool input: {input_str}")
return super().on_tool_start(serialized, input_str, *args, **kwargs)
def on_tool_end(self, output: str, *args, **kwargs):
self._stream(output)
self._reset_active()
return super().on_tool_end(output, *args, **kwargs)
def on_tool_error(
self, error: Union[Exception, KeyboardInterrupt], *args, **kwargs
):
return super().on_tool_error(error, *args, **kwargs)
def on_chain_start(
self, serialized: dict[str, Any], inputs: dict[str, Any], *args, **kwargs
):
self._disabled_state = self.instance.disabled
self.instance.disabled = True
return super().on_chain_start(serialized, inputs, *args, **kwargs)
def on_chain_end(self, outputs: dict[str, Any], *args, **kwargs):
self.instance.disabled = self._disabled_state
return super().on_chain_end(outputs, *args, **kwargs)
def on_retriever_error(
self, error: Union[Exception, KeyboardInterrupt], **kwargs: Any
) -> Any:
"""Run when Retriever errors."""
return super().on_retriever_error(error, **kwargs)
def on_retriever_end(self, documents, **kwargs: Any) -> Any:
"""Run when Retriever ends running."""
objects = [(f"Document {index}", document.page_content) for index, document in enumerate(documents)]
message = Accordion(*objects, sizing_mode="stretch_width", margin=(10,13,10,5))
self.instance.send(
message,
user="LangChain (retriever)",
avatar=DEFAULT_AVATARS["retriever"],
respond=False,
)
return super().on_retriever_end(documents=documents, **kwargs)
def on_text(self, text: str, **kwargs: Any):
"""Run when text is received."""
return super().on_text(text, **kwargs)
def on_chat_model_start(
self,
serialized: dict[str, Any],
messages: list,
**kwargs: Any
) -> None:
"""
To prevent the inherited class from raising
NotImplementedError, will not call super() here.
"""
self._on_start(serialized, kwargs)