feat: 이벤트 로깅 시스템 추가 및 주요 컴포넌트 로깅 통합

.vscode/settings.json

@@ -0,0 +1,3 @@
+{
+    "chatgpt.openOnStartup": false
+}

AI_Web_Scraper/README.md

@@ -39,6 +39,13 @@ pip install -r requirements.txt
 
 자세한 실행 방법은 `run_guide.md` 파일을 참고하세요.
 
+## 로그
+
+- 실행 시 `./logs/run_*.jsonl`에 구조화된 이벤트 로그가 저장됩니다.
+- LLM 내부 추론(Thought) 로그는 기본 비활성화입니다. 필요 시 환경변수로 활성화할 수 있습니다:
+  - `AIWS_SHOW_THOUGHTS=1`
+  - 저장 파일 미리보기 로그: `AIWS_LOG_FILE_PREVIEW=1`
+
 ## 파일 구조
 
 ```

AI_Web_Scraper/ai_agent.py

@@ -9,6 +9,7 @@ from langchain.tools import Tool
 from langchain.memory import ConversationBufferMemory
 from web_scraper import WebScraper
 from google_drive_uploader import GoogleDriveUploader, SimpleDriveSaver
+from event_logger import get_logger, LangChainEventsHandler
 
 class AIAgent:
     def __init__(self, config_path='./config.json'):
@@ -85,6 +86,9 @@ class AIAgent:
             memory=self.memory,
             verbose=True
         )
+        # 콜백 핸들러 구성 (이벤트 로깅)
+        logger = get_logger()
+        self.callback_handler = LangChainEventsHandler(logger) if logger else None
 
     def load_model(self):
         """
@@ -349,9 +353,15 @@ class AIAgent:
         주제별로 웹 검색 → 스크래핑 → 요약 → 저장까지 수행
         반환: [{ topic, response }]
         """
+        from event_logger import get_logger
+        logger = get_logger()
         results = []
         for topic in topics:
+            if logger:
+                logger.log_event("topic_start", topic=topic)
             urls = self._search_urls(topic, k=5)
+            if logger:
+                logger.log_event("search_done", topic=topic, url_count=len(urls))
             collected = []
             for u in urls[:5]:
                 data = self.web_scraper.scrape_website(u)
@@ -372,11 +382,18 @@ class AIAgent:
                 자료:
                 {snippet}
                 """
-                summary = self.llm(prompt)
+                # LangChain 0.1+: __call__ deprecated → use invoke
+                if logger:
+                    logger.log_event("llm_summary_start", topic=topic)
+                summary = self.llm.invoke(prompt)
+                if logger:
+                    logger.log_event("llm_summary_end", topic=topic)
             except Exception as e:
                 summary = f"요약 실패: {e}"
 
             results.append({"topic": topic, "response": summary})
+            if logger:
+                logger.log_event("topic_done", topic=topic)
 
         return results
 
@@ -432,7 +449,10 @@ class AIAgent:
         AI 에이전트를 실행합니다.
         """
         try:
-            response = self.agent.run(task_description)
+            if self.callback_handler:
+                response = self.agent.run(task_description, callbacks=[self.callback_handler])
+            else:
+                response = self.agent.run(task_description)
             return response
         except Exception as e:
             print(f"에이전트 실행 실패: {e}")
@@ -459,7 +479,14 @@ class AIAgent:
         """
 
         try:
-            response = self.llm(prompt)
+            # LangChain 0.1+: __call__ deprecated → use invoke
+            from event_logger import get_logger
+            logger = get_logger()
+            if logger:
+                logger.log_event("llm_topics_start", count=num_topics)
+            response = self.llm.invoke(prompt)
+            if logger:
+                logger.log_event("llm_topics_end", count=num_topics)
             # 응답에서 주제들을 추출 (줄 단위로 분리)
             topics = [line.strip() for line in response.split('\n') if line.strip() and not line.startswith(('1.', '2.', '3.', '-'))]
             # 최대 num_topics개 반환

AI_Web_Scraper/event_logger.py

@@ -0,0 +1,181 @@
+import os
+import json
+import uuid
+import datetime as _dt
+from typing import Any, Dict, Optional
+
+_LOGGER_INSTANCE = None
+
+
+class EventLogger:
+    """
+    Lightweight JSONL + console event logger for the app.
+    Avoids exposing LLM chain-of-thought by default; can be opted-in via config/env.
+    """
+
+    def __init__(self, log_dir: str = "./logs", enable_file: bool = True,
+                 show_llm_thoughts: bool = False, console_level: str = "INFO",
+                 preview_saved_files: bool = False, preview_limit: int = 500):
+        self.log_dir = log_dir
+        self.enable_file = enable_file
+        self.show_llm_thoughts = show_llm_thoughts
+        self.console_level = console_level.upper()
+        self.preview_saved_files = preview_saved_files
+        self.preview_limit = preview_limit
+        self.run_id = _dt.datetime.now().strftime("%Y%m%d_%H%M%S") + "_" + uuid.uuid4().hex[:6]
+        self.file_path = None
+        if enable_file:
+            os.makedirs(log_dir, exist_ok=True)
+            self.file_path = os.path.join(log_dir, f"run_{self.run_id}.jsonl")
+
+    def _now(self) -> str:
+        return _dt.datetime.now().isoformat(timespec='seconds')
+
+    def _console_enabled(self, level: str) -> bool:
+        order = ["DEBUG", "INFO", "WARN", "ERROR"]
+        try:
+            return order.index(level) >= order.index(self.console_level)
+        except Exception:
+            return True
+
+    def log_event(self, event: str, message: Optional[str] = None, **fields: Any) -> None:
+        rec: Dict[str, Any] = {
+            "ts": self._now(),
+            "run_id": self.run_id,
+            "event": event,
+        }
+        if message:
+            rec["message"] = message
+        if fields:
+            rec.update(fields)
+
+        # console (pretty one-liner)
+        if self._console_enabled("INFO"):
+            kv = " ".join(
+                f"{k}={str(v)[:120]}" for k, v in fields.items() if v is not None
+            )
+            line = f"[{rec['ts']}] {event}"
+            if message:
+                line += f" | {message}"
+            if kv:
+                line += f" | {kv}"
+            print(line)
+
+        # JSONL file
+        if self.enable_file and self.file_path:
+            try:
+                with open(self.file_path, "a", encoding="utf-8") as f:
+                    f.write(json.dumps(rec, ensure_ascii=False) + "\n")
+            except Exception:
+                # Do not crash on logging errors
+                pass
+
+
+def init_from_config(config: Dict[str, Any]) -> EventLogger:
+    global _LOGGER_INSTANCE
+    lg = config.get("logging", {}) if isinstance(config, dict) else {}
+    log_dir = lg.get("log_dir", "./logs")
+    enable_file = bool(lg.get("log_to_file", True))
+    console_level = str(lg.get("console_level", "INFO")).upper()
+    # env override for showing LLM thoughts
+    env_flag = os.environ.get("AIWS_SHOW_THOUGHTS")
+    show_llm_thoughts = bool(lg.get("show_thoughts", False)) or (str(env_flag).lower() in ("1", "true", "yes"))
+    from os import environ
+    preview_files = bool(lg.get("preview_saved_files", False)) or (str(environ.get("AIWS_LOG_FILE_PREVIEW")).lower() in ("1", "true", "yes"))
+    preview_limit = int(lg.get("preview_limit", 500)) if str(lg.get("preview_limit", "")).isdigit() else 500
+    _LOGGER_INSTANCE = EventLogger(
+        log_dir=log_dir,
+        enable_file=enable_file,
+        show_llm_thoughts=show_llm_thoughts,
+        console_level=console_level,
+        preview_saved_files=preview_files,
+        preview_limit=preview_limit,
+    )
+    _LOGGER_INSTANCE.log_event("run_start", message="Application run started")
+    return _LOGGER_INSTANCE
+
+
+def get_logger() -> Optional[EventLogger]:
+    return _LOGGER_INSTANCE
+
+
+# LangChain callback handler
+try:
+    from langchain.callbacks.base import BaseCallbackHandler
+except Exception:  # pragma: no cover - fallback for newer versions
+    try:
+        from langchain_core.callbacks.base import BaseCallbackHandler  # type: ignore
+    except Exception:
+        BaseCallbackHandler = object  # minimal fallback
+
+
+class LangChainEventsHandler(BaseCallbackHandler):
+    def __init__(self, logger: EventLogger):
+        super().__init__()
+        self.logger = logger
+
+    # Chains
+    def on_chain_start(self, serialized, inputs, **kwargs):
+        name = serialized.get("name") if isinstance(serialized, dict) else str(serialized)
+        self.logger.log_event("chain_start", name=name, inputs=_truncate(inputs))
+
+    def on_chain_end(self, outputs, **kwargs):
+        self.logger.log_event("chain_end", outputs=_truncate(outputs))
+
+    # LLMs
+    def on_llm_start(self, serialized, prompts, **kwargs):
+        if self.logger.show_llm_thoughts:
+            self.logger.log_event("llm_start", prompts=_truncate(prompts))
+        else:
+            self.logger.log_event("llm_start", message="prompt issued", prompt_count=len(prompts) if prompts else 0)
+
+    def on_llm_end(self, response, **kwargs):
+        try:
+            if self.logger.show_llm_thoughts:
+                texts = [g[0].text for g in response.generations]  # type: ignore[attr-defined]
+                self.logger.log_event("llm_end", outputs=_truncate(texts))
+            else:
+                # token/length only when possible
+                usage = getattr(response, 'llm_output', None) or {}
+                self.logger.log_event("llm_end", message="llm completed", meta=_truncate(usage))
+        except Exception:
+            self.logger.log_event("llm_end")
+
+    # Tools
+    def on_tool_start(self, serialized, input_str, **kwargs):
+        name = serialized.get("name") if isinstance(serialized, dict) else str(serialized)
+        self.logger.log_event("tool_start", name=name, input=_truncate(input_str))
+
+    def on_tool_end(self, output, **kwargs):
+        self.logger.log_event("tool_end", output=_truncate(output))
+
+    # Agent actions
+    def on_agent_action(self, action, **kwargs):
+        try:
+            self.logger.log_event(
+                "agent_action",
+                tool=getattr(action, 'tool', None),
+                tool_input=_truncate(getattr(action, 'tool_input', None)),
+                log=_truncate(getattr(action, 'log', None)) if self.logger.show_llm_thoughts else None,
+            )
+        except Exception:
+            self.logger.log_event("agent_action")
+
+    def on_agent_finish(self, finish, **kwargs):
+        try:
+            out = getattr(finish, 'return_values', {}).get('output')
+            self.logger.log_event("agent_finish", output=_truncate(out))
+        except Exception:
+            self.logger.log_event("agent_finish")
+
+
+def _truncate(obj: Any, limit: int = 800) -> Any:
+    try:
+        s = obj if isinstance(obj, str) else json.dumps(obj, ensure_ascii=False)
+        return s if len(s) <= limit else (s[:limit] + "…")
+    except Exception:
+        try:
+            s = str(obj)
+            return s if len(s) <= limit else (s[:limit] + "…")
+        except Exception:
+            return None

AI_Web_Scraper/google_drive_uploader.py

@@ -5,6 +5,7 @@ from googleapiclient.http import MediaFileUpload
 from google.oauth2.credentials import Credentials
 from google_auth_oauthlib.flow import InstalledAppFlow
 from google.auth.transport.requests import Request
+from event_logger import get_logger
 
 class GoogleDriveUploader:
     def __init__(self, config_path='./config.json'):
@@ -52,6 +53,7 @@ class GoogleDriveUploader:
         """
         파일을 Google Drive에 업로드
         """
+        logger = get_logger()
         if self.service is None:
             raise RuntimeError('Google Drive API가 초기화되지 않았습니다. credentials.json과 folder_id를 설정하세요.')
 
@@ -71,11 +73,15 @@ class GoogleDriveUploader:
                 media_body=media,
                 fields='id'
             ).execute()
-
-            print(f'파일 업로드 완료: {file_name} (ID: {file.get("id")})')
-            return file.get('id')
+            fid = file.get('id')
+            print(f'파일 업로드 완료: {file_name} (ID: {fid})')
+            if logger:
+                logger.log_event("drive_upload", name=file_name, id=fid)
+            return fid
         except Exception as e:
             print(f'업로드 실패: {e}')
+            if logger:
+                logger.log_event("drive_upload_error", name=file_name, error=str(e))
             return None
 
     def upload_data_as_json(self, data, filename='collected_data.json'):
@@ -83,6 +89,7 @@ class GoogleDriveUploader:
         데이터를 JSON 파일로 변환하여 업로드
         """
         import tempfile
+        logger = get_logger()
 
         with tempfile.NamedTemporaryFile(mode='w', suffix='.json', delete=False) as f:
             json.dump(data, f, ensure_ascii=False, indent=2)
@@ -90,9 +97,19 @@ class GoogleDriveUploader:
 
         try:
             file_id = self.upload_file(temp_path, filename)
+            logger = get_logger()
+            if logger and logger.preview_saved_files:
+                try:
+                    with open(temp_path, 'r', encoding='utf-8') as rf:
+                        content = rf.read(logger.preview_limit)
+                    logger.log_event("file_preview", name=filename, preview=content)
+                except Exception:
+                    pass
             return file_id
         finally:
             os.unlink(temp_path)
+            if logger:
+                logger.log_event("tempfile_cleanup", path=temp_path)
 
     def list_files(self):
         """
@@ -109,9 +126,15 @@ class GoogleDriveUploader:
             ).execute()
 
             items = results.get('files', [])
+            logger = get_logger()
+            if logger:
+                logger.log_event("drive_list", count=len(items))
             return items
         except Exception as e:
             print(f'파일 목록 조회 실패: {e}')
+            logger = get_logger()
+            if logger:
+                logger.log_event("drive_list_error", error=str(e))
             return []
 
 class SimpleDriveSaver:
@@ -132,9 +155,22 @@ class SimpleDriveSaver:
             with open(filepath, 'w', encoding='utf-8') as f:
                 json.dump(data, f, ensure_ascii=False, indent=2)
             print(f'데이터 저장 완료: {filepath}')
+            logger = get_logger()
+            if logger:
+                logger.log_event("file_saved", path=filepath, bytes=os.path.getsize(filepath))
+                if logger.preview_saved_files:
+                    try:
+                        with open(filepath, 'r', encoding='utf-8') as rf:
+                            preview = rf.read(logger.preview_limit)
+                        logger.log_event("file_preview", path=filepath, preview=preview)
+                    except Exception:
+                        pass
             return filepath
         except Exception as e:
             print(f'저장 실패: {e}')
+            logger = get_logger()
+            if logger:
+                logger.log_event("file_save_error", path=filepath, error=str(e))
             return None
 
     def save_text_data(self, data, filename='collected_data.txt'):
@@ -150,9 +186,15 @@ class SimpleDriveSaver:
                 else:
                     f.write(str(data))
             print(f'텍스트 데이터 저장 완료: {filepath}')
+            logger = get_logger()
+            if logger:
+                logger.log_event("file_saved", path=filepath, bytes=os.path.getsize(filepath))
             return filepath
         except Exception as e:
             print(f'저장 실패: {e}')
+            logger = get_logger()
+            if logger:
+                logger.log_event("file_save_error", path=filepath, error=str(e))
             return None
 
 def save_to_drive_simple(data, filename='collected_data.json', mount_path='/content/drive/MyDrive/AI_Data'):

AI_Web_Scraper/main.py

@@ -3,6 +3,7 @@ import json
 from model_downloader import download_model
 from ai_agent import AIAgent
 import argparse
+from event_logger import init_from_config, get_logger
 
 def main():
     parser = argparse.ArgumentParser(description='AI 웹 정보 수집 시스템')
@@ -19,8 +20,10 @@ def main():
     with open(args.config, 'r') as f:
         config = json.load(f)
 
-    # 2. 모델 다운로드 (필요한 경우)
+    # 2. 로깅 초기화 및 모델 다운로드 (필요한 경우)
+    logger = init_from_config(config)
     print("모델 확인 중...")
+    logger.log_event("model_check")
     model, tokenizer = download_model(args.config)
 
     # 모델 로딩은 AIAgent에서 수행하므로, 다운로드만 성공해도 계속 진행
@@ -42,22 +45,28 @@ def main():
         with open(args.config, 'w') as f:
             json.dump(config, f, indent=2)
         print(f"저장 경로 설정됨: {args.save_path}")
+        logger.log_event("save_path_set", path=args.save_path)
 
     # 3. AI 에이전트 초기화
     print("AI 에이전트 초기화 중...")
+    logger.log_event("agent_init_start")
     agent = AIAgent(args.config)
+    logger.log_event("agent_init_done")
 
     # 3. 주제 결정
     if args.auto_topics or args.topics is None:
         print("AI가 스스로 주제를 선정합니다...")
         topics = agent.generate_topics(num_topics=3)
         print(f"선정된 주제: {topics}")
+        logger.log_event("auto_topics", topics=topics)
     else:
         topics = args.topics
+        logger.log_event("user_topics", topics=topics)
 
     # 4. 정보 수집 실행
     print(f"다음 주제들에 대해 정보를 수집합니다: {topics}")
     results = agent.collect_information(topics)
+    logger.log_event("collect_done", topic_count=len(topics))
 
     # 5. 결과 출력
     print("\n=== 수집 결과 ===")
@@ -68,6 +77,7 @@ def main():
 
     # 6. 정리
     agent.close()
+    logger.log_event("run_complete", message="Program finished")
     print("프로그램 완료")
 
 if __name__ == "__main__":

AI_Web_Scraper/model_downloader.py

@@ -3,6 +3,7 @@ import json
 from typing import Tuple, Optional
 from transformers import AutoTokenizer
 from huggingface_hub import snapshot_download
+from event_logger import get_logger
 
 def download_model(config_path: str = './config.json') -> Tuple[Optional[object], Optional[AutoTokenizer]]:
     """
@@ -22,6 +23,9 @@ def download_model(config_path: str = './config.json') -> Tuple[Optional[object]
 
     os.makedirs(local_path, exist_ok=True)
     print(f"모델 {model_name}을 {local_path}에 다운로드 중...")
+    logger = get_logger()
+    if logger:
+        logger.log_event("model_download_start", model=model_name, path=local_path)
 
     try:
         # 인증 토큰(필요 시) 지원: 환경변수 HF_TOKEN 사용
@@ -45,9 +49,13 @@ def download_model(config_path: str = './config.json') -> Tuple[Optional[object]
             print("토크나이저 확인 실패(계속 진행): 로컬 경로에 tokenizer 파일이 없을 수 있습니다.")
 
         print(f"모델 다운로드 완료: {local_path}")
+        if logger:
+            logger.log_event("model_download_done", path=local_path)
         return None, tokenizer
     except Exception as e:
         print(f"모델 다운로드 실패: {e}")
+        if logger:
+            logger.log_event("model_download_error", error=str(e))
         return None, None
 
 if __name__ == "__main__":

AI_Web_Scraper/run_guide.md

@@ -139,6 +139,19 @@ os.environ["HF_TOKEN"] = "hf_********************************"
 !free -h
 ```
 
+### 5.4 이벤트 로그(JSONL)
+
+- 실행 시 `./logs/` 폴더에 `run_YYYYMMDD_HHMMSS_xxxxxx.jsonl` 파일이 생성됩니다.
+- 각 단계(모델 다운로드, 에이전트/LLM 호출, 도구 실행, 스크래핑 시작/완료, 파일 저장 등)가 구조화된 레코드로 기록됩니다.
+- 실시간 확인 예시:
+  ```bash
+  tail -f logs/run_*.jsonl
+  ```
+- LLM의 내부 추론(Thought) 노출은 기본 비활성화입니다. 필요 시 다음 환경변수로 활성화할 수 있습니다(권장하지 않음):
+  ```bash
+  export AIWS_SHOW_THOUGHTS=1
+  ```
+
 ## 6. 문제 해결
 
 ### 6.1 모델 다운로드 실패

AI_Web_Scraper/web_scraper.py

@@ -3,6 +3,7 @@ from bs4 import BeautifulSoup
 import json
 import time
 import os
+from event_logger import get_logger
 
 try:
     from selenium import webdriver
@@ -64,7 +65,10 @@ class WebScraper:
         """
         웹사이트에서 정보를 수집합니다.
         """
+        logger = get_logger()
         try:
+            if logger:
+                logger.log_event("scrape_start", url=url)
             if self.use_selenium and self.driver is not None:
                 self.driver.get(url)
                 time.sleep(self.delay)
@@ -88,9 +92,13 @@ class WebScraper:
                 'content': text_content[:5000],
                 'timestamp': time.time()
             }
+            if logger:
+                logger.log_event("scrape_done", url=url, title=title, size=len(data['content']))
             return data
         except Exception as e:
             print(f"스크래핑 실패: {url} - {e}")
+            if logger:
+                logger.log_event("scrape_error", url=url, error=str(e))
             return None
 
     def crawl_multiple_pages(self, start_urls, keywords=None):