File size: 5,292 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
from typing import Any, Dict, List, Optional, cast

import requests
from langchain_core.callbacks import CallbackManagerForLLMRun
from langchain_core.language_models.llms import LLM
from langchain_core.pydantic_v1 import BaseModel, Extra, SecretStr, root_validator
from langchain_core.utils import convert_to_secret_str, get_from_dict_or_env


class AI21PenaltyData(BaseModel):
    """Parameters for AI21 penalty data."""

    scale: int = 0
    applyToWhitespaces: bool = True
    applyToPunctuations: bool = True
    applyToNumbers: bool = True
    applyToStopwords: bool = True
    applyToEmojis: bool = True


class AI21(LLM):
    """AI21 large language models.

    To use, you should have the environment variable ``AI21_API_KEY``
    set with your API key or pass it as a named parameter to the constructor.

    Example:
        .. code-block:: python

            from langchain_community.llms import AI21
            ai21 = AI21(ai21_api_key="my-api-key", model="j2-jumbo-instruct")
    """

    model: str = "j2-jumbo-instruct"
    """Model name to use."""

    temperature: float = 0.7
    """What sampling temperature to use."""

    maxTokens: int = 256
    """The maximum number of tokens to generate in the completion."""

    minTokens: int = 0
    """The minimum number of tokens to generate in the completion."""

    topP: float = 1.0
    """Total probability mass of tokens to consider at each step."""

    presencePenalty: AI21PenaltyData = AI21PenaltyData()
    """Penalizes repeated tokens."""

    countPenalty: AI21PenaltyData = AI21PenaltyData()
    """Penalizes repeated tokens according to count."""

    frequencyPenalty: AI21PenaltyData = AI21PenaltyData()
    """Penalizes repeated tokens according to frequency."""

    numResults: int = 1
    """How many completions to generate for each prompt."""

    logitBias: Optional[Dict[str, float]] = None
    """Adjust the probability of specific tokens being generated."""

    ai21_api_key: Optional[SecretStr] = None

    stop: Optional[List[str]] = None

    base_url: Optional[str] = None
    """Base url to use, if None decides based on model name."""

    class Config:
        """Configuration for this pydantic object."""

        extra = Extra.forbid

    @root_validator()
    def validate_environment(cls, values: Dict) -> Dict:
        """Validate that api key exists in environment."""
        ai21_api_key = convert_to_secret_str(
            get_from_dict_or_env(values, "ai21_api_key", "AI21_API_KEY")
        )
        values["ai21_api_key"] = ai21_api_key
        return values

    @property
    def _default_params(self) -> Dict[str, Any]:
        """Get the default parameters for calling AI21 API."""
        return {
            "temperature": self.temperature,
            "maxTokens": self.maxTokens,
            "minTokens": self.minTokens,
            "topP": self.topP,
            "presencePenalty": self.presencePenalty.dict(),
            "countPenalty": self.countPenalty.dict(),
            "frequencyPenalty": self.frequencyPenalty.dict(),
            "numResults": self.numResults,
            "logitBias": self.logitBias,
        }

    @property
    def _identifying_params(self) -> Dict[str, Any]:
        """Get the identifying parameters."""
        return {**{"model": self.model}, **self._default_params}

    @property
    def _llm_type(self) -> str:
        """Return type of llm."""
        return "ai21"

    def _call(
        self,
        prompt: str,
        stop: Optional[List[str]] = None,
        run_manager: Optional[CallbackManagerForLLMRun] = None,
        **kwargs: Any,
    ) -> str:
        """Call out to AI21's complete endpoint.

        Args:
            prompt: The prompt to pass into the model.
            stop: Optional list of stop words to use when generating.

        Returns:
            The string generated by the model.

        Example:
            .. code-block:: python

                response = ai21("Tell me a joke.")
        """
        if self.stop is not None and stop is not None:
            raise ValueError("`stop` found in both the input and default params.")
        elif self.stop is not None:
            stop = self.stop
        elif stop is None:
            stop = []
        if self.base_url is not None:
            base_url = self.base_url
        else:
            if self.model in ("j1-grande-instruct",):
                base_url = "https://api.ai21.com/studio/v1/experimental"
            else:
                base_url = "https://api.ai21.com/studio/v1"
        params = {**self._default_params, **kwargs}
        self.ai21_api_key = cast(SecretStr, self.ai21_api_key)
        response = requests.post(
            url=f"{base_url}/{self.model}/complete",
            headers={"Authorization": f"Bearer {self.ai21_api_key.get_secret_value()}"},
            json={"prompt": prompt, "stopSequences": stop, **params},
        )
        if response.status_code != 200:
            optional_detail = response.json().get("error")
            raise ValueError(
                f"AI21 /complete call failed with status code {response.status_code}."
                f" Details: {optional_detail}"
            )
        response_json = response.json()
        return response_json["completions"][0]["data"]["text"]