remove pysbd requirement

pyproject.toml

@@ -13,7 +13,6 @@ dependencies = [
     "appdirs",
     "pydantic>=2",
     "openai>1",
-    "pysbd>=0.3.4",
     "diskcache>=5.6.3",
 ]
 dynamic = ["version", "readme"]

src/ragas/metrics/base.py

@@ -9,7 +9,6 @@
 from enum import Enum
 
 from pydantic import ValidationError
-from pysbd import Segmenter
 from tqdm import tqdm
 
 from ragas._analytics import EvaluationEvent, _analytics_batcher
@@ -19,12 +18,7 @@
 from ragas.losses import BinaryMetricLoss, MSELoss
 from ragas.prompt import FewShotPydanticPrompt, PromptMixin
 from ragas.run_config import RunConfig
-from ragas.utils import (
-    RAGAS_SUPPORTED_LANGUAGE_CODES,
-    camel_to_snake,
-    deprecated,
-    get_metric_language,
-)
+from ragas.utils import camel_to_snake, deprecated, get_metric_language
 
 if t.TYPE_CHECKING:
     from langchain_core.callbacks import Callbacks
@@ -694,22 +688,4 @@ def from_discrete(
         return verdict_agg
 
 
-def get_segmenter(
-    language: str = "english", clean: bool = False, char_span: bool = False
-):
-    """
-    Get a sentence segmenter for a given language
-    """
-    language = language.lower()
-    if language not in RAGAS_SUPPORTED_LANGUAGE_CODES:
-        raise ValueError(
-            f"Language '{language}' not supported. Supported languages: {RAGAS_SUPPORTED_LANGUAGE_CODES.keys()}"
-        )
-    return Segmenter(
-        language=RAGAS_SUPPORTED_LANGUAGE_CODES[language],
-        clean=clean,
-        char_span=char_span,
-    )
-
-
 ensembler = Ensember()

src/ragas/prompt/base.py

@@ -7,7 +7,7 @@
 from langchain_core.prompt_values import StringPromptValue
 from pydantic import BaseModel
 
-from ragas.utils import RAGAS_SUPPORTED_LANGUAGE_CODES, camel_to_snake
+from ragas.utils import camel_to_snake
 
 if t.TYPE_CHECKING:
     from langchain_core.callbacks import Callbacks
@@ -17,13 +17,6 @@
 logger = logging.getLogger(__name__)
 
 
-def _check_if_language_is_supported(language: str):
-    if language not in RAGAS_SUPPORTED_LANGUAGE_CODES:
-        raise ValueError(
-            f"Language '{language}' not supported. Supported languages: {RAGAS_SUPPORTED_LANGUAGE_CODES.keys()}"
-        )
-
-
 class BasePrompt(ABC):
     def __init__(
         self,
@@ -34,7 +27,6 @@ def __init__(
         if name is None:
             self.name = camel_to_snake(self.__class__.__name__)
 
-        _check_if_language_is_supported(language)
         self.language = language
         self.original_hash = original_hash
 

src/ragas/prompt/mixin.py

@@ -5,7 +5,6 @@
 import os
 import typing as t
 
-from .base import _check_if_language_is_supported
 from .pydantic_prompt import PydanticPrompt
 
 if t.TYPE_CHECKING:
@@ -111,7 +110,6 @@ def load_prompts(self, path: str, language: t.Optional[str] = None):
                 "Language not specified, loading prompts for default language: %s",
                 language,
             )
-        _check_if_language_is_supported(language)
 
         loaded_prompts = {}
         for prompt_name, prompt in self.get_prompts().items():

src/ragas/prompt/pydantic_prompt.py

@@ -15,7 +15,7 @@
 from ragas.callbacks import ChainType, new_group
 from ragas.exceptions import RagasOutputParserException
 
-from .base import BasePrompt, StringIO, _check_if_language_is_supported
+from .base import BasePrompt, StringIO
 from .utils import extract_json, get_all_strings, update_strings
 
 if t.TYPE_CHECKING:
@@ -227,9 +227,6 @@ async def adapt(
         Adapt the prompt to a new language.
         """
 
-        # throws ValueError if language is not supported
-        _check_if_language_is_supported(target_language)
-
         # set the original hash, this is used to
         # identify the original prompt object when loading from file
         if self.original_hash is None:

src/ragas/utils.py

@@ -11,15 +11,11 @@
 import numpy as np
 import tiktoken
 from datasets import Dataset
-from pysbd.languages import LANGUAGE_CODES
 
 if t.TYPE_CHECKING:
     from ragas.metrics.base import Metric
 
 DEBUG_ENV_VAR = "RAGAS_DEBUG"
-RAGAS_SUPPORTED_LANGUAGE_CODES = {
-    v.__name__.lower(): k for k, v in LANGUAGE_CODES.items()
-}
 
 
 @lru_cache(maxsize=1)