feat (experimental) added new prompt and metric into ragas.experimental (#1240)

you can use it like
```py
from ragas.experimental.metrics import FaithfulnessExperimental
from ragas.metrics import faithfulness
from ragas import evaluate


f = FaithfulnessExperimental(llm=LangchainLLMWrapper(gpt4o))
faithfulness.llm = LangchainLLMWrapper(gpt4o)

# row = amnesty_qa["eval"][0]
# await f.ascore(row)
# await faithfulness.ascore(row)

r = evaluate(
    amnesty_qa["eval"].select(range(10)),
    metrics=[f, faithfulness],
    raise_exceptions=True,
    callbacks=[]
)
```

Author: jjmachan

src/experimental/tests/test_prompt.py

@@ -1,11 +1,11 @@
-from ragas_experimental.llms.prompt import StringPrompt, StringIO
+import pytest
+from langchain_core.outputs import Generation, LLMResult
+from ragas_experimental.llms.prompt import StringIO, StringPrompt
+
 from ragas.llms.base import BaseRagasLLM
-from langchain_core.outputs import LLMResult, Generation
 from ragas.llms.prompt import PromptValue
 from ragas.run_config import RunConfig
 
-import pytest
-
 
 class EchoLLM(BaseRagasLLM):
     def generate_text(  # type: ignore
@@ -37,10 +37,11 @@ async def test_string_prompt():
 
 
 def test_process_fields():
-    from ragas_experimental.llms.prompt import PydanticPrompt, StringIO
-    from pydantic import BaseModel
     from enum import Enum
 
+    from pydantic import BaseModel
+    from ragas_experimental.llms.prompt import PydanticPrompt, StringIO
+
     class Categories(str, Enum):
         science = "science"
         commerce = "commerce"
@@ -63,10 +64,7 @@ class JokeGenerator(PydanticPrompt[InputModel, StringIO]):
 
 @pytest.mark.asyncio
 async def test_pydantic_prompt_io():
-    from ragas_experimental.llms.prompt import (
-        PydanticPrompt,
-        StringIO,
-    )
+    from ragas_experimental.llms.prompt import PydanticPrompt, StringIO
 
     class Prompt(PydanticPrompt[StringIO, StringIO]):
         instruction = ""
@@ -82,9 +80,7 @@ class Prompt(PydanticPrompt[StringIO, StringIO]):
 
 
 def test_pydantic_prompt_examples():
-    from ragas_experimental.llms.prompt import (
-        PydanticPrompt,
-    )
+    from ragas_experimental.llms.prompt import PydanticPrompt
 
     class Prompt(PydanticPrompt[StringIO, StringIO]):
         instruction = ""

src/ragas/experimental/llms/__init__.py

@@ -1,21 +1,21 @@
 from __future__ import annotations
 
-from abc import ABC, abstractmethod
-from dataclasses import dataclass
-import json
 import typing as t
+from abc import ABC, abstractmethod
 
-from ragas.llms.output_parser import RagasoutputParser
-from ragas.llms.prompt import PromptValue
+import pydantic
 
 # Check Pydantic version
 from pydantic import BaseModel
-import pydantic
+
+from ragas.llms.output_parser import RagasoutputParser
+from ragas.llms.prompt import PromptValue
 
 if t.TYPE_CHECKING:
-    from ragas.llms.base import BaseRagasLLM
     from langchain_core.callbacks import Callbacks
 
+    from ragas.llms.base import BaseRagasLLM
+
 PYDANTIC_V2 = pydantic.VERSION.startswith("2.")
 
 
@@ -24,7 +24,7 @@ def __init__(self, llm):
         self.llm: BaseRagasLLM = llm
 
     @abstractmethod
-    async def generate(self, data: t.Any) -> t.Any:
+    async def generate(self, data: t.Any, callbacks: Callbacks = None) -> t.Any:
         pass
 
 
@@ -57,12 +57,14 @@ def to_json(model: t.Any, indent: int = 4) -> str:
         return model.json(indent=indent)
 
 
-def model_to_json_schema(model: t.Type[BaseModel]) -> dict:
+def model_to_json_schema(model: t.Type[BaseModel]) -> str:
     if PYDANTIC_V2:
-        return model.model_json_schema()
+        # NOTE: this is not the same as model.schema_json()
+        return model.model_json_schema()  # type: ignore
     else:
         return model.schema_json()
 
+
 InputModel = t.TypeVar("InputModel", bound=BaseModel)
 OutputModel = t.TypeVar("OutputModel", bound=BaseModel)
 
@@ -96,9 +98,11 @@ def generate_examples(self):
                 example_strings.append(
                     self.instruction
                     + "\n"
-                    + "input: " + to_json(input_data, indent=4)
+                    + "input: "
+                    + to_json(input_data, indent=4)
                     + "\n"
-                    + "output: " + to_json(output_data, indent=4)
+                    + "output: "
+                    + to_json(output_data, indent=4)
                 )
 
             return (
@@ -118,12 +122,15 @@ def to_string(self, data: InputModel) -> str:
             + "\n"
             + self.generate_examples()
             + "\nNow perform the above instruction with the following input\n"
-            + "input: " + to_json(data, indent=4)
+            + "input: "
+            + to_json(data, indent=4)
             + "\n"
             + "output: "
         )
 
-    async def generate(self, data: InputModel, callbacks: Callbacks) -> OutputModel:
+    async def generate(
+        self, data: InputModel, callbacks: Callbacks = None
+    ) -> OutputModel:
         prompt_value = PromptValue(prompt_str=self.to_string(data))
         resp = await self.llm.generate(prompt_value, callbacks=callbacks)
         resp_text = resp.generations[0][0].text
@@ -135,7 +142,7 @@ async def generate(self, data: InputModel, callbacks: Callbacks) -> OutputModel:
 
 
 class StringPrompt(BasePrompt):
-    async def generate(self, data: str) -> str:
+    async def generate(self, data: str, callbacks: Callbacks = None) -> str:
         prompt_value = PromptValue(prompt_str=data)
-        llm_result = await self.llm.agenerate_text(prompt_value)
-        return llm_result.generations[0][0].text
+        llm_result = await self.llm.agenerate_text(prompt_value, callbacks=callbacks)
+        return llm_result.generations[0][0].text

src/experimental/ragas_experimental/llms/prompt.py renamed to src/ragas/experimental/llms/prompt.py

@@ -1,3 +1,3 @@
 from ._faithfulness import FaithfulnessExperimental
 
-__all__ = ["FaithfulnessExperimental"]
+__all__ = ["FaithfulnessExperimental"]

src/experimental/ragas_experimental/metrics/__init__.py renamed to src/ragas/experimental/metrics/__init__.py

@@ -1,17 +1,18 @@
 from __future__ import annotations
 
-import typing as t
 import logging
+import typing as t
 from dataclasses import dataclass
 
-from pydantic import BaseModel, Field
 import numpy as np
+from pydantic import BaseModel, Field
 
+from ragas.experimental.llms.prompt import PydanticPrompt
 from ragas.metrics.base import EvaluationMode, MetricWithLLM, get_segmenter
-from ragas_experimental.llms.prompt import PydanticPrompt
 
 if t.TYPE_CHECKING:
     from langchain_core.callbacks import Callbacks
+
     from ragas.metrics._faithfulness import HasSegmentMethod
 
 
@@ -187,6 +188,8 @@ async def _ascore(self, row: t.Dict, callbacks: Callbacks) -> float:
         answer, question, contexts = row["answer"], row["question"], row["contexts"]
 
         # get the sentences from the answer
+        if self.sentence_segmenter is None:
+            raise ValueError("Sentence segmenter is not set")
         sentences = self.sentence_segmenter.segment(answer)
         # TODO: why do we do this?
         sentences = [
@@ -198,9 +201,9 @@ async def _ascore(self, row: t.Dict, callbacks: Callbacks) -> float:
                 answer=answer,
                 sentences={i: sentence for i, sentence in enumerate(sentences)},
             ),
-            callbacks=callbacks
+            callbacks=callbacks,
         )
-        
+
         statements = [
             statement
             for component in sentence_components.sentences
@@ -211,9 +214,9 @@ async def _ascore(self, row: t.Dict, callbacks: Callbacks) -> float:
                 context="\n".join(contexts),
                 statements=statements,
             ),
-            callbacks=callbacks
+            callbacks=callbacks,
         )
-        
+
         # compute the score
         num_faithful_statements = sum(
             verdict.verdict for verdict in verdicts.statements
@@ -223,4 +226,3 @@ async def _ascore(self, row: t.Dict, callbacks: Callbacks) -> float:
         else:
             score = np.nan
         return score
-

src/experimental/ragas_experimental/metrics/_faithfulness.py renamed to src/ragas/experimental/metrics/_faithfulness.py

@@ -63,7 +63,8 @@ def generate_text(
         temperature: float = 1e-8,
         stop: t.Optional[t.List[str]] = None,
         callbacks: Callbacks = None,
-    ) -> LLMResult: ...
+    ) -> LLMResult:
+        ...
 
     @abstractmethod
     async def agenerate_text(
@@ -73,7 +74,8 @@ async def agenerate_text(
         temperature: t.Optional[float] = None,
         stop: t.Optional[t.List[str]] = None,
         callbacks: Callbacks = None,
-    ) -> LLMResult: ...
+    ) -> LLMResult:
+        ...
 
     async def generate(
         self,