langchain/libs/community/langchain_community/llms/solar.py

from typing import Any, Dict, List, Optional

import requests
from langchain_core.callbacks import CallbackManagerForLLMRun
from langchain_core.language_models import LLM
from langchain_core.utils import convert_to_secret_str, get_from_dict_or_env, pre_init
from pydantic import (
    BaseModel,
    ConfigDict,
    Field,
    SecretStr,
    model_validator,
)

from langchain_community.llms.utils import enforce_stop_tokens

SOLAR_SERVICE_URL_BASE = "https://api.upstage.ai/v1/solar"
SOLAR_SERVICE = "https://api.upstage.ai"


class _SolarClient(BaseModel):
    """An API client that talks to the Solar server."""

    api_key: SecretStr
    """The API key to use for authentication."""
    base_url: str = SOLAR_SERVICE_URL_BASE

    def completion(self, request: Any) -> Any:
        headers = {"Authorization": f"Bearer {self.api_key.get_secret_value()}"}
        response = requests.post(
            f"{self.base_url}/chat/completions",
            headers=headers,
            json=request,
        )
        if not response.ok:
            raise ValueError(f"HTTP {response.status_code} error: {response.text}")
        return response.json()["choices"][0]["message"]["content"]


class SolarCommon(BaseModel):
    """Common configuration for Solar LLMs."""

    _client: _SolarClient
    base_url: str = SOLAR_SERVICE_URL_BASE
    solar_api_key: Optional[SecretStr] = Field(default=None, alias="api_key")
    """Solar API key. Get it here: https://console.upstage.ai/services/solar"""
    model_name: str = Field(default="solar-mini", alias="model")
    """Model name. Available models listed here: https://console.upstage.ai/services/solar"""
    max_tokens: int = Field(default=1024)
    temperature: float = 0.3

    model_config = ConfigDict(
        populate_by_name=True,
        arbitrary_types_allowed=True,
        extra="ignore",
        protected_namespaces=(),
    )

    @property
    def lc_secrets(self) -> dict:
        return {"solar_api_key": "SOLAR_API_KEY"}

    @property
    def _default_params(self) -> Dict[str, Any]:
        return {
            "model": self.model_name,
            "max_tokens": self.max_tokens,
            "temperature": self.temperature,
        }

    @property
    def _invocation_params(self) -> Dict[str, Any]:
        return {**{"model": self.model_name}, **self._default_params}

    @model_validator(mode="before")
    @classmethod
    def build_extra(cls, values: Dict[str, Any]) -> Any:
        return values

    @pre_init
    def validate_environment(cls, values: Dict) -> Dict:
        api_key = get_from_dict_or_env(values, "solar_api_key", "SOLAR_API_KEY")
        if api_key is None or len(api_key) == 0:
            raise ValueError("SOLAR_API_KEY must be configured")

        values["solar_api_key"] = convert_to_secret_str(api_key)

        if "base_url" not in values:
            values["base_url"] = SOLAR_SERVICE_URL_BASE

        if "base_url" in values and not values["base_url"].startswith(SOLAR_SERVICE):
            raise ValueError("base_url must match with: " + SOLAR_SERVICE)

        values["_client"] = _SolarClient(
            api_key=values["solar_api_key"], base_url=values["base_url"]
        )
        return values

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


class Solar(SolarCommon, LLM):
    """Solar large language models.

    To use, you should have the environment variable
    ``SOLAR_API_KEY`` set with your API key.
    Referenced from https://console.upstage.ai/services/solar
    """

    model_config = ConfigDict(
        populate_by_name=True,
    )

    def _call(
        self,
        prompt: str,
        stop: Optional[List[str]] = None,
        run_manager: Optional[CallbackManagerForLLMRun] = None,
        **kwargs: Any,
    ) -> str:
        request = self._invocation_params
        request["messages"] = [{"role": "user", "content": prompt}]
        request.update(kwargs)
        text = self._client.completion(request)
        if stop is not None:
            # This is required since the stop tokens
            # are not enforced by the model parameters
            text = enforce_stop_tokens(text, stop)

        return text