fix(type): Resolve all mypy type errors across 603 source files

Systematically fix 497 mypy errors across 147 files:
- Add explicit type casts (cast()) for no-any-return, arg-type errors
- Add proper type annotations for class attributes and variables
- Fix override signature mismatches in evaluation/distributed interfaces
- Handle Optional types with None guards for union-attr errors
- Use object.__setattr__ for frozen dataclass property assignments
- Add missing imports (Any, cast, AsyncGenerator, etc.)
- Configure mypy overrides for 100+ optional third-party libraries

Result: Success - no issues found in 603 source files
Co-authored-by: Cursor <cursoragent@cursor.com>

Author: leebeanbin

pyproject.toml

@@ -222,6 +222,7 @@ dev = [
     "black>=23.0.0,<26.0.0",
     "ruff>=0.1.0,<1.0.0",
     "mypy>=1.0.0,<2.0.0",
+    "types-PyYAML>=6.0.0",  # type stubs for yaml (finetuning/provider_axolotl)
     "bandit>=1.7.0,<2.0.0",  # 보안 검사 (pre-commit)
     "pre-commit>=3.0.0,<4.0.0",  # pre-commit 훅
     "faker>=20.0.0,<25.0.0",  # 테스트 데이터 생성
@@ -285,7 +286,127 @@ warn_unused_configs = true
 disallow_untyped_defs = false
 
 [[tool.mypy.overrides]]
-module = "prompt_toolkit.*"
+module = [
+    # CLI / REPL
+    "prompt_toolkit.*",
+    # AI / ML frameworks
+    "torch.*",
+    "transformers.*",
+    "sentence_transformers.*",
+    "tensorflow.*",
+    # Computer Vision
+    "cv2.*",
+    "ultralytics.*",
+    "segment_anything.*",
+    "sam2.*",
+    "sam3.*",
+    "surya.*",
+    "easyocr.*",
+    "paddleocr.*",
+    "pytesseract.*",
+    # NLP / NER
+    "spacy.*",
+    "flair.*",
+    "gliner.*",
+    # Document / PDF processing
+    "pypdf.*",
+    "fitz.*",
+    "docling.*",
+    "marker.*",
+    "nbformat.*",
+    "readability.*",
+    "trafilatura.*",
+    # RAG / Retrieval
+    "faiss.*",
+    "ragatouille.*",
+    "byaldi.*",
+    "rank_bm25.*",
+    "semchunk.*",
+    "cohere.*",
+    # Evaluation frameworks
+    "deepeval.*",
+    "ragas.*",
+    "lm_eval.*",
+    "trulens_eval.*",
+    "datasets.*",
+    # LLM providers
+    "anthropic.*",
+    "mistralai.*",
+    "voyageai.*",
+    "google.*",
+    "google.cloud.*",
+    "google.generativeai.*",
+    # Vector stores
+    "pinecone.*",
+    "qdrant_client.*",
+    "pymilvus.*",
+    "weaviate.*",
+    "lancedb.*",
+    "pgvector.*",
+    "chromadb.*",
+    # Databases
+    "pymongo.*",
+    "motor.*",
+    "psycopg2.*",
+    "asyncpg.*",
+    "neo4j.*",
+    # Distributed / Messaging
+    "redis.*",
+    "kafka.*",
+    # Data Science
+    "numpy.*",
+    "pandas.*",
+    "scipy.*",
+    "sklearn.*",
+    "matplotlib.*",
+    "plotly.*",
+    "seaborn.*",
+    "networkx.*",
+    "hdbscan.*",
+    "umap.*",
+    "joblib.*",
+    "cupy.*",
+    # Visualization / UI
+    "graphviz.*",
+    "streamlit.*",
+    "ipywidgets.*",
+    "IPython.*",
+    # Templating
+    "jinja2.*",
+    # Optimization
+    "bayes_opt.*",
+    # LangChain / LangGraph
+    "langgraph.*",
+    "langchain_openai.*",
+    "llama_index.*",
+    # Fine-tuning
+    "axolotl.*",
+    # Audio
+    "funasr.*",
+    "nemo.*",
+    # AWS
+    "boto3.*",
+    # Scheduler
+    "apscheduler.*",
+    # Search
+    "duckduckgo_search.*",
+    # Fine-tuning (additional)
+    "unsloth.*",
+    # Audio
+    "whisper.*",
+    # Cloud providers
+    "azure.*",
+    # Internal modules without py.typed
+    "beanllm.utils.logger",
+    "beanllm.domain.loaders.splitters",
+    "beanllm.domain.loaders.core.factory",
+    "beanllm.domain.tools.web_search",
+    "beanllm.vision_embeddings",
+    "beanllm.vision_loaders",
+    "beanllm.handler.ml.knowledge_graph_handler",
+    "beanllm.facade.service.types",
+    "beanllm.facade.vector_stores",
+]
 ignore_missing_imports = true
 
 # Pytest 설정

src/beanllm/domain/audio/engines/distil_whisper_engine.py

@@ -17,7 +17,7 @@
 
 import time
 from pathlib import Path
-from typing import Dict, Union
+from typing import Any, Dict, Union
 
 import numpy as np
 
@@ -136,7 +136,7 @@ def transcribe(self, audio_path: Union[str, Path, np.ndarray], config: STTConfig
             audio_path = str(audio_path)
 
         # Pipeline 옵션 설정
-        generate_kwargs = {
+        generate_kwargs: Dict[str, Any] = {
             "task": config.task,
             "language": None if config.language == "auto" else config.language,
         }
@@ -146,6 +146,7 @@ def transcribe(self, audio_path: Union[str, Path, np.ndarray], config: STTConfig
             generate_kwargs["num_beams"] = config.beam_size
 
         # 전사 실행
+        assert self._pipeline is not None, "Pipeline not initialized"
         if config.timestamp:
             result = self._pipeline(
                 audio_path,

src/beanllm/domain/audio/engines/whisper_engine.py

@@ -17,7 +17,7 @@
 
 import time
 from pathlib import Path
-from typing import Dict, Union
+from typing import Any, Dict, Union
 
 import numpy as np
 
@@ -148,7 +148,7 @@ def transcribe(self, audio_path: Union[str, Path, np.ndarray], config: STTConfig
             audio_path = str(audio_path)
 
         # Pipeline 옵션 설정
-        generate_kwargs = {
+        generate_kwargs: Dict[str, Any] = {
             "task": config.task,
             "language": None if config.language == "auto" else config.language,
         }

src/beanllm/domain/embeddings/api/api_embeddings.py

@@ -13,7 +13,7 @@
 Template Method Pattern을 사용하여 중복 코드 제거
 """
 
-from typing import List, Optional
+from typing import Any, List, Optional, cast
 
 from beanllm.domain.embeddings.base import BaseAPIEmbedding
 
@@ -79,6 +79,7 @@ async def embed(self, texts: List[str]) -> List[List[float]]:
 
         except Exception as e:
             self._handle_embed_error("OpenAI", e)
+            raise
 
     def embed_sync(self, texts: List[str]) -> List[List[float]]:
         """텍스트들을 임베딩 (동기)"""
@@ -94,6 +95,7 @@ def embed_sync(self, texts: List[str]) -> List[List[float]]:
 
         except Exception as e:
             self._handle_embed_error("OpenAI", e)
+            raise
 
 
 class GeminiEmbedding(BaseAPIEmbedding):
@@ -183,10 +185,11 @@ def embed_sync(self, texts: List[str]) -> List[List[float]]:
 
                 self._log_embed_success(len(texts), f"sequential mode, {len(texts)} API calls")
 
-            return embeddings
+            return cast(List[List[float]], embeddings)
 
         except Exception as e:
             self._handle_embed_error("Gemini", e)
+            raise
 
 
 class OllamaEmbedding(BaseAPIEmbedding):
@@ -270,7 +273,10 @@ def embed_sync(self, texts: List[str]) -> List[List[float]]:
 
                 embeddings = []
                 for text in texts:
-                    response = self.client.embeddings(model=self.model, prompt=text)
+                    response = cast(
+                        Any,
+                        self.client.embeddings(model=self.model, prompt=text),
+                    )
                     embeddings.append(response["embedding"])
 
                 self._log_embed_success(len(texts), f"sequential mode, {len(texts)} requests")
@@ -279,6 +285,7 @@ def embed_sync(self, texts: List[str]) -> List[List[float]]:
 
         except Exception as e:
             self._handle_embed_error("Ollama", e)
+            raise
 
 
 class VoyageEmbedding(BaseAPIEmbedding):
@@ -340,13 +347,13 @@ def __init__(self, model: str = "voyage-3", api_key: Optional[str] = None, **kwa
     def embed_sync(self, texts: List[str]) -> List[List[float]]:
         """텍스트들을 임베딩 (동기)"""
         try:
-            response = self.client.embed(texts=texts, model=self.model, **self.kwargs)
-
+            response: Any = self.client.embed(texts=texts, model=self.model, **self.kwargs)
             self._log_embed_success(len(texts))
-            return response.embeddings
+            return cast(List[List[float]], response.embeddings)
 
         except Exception as e:
             self._handle_embed_error("Voyage AI", e)
+            raise
 
 
 class JinaEmbedding(BaseAPIEmbedding):
@@ -428,6 +435,7 @@ def embed_sync(self, texts: List[str]) -> List[List[float]]:
 
         except Exception as e:
             self._handle_embed_error("Jina AI", e)
+            raise
 
 
 class MistralEmbedding(BaseAPIEmbedding):
@@ -470,10 +478,11 @@ def embed_sync(self, texts: List[str]) -> List[List[float]]:
 
             embeddings = [item.embedding for item in response.data]
             self._log_embed_success(len(texts))
-            return embeddings
+            return cast(List[List[float]], embeddings)
 
         except Exception as e:
             self._handle_embed_error("Mistral AI", e)
+            raise
 
 
 class CohereEmbedding(BaseAPIEmbedding):
@@ -520,12 +529,18 @@ def __init__(
     def embed_sync(self, texts: List[str]) -> List[List[float]]:
         """텍스트들을 임베딩 (동기)"""
         try:
-            response = self.client.embed(
-                texts=texts, model=self.model, input_type=self.input_type, **self.kwargs
+            response = cast(
+                Any,
+                self.client.embed(
+                    texts=texts,
+                    model=self.model,
+                    input_type=self.input_type,
+                    **self.kwargs,
+                ),
             )
-
             self._log_embed_success(len(texts))
-            return response.embeddings
+            return cast(List[List[float]], response.embeddings)
 
         except Exception as e:
             self._handle_embed_error("Cohere", e)
+            raise

src/beanllm/domain/embeddings/factory.py

@@ -3,9 +3,9 @@
 """
 
 import os
-from typing import List, Optional, Union
+from typing import List, Optional, Union, cast
 
-from .api.providers import (
+from beanllm.domain.embeddings.api.providers import (
     CohereEmbedding,
     GeminiEmbedding,
     JinaEmbedding,
@@ -14,7 +14,7 @@
     OpenAIEmbedding,
     VoyageEmbedding,
 )
-from .base import BaseEmbedding
+from beanllm.domain.embeddings.base import BaseEmbedding
 
 try:
     from beanllm.utils.logging import get_logger
@@ -113,7 +113,7 @@ class Embedding:
         "cohere": "COHERE_API_KEY",
     }
 
-    def __new__(cls, model: str, provider: Optional[str] = None, **kwargs) -> BaseEmbedding:
+    def __new__(cls, model: str, provider: Optional[str] = None, **kwargs) -> "Embedding":
         """
         Embedding 인스턴스 생성 (자동 provider 감지)
 
@@ -144,7 +144,15 @@ def __new__(cls, model: str, provider: Optional[str] = None, **kwargs) -> BaseEm
             )
 
         embedding_class = cls.PROVIDERS[provider]
-        return embedding_class(model=model, **kwargs)
+        return cast("Embedding", embedding_class(model=model, **kwargs))
+
+    async def embed(self, texts: List[str]) -> List[List[float]]:
+        """Not implemented; use Embedding(model=...) to get a concrete provider."""
+        raise NotImplementedError("Use Embedding(model=...) to get a concrete provider")
+
+    def embed_sync(self, texts: List[str]) -> List[List[float]]:
+        """Not implemented; use Embedding(model=...) to get a concrete provider."""
+        raise NotImplementedError("Use Embedding(model=...) to get a concrete provider")
 
     @classmethod
     def _detect_provider(cls, model: str) -> Optional[str]:
@@ -337,8 +345,8 @@ async def embed(
     if isinstance(texts, str):
         texts = [texts]
 
-    embedding = Embedding(model=model, **kwargs)
-    return await embedding.embed(texts)
+    embedding = cast(BaseEmbedding, Embedding(model=model, **kwargs))
+    return cast(List[List[float]], await embedding.embed(texts))
 
 
 def embed_sync(
@@ -370,5 +378,5 @@ def embed_sync(
     if isinstance(texts, str):
         texts = [texts]
 
-    embedding = Embedding(model=model, **kwargs)
-    return embedding.embed_sync(texts)
+    embedding = cast(BaseEmbedding, Embedding(model=model, **kwargs))
+    return cast(List[List[float]], embedding.embed_sync(texts))