File size: 5,094 Bytes
ed4d993
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import logging
from typing import Any, Dict, List, Optional

import requests
from langchain_core.callbacks import CallbackManagerForLLMRun
from langchain_core.language_models.llms import LLM

logger = logging.getLogger(__name__)


def clean_url(url: str) -> str:
    """Remove trailing slash and /api from url if present."""
    if url.endswith("/api"):
        return url[:-4]
    elif url.endswith("/"):
        return url[:-1]
    else:
        return url


class KoboldApiLLM(LLM):
    """Kobold API language model.

    It includes several fields that can be used to control the text generation process.

    To use this class, instantiate it with the required parameters and call it with a
    prompt to generate text. For example:

        kobold = KoboldApiLLM(endpoint="http://localhost:5000")
        result = kobold("Write a story about a dragon.")

    This will send a POST request to the Kobold API with the provided prompt and
    generate text.
    """

    endpoint: str
    """The API endpoint to use for generating text."""

    use_story: Optional[bool] = False
    """ Whether or not to use the story from the KoboldAI GUI when generating text. """

    use_authors_note: Optional[bool] = False
    """Whether to use the author's note from the KoboldAI GUI when generating text.
    
    This has no effect unless use_story is also enabled.
    """

    use_world_info: Optional[bool] = False
    """Whether to use the world info from the KoboldAI GUI when generating text."""

    use_memory: Optional[bool] = False
    """Whether to use the memory from the KoboldAI GUI when generating text."""

    max_context_length: Optional[int] = 1600
    """Maximum number of tokens to send to the model.
    
    minimum: 1
    """

    max_length: Optional[int] = 80
    """Number of tokens to generate.
    
    maximum: 512
    minimum: 1
    """

    rep_pen: Optional[float] = 1.12
    """Base repetition penalty value.
    
    minimum: 1
    """

    rep_pen_range: Optional[int] = 1024
    """Repetition penalty range.
    
    minimum: 0
    """

    rep_pen_slope: Optional[float] = 0.9
    """Repetition penalty slope.
    
    minimum: 0
    """

    temperature: Optional[float] = 0.6
    """Temperature value.
    
    exclusiveMinimum: 0
    """

    tfs: Optional[float] = 0.9
    """Tail free sampling value.
    
    maximum: 1
    minimum: 0
    """

    top_a: Optional[float] = 0.9
    """Top-a sampling value.
    
    minimum: 0
    """

    top_p: Optional[float] = 0.95
    """Top-p sampling value.
    
    maximum: 1
    minimum: 0
    """

    top_k: Optional[int] = 0
    """Top-k sampling value.
    
    minimum: 0
    """

    typical: Optional[float] = 0.5
    """Typical sampling value.
    
    maximum: 1
    minimum: 0
    """

    @property
    def _llm_type(self) -> str:
        return "koboldai"

    def _call(
        self,
        prompt: str,
        stop: Optional[List[str]] = None,
        run_manager: Optional[CallbackManagerForLLMRun] = None,
        **kwargs: Any,
    ) -> str:
        """Call the API and return the output.

        Args:
            prompt: The prompt to use for generation.
            stop: A list of strings to stop generation when encountered.

        Returns:
            The generated text.

        Example:
            .. code-block:: python

                from langchain_community.llms import KoboldApiLLM

                llm = KoboldApiLLM(endpoint="http://localhost:5000")
                llm.invoke("Write a story about dragons.")
        """
        data: Dict[str, Any] = {
            "prompt": prompt,
            "use_story": self.use_story,
            "use_authors_note": self.use_authors_note,
            "use_world_info": self.use_world_info,
            "use_memory": self.use_memory,
            "max_context_length": self.max_context_length,
            "max_length": self.max_length,
            "rep_pen": self.rep_pen,
            "rep_pen_range": self.rep_pen_range,
            "rep_pen_slope": self.rep_pen_slope,
            "temperature": self.temperature,
            "tfs": self.tfs,
            "top_a": self.top_a,
            "top_p": self.top_p,
            "top_k": self.top_k,
            "typical": self.typical,
        }

        if stop is not None:
            data["stop_sequence"] = stop

        response = requests.post(
            f"{clean_url(self.endpoint)}/api/v1/generate", json=data
        )

        response.raise_for_status()
        json_response = response.json()

        if (
            "results" in json_response
            and len(json_response["results"]) > 0
            and "text" in json_response["results"][0]
        ):
            text = json_response["results"][0]["text"].strip()

            if stop is not None:
                for sequence in stop:
                    if text.endswith(sequence):
                        text = text[: -len(sequence)].rstrip()

            return text
        else:
            raise ValueError(
                f"Unexpected response format from Kobold API:  {json_response}"
            )