File size: 10,136 Bytes
bec8a7b
25cdf03
 
bec8a7b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25cdf03
 
 
 
bec8a7b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25cdf03
bec8a7b
 
 
 
 
 
 
 
25cdf03
bec8a7b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25cdf03
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
from app.schemas.message_schema import IChatResponse
from app.utils.adaptive_cards.cards import create_adaptive_card, create_image_card
from app.utils.chains import get_suggestions_questions
from langchain.callbacks.base import AsyncCallbackHandler
from app.utils.utils import generate_uuid
from fastapi import WebSocket
from uuid import UUID
from typing import Any
from langchain.schema.agent import AgentFinish
from langchain.schema.output import LLMResult


DEFAULT_ANSWER_PREFIX_TOKENS = ["Final", " Answer", ":"]


class CustomAsyncCallbackHandler(AsyncCallbackHandler):
    def append_to_last_tokens(self, token: str) -> None:
        self.last_tokens.append(token)
        self.last_tokens_stripped.append(token.strip())
        if len(self.last_tokens) > len(self.answer_prefix_tokens):
            self.last_tokens.pop(0)
            self.last_tokens_stripped.pop(0)

    def check_if_answer_reached(self) -> bool:
        if self.strip_tokens:
            return self.last_tokens_stripped == self.answer_prefix_tokens_stripped
        else:
            return self.last_tokens == self.answer_prefix_tokens

    def update_message_id(self, message_id: str = generate_uuid()):
        self.message_id = message_id

    def __init__(
        self,
        websocket: WebSocket,
        *,
        message_id: str = generate_uuid(),
        answer_prefix_tokens: list[str] | None = None,
        strip_tokens: bool = True,
        stream_prefix: bool = False,
    ) -> None:
        """Instantiate FinalStreamingStdOutCallbackHandler.

        Args:
            answer_prefix_tokens: Token sequence that prefixes the answer.
                Default is ["Final", "Answer", ":"]
            strip_tokens: Ignore white spaces and new lines when comparing
                answer_prefix_tokens to last tokens? (to determine if answer has been
                reached)
            stream_prefix: Should answer prefix itself also be streamed?
        """
        self.websocket: WebSocket = websocket
        self.message_id: str = message_id
        self.text: str = ""
        self.started: bool = False
        self.loading_card = create_image_card(
            "https://res.cloudinary.com/dnv0qwkrk/image/upload/v1691005682/Alita/Ellipsis-2.4s-81px_1_nja8hq.gif"
        )
        self.adaptive_card = self.loading_card

        if answer_prefix_tokens is None:
            self.answer_prefix_tokens = DEFAULT_ANSWER_PREFIX_TOKENS
        else:
            self.answer_prefix_tokens = answer_prefix_tokens
        if strip_tokens:
            self.answer_prefix_tokens_stripped = [
                token.strip() for token in self.answer_prefix_tokens
            ]
        else:
            self.answer_prefix_tokens_stripped = self.answer_prefix_tokens
        self.last_tokens = [""] * len(self.answer_prefix_tokens)
        self.last_tokens_stripped = [""] * len(self.answer_prefix_tokens)
        self.strip_tokens = strip_tokens
        self.stream_prefix = stream_prefix
        self.answer_reached = False

    async def on_llm_start(
        self, serialized: dict[str, Any], prompts: list[str], **kwargs: Any
    ) -> None:
        """Run when LLM starts running."""

        resp = IChatResponse(
            id="",
            message_id=self.message_id,
            sender="bot",
            message=self.loading_card.to_dict(),
            type="start",
        )
        await self.websocket.send_json(resp.dict())

    async def on_llm_new_token(self, token: str, **kwargs: Any) -> None:
        """Run on new LLM token. Only available when streaming is enabled."""
        # Remember the last n tokens, where n = len(answer_prefix_tokens)
        self.append_to_last_tokens(token)

        self.text += f"{token}"
        self.adaptive_card = create_adaptive_card(self.text)
        resp = IChatResponse(
            # id=generate_uuid(),
            id="",
            message_id=self.message_id,
            sender="bot",
            message=self.adaptive_card.to_dict(),
            type="stream",
        )
        await self.websocket.send_json(resp.model_dump())

    async def on_llm_end(
        self,
        response: LLMResult,
        *,
        run_id: UUID,
        parent_run_id: UUID | None = None,
        tags: list[str] | None = None,
        **kwargs: Any,
    ) -> None:
        """Run when LLM ends running."""
        resp = IChatResponse(
            id="",
            message_id=self.message_id,
            sender="bot",
            message=self.adaptive_card.to_dict(),
            type="end",
        )
        await self.websocket.send_json(resp.model_dump())


class CustomFinalStreamingStdOutCallbackHandler(AsyncCallbackHandler):
    """Callback handler for streaming in agents.
    Only works with agents using LLMs that support streaming.

    Only the final output of the agent will be streamed.
    """

    def append_to_last_tokens(self, token: str) -> None:
        self.last_tokens.append(token)
        self.last_tokens_stripped.append(token.strip())
        if len(self.last_tokens) > len(self.answer_prefix_tokens):
            self.last_tokens.pop(0)
            self.last_tokens_stripped.pop(0)

    def check_if_answer_reached(self) -> bool:
        if self.strip_tokens:
            return self.last_tokens_stripped == self.answer_prefix_tokens_stripped
        else:
            return self.last_tokens == self.answer_prefix_tokens

    def update_message_id(self, message_id: str = generate_uuid()):
        self.message_id = message_id

    def __init__(
        self,
        websocket: WebSocket,
        *,
        message_id: str = generate_uuid(),
        answer_prefix_tokens: list[str] | None = None,
        strip_tokens: bool = True,
        stream_prefix: bool = False,
    ) -> None:
        """Instantiate FinalStreamingStdOutCallbackHandler.

        Args:
            answer_prefix_tokens: Token sequence that prefixes the answer.
                Default is ["Final", "Answer", ":"]
            strip_tokens: Ignore white spaces and new lines when comparing
                answer_prefix_tokens to last tokens? (to determine if answer has been
                reached)
            stream_prefix: Should answer prefix itself also be streamed?
        """
        self.websocket: WebSocket = websocket
        self.message_id: str = message_id
        self.text: str = ""
        self.started: bool = False
        self.loading_card = create_image_card(
            "https://res.cloudinary.com/dnv0qwkrk/image/upload/v1691005682/Alita/Ellipsis-2.4s-81px_1_nja8hq.gif"
        )
        self.adaptive_card = self.loading_card

        if answer_prefix_tokens is None:
            self.answer_prefix_tokens = DEFAULT_ANSWER_PREFIX_TOKENS
        else:
            self.answer_prefix_tokens = answer_prefix_tokens
        if strip_tokens:
            self.answer_prefix_tokens_stripped = [
                token.strip() for token in self.answer_prefix_tokens
            ]
        else:
            self.answer_prefix_tokens_stripped = self.answer_prefix_tokens
        self.last_tokens = [""] * len(self.answer_prefix_tokens)
        self.last_tokens_stripped = [""] * len(self.answer_prefix_tokens)
        self.strip_tokens = strip_tokens
        self.stream_prefix = stream_prefix
        self.answer_reached = False

    async def on_llm_start(
        self, serialized: dict[str, Any], prompts: list[str], **kwargs: Any
    ) -> None:
        """Run when LLM starts running."""
        if self.started == False:
            self.started = True
            resp = IChatResponse(
                id="",
                message_id=self.message_id,
                sender="bot",
                message=self.loading_card.to_dict(),
                type="start",
            )
            await self.websocket.send_json(resp.model_dump())

    async def on_agent_finish(
        self,
        finish: AgentFinish,
        *,
        run_id: UUID,
        parent_run_id: UUID | None = None,
        **kwargs: Any,
    ) -> Any:
        """Run on agent end."""
        message: str = (
            self.text
            if self.text != ""
            # else "😕 Lo siento no he podido hallar lo que buscabas"
            else finish.return_values["output"]
        )
        self.adaptive_card = create_adaptive_card(message)

        resp = IChatResponse(
            id="",
            message_id=self.message_id,
            sender="bot",
            message=self.adaptive_card.to_dict(),
            type="stream",
        )
        await self.websocket.send_json(resp.dict())

        suggested_responses = await get_suggestions_questions(message)
        if len(suggested_responses) > 0:
            self.adaptive_card = create_adaptive_card(
                answer=message,
            )
        medium_resp = IChatResponse(
            id="",
            message_id=self.message_id,
            sender="bot",
            message=self.adaptive_card.to_dict(),
            type="end",
            suggested_responses=suggested_responses,
        )
        await self.websocket.send_json(medium_resp.model_dump())

        # Reset values
        self.text = ""
        self.answer_reached = False
        self.started = False

    async def on_llm_new_token(self, token: str, **kwargs: Any) -> None:
        """Run on new LLM token. Only available when streaming is enabled."""
        # Remember the last n tokens, where n = len(answer_prefix_tokens)
        self.append_to_last_tokens(token)

        # Check if the last n tokens match the answer_prefix_tokens list ...
        if self.check_if_answer_reached():
            self.answer_reached = True
            return

        # ... if yes, then print tokens from now on
        if self.answer_reached:
            self.text += f"{token}"
            self.adaptive_card = create_adaptive_card(self.text)

            resp = IChatResponse(
                id="",
                message_id=self.message_id,
                sender="bot",
                message=self.adaptive_card.to_dict(),
                type="stream",
            )
            await self.websocket.send_json(resp.model_dump())