обновлен формат генерации датасета

llm/factory.py

@@ -1,6 +1,5 @@
 import os
 import webbrowser
-import pandas as pd
 from pathlib import Path
 
 from pydantic_ai.models.openai import OpenAIChatModel
@@ -13,7 +12,6 @@ from pydantic_ai.providers.openai_codex import (
 
 from .config import CodexConfig, LLMConfig, OpenAiConfig
 from .credentials import FileCodexCredentionalSource
-from .schemas import BaseDatasetRecord, ChatRecord, ChatMessage
 
 
 class LLMModelProvider:
@@ -22,7 +20,7 @@ class LLMModelProvider:
     def create_openai_compatible(model_name: str,
                                  api_key: str,
                                  base_url: str
-                                ) -> OpenAICodexModel:
+                                ) -> OpenAIChatModel:
 
         provider = OpenAIProvider(base_url=base_url,
                                 api_key=api_key)
@@ -84,68 +82,3 @@ class LLMModelProvider:
             )
 
         raise ValueError(f"Unsupported LLM config: {type(config)}")
-
-class DatasetWriter:
-    def __init__(self,
-                 dataset_record: BaseDatasetRecord,
-                 base_path: str | Path):
-
-
-        self._dataset_record = dataset_record    
-        self._path = Path(base_path)
-        self._path.mkdir(parents=True, exist_ok=True)
-
-    def __check_file_exists(self, filename :str)-> bool:
-        file_path = self._path / filename
-        return file_path.exists()
-
-    def to_csv(self,
-               filename: str) -> None:
-
-        exists=self.__check_file_exists(filename)
-        
-        df = pd.json_normalize(
-            self._dataset_record.model_dump()
-        )
-
-        df.to_csv(self._path / filename,
-                  mode="a" if exists else "w",
-                  header=not exists,
-                  index=False,
-                  encoding="utf-8")
-
-    def to_chat_record(self) -> ChatRecord:
-        response = self._dataset_record.response
-        prompts = self._dataset_record.prompts
-
-        assistant_content = (
-            response.assistant_content
-            or response.model_response
-        )
-
-        return ChatRecord(
-            messages=[
-                ChatMessage(
-                    role="system",
-                    content=prompts.system_prompt,
-                ),
-                ChatMessage(
-                    role="user",
-                    content=prompts.user_prompt,
-                ),
-                ChatMessage(
-                    role="assistant",
-                    content=assistant_content,
-                ),
-            ]
-        )
-
-    def to_jsonl(self,
-                 filename: str) -> None:
-        
-        chat_record = self.to_chat_record()
-        path = self._path/filename
-
-        with path.open("a", encoding='utf-8') as file:
-            file.write(chat_record.model_dump_json())
-            file.write("\n")