botrun-flow-lang 5.11.11__py3-none-any.whl → 5.11.281__py3-none-any.whl

botrun_flow_lang/api/line_bot_api.py

@@ -100,6 +100,11 @@ BOTRUN_FRONT_URL = os.getenv("BOTRUN_FRONT_URL", None)
 SUBSIDY_API_TOKEN = os.getenv("SUBSIDY_API_TOKEN", None)
 SUBSIDY_API_URL = os.getenv("SUBSIDY_API_URL", "https://p271-subsidy-ie7vwovclq-de.a.run.app/v1/generateContent")
 
+# BigQuery Token Logging API 相關環境變數
+BIGQUERY_TOKEN_LOG_API_URL = os.getenv("BIGQUERY_TOKEN_LOG_API_URL", "http://localhost:8002/api/v1/logs/text")
+BIGQUERY_TOKEN_LOG_ENABLED = os.getenv("BIGQUERY_TOKEN_LOG_ENABLED", "true").lower() == "true"
+SUBSIDY_LINE_BOT_MODEL_NAME = os.getenv("SUBSIDY_LINE_BOT_MODEL_NAME", "gemini-2.0-flash-thinking-exp")
+
 # 全局變數
 # 用於追蹤正在處理訊息的使用者，避免同一使用者同時發送多條訊息造成處理衝突
 _processing_users = set()
@@ -202,6 +207,111 @@ async def log_to_bigquery(
         )
 
 
+async def log_tokens_to_bigquery(
+    user_id: str,
+    display_name: str,
+    log_content: str,
+    model: str,
+    input_tokens: int | None,
+    output_tokens: int | None,
+    total_tokens: int | None,
+    request: Request,
+    session_id: str = "",
+) -> None:
+    """
+    記錄 token 使用量到 BigQuery logging API
+
+    Args:
+        user_id: LINE 使用者 ID
+        display_name: 使用者顯示名稱
+        log_content: 使用者輸入訊息
+        model: 使用的 AI 模型
+        input_tokens: 輸入 token 數量
+        output_tokens: 輸出 token 數量
+        total_tokens: 總 token 數量
+        request: FastAPI Request 物件
+        session_id: Session ID (可選，預設使用 user_id)
+    """
+    # 檢查功能是否啟用
+    if not BIGQUERY_TOKEN_LOG_ENABLED:
+        logging.debug("[Token Logger] BigQuery token logging is disabled")
+        return
+
+    start_time = time.time()
+
+    try:
+        tz = pytz.timezone("Asia/Taipei")
+        current_time = datetime.now(tz)
+
+        # 組裝 payload
+        payload = {
+            "action_details": log_content,
+            "action_type": "call_subsidy_api",
+            "botrun": "subsidy_line_bot",
+            "dataset_name": os.getenv("BOTRUN_LOG_DATASET_NAME", "subsidy_line_bot"),
+            "department": os.getenv("BOTRUN_LOG_DEPARTMENT", "subsidy_line_bot"),
+            "developer": "",
+            "domain_name": "subsidy_line_bot",
+            "input_tokens": input_tokens,
+            "model": model,
+            "output_tokens": output_tokens,
+            "resource_id": json.dumps({
+                "user_id": user_id,
+                "timestamp": current_time.strftime("%Y-%m-%dT%H:%M:%SZ")
+            }),
+            "session_id": session_id or user_id,
+            "total_tokens": total_tokens,
+            "user_agent": request.headers.get("user-agent", "Line Platform"),
+            "user_name": display_name
+        }
+
+        logging.info(
+            f"[Token Logger] Logging tokens for user {display_name} ({user_id}): "
+            f"input={input_tokens}, output={output_tokens}, total={total_tokens}"
+        )
+
+        # 使用 aiohttp 非同步呼叫 API
+        timeout = aiohttp.ClientTimeout(total=10)  # 10 秒超時
+        async with aiohttp.ClientSession(timeout=timeout) as session:
+            async with session.post(
+                BIGQUERY_TOKEN_LOG_API_URL,
+                json=payload,
+                headers={"Content-Type": "application/json"}
+            ) as response:
+                response_text = await response.text()
+
+                if response.status == 200:
+                    elapsed_time = time.time() - start_time
+                    logging.info(
+                        f"[Token Logger] Successfully logged tokens to BigQuery for user "
+                        f"{display_name} ({user_id}), elapsed time: {elapsed_time:.3f}s"
+                    )
+                else:
+                    logging.error(
+                        f"[Token Logger] Failed to log tokens, API returned status {response.status}: {response_text}"
+                    )
+
+    except asyncio.TimeoutError:
+        elapsed_time = time.time() - start_time
+        logging.error(
+            f"[Token Logger] Timeout while logging tokens for user {display_name} ({user_id}), "
+            f"elapsed time: {elapsed_time:.3f}s"
+        )
+    except aiohttp.ClientError as e:
+        elapsed_time = time.time() - start_time
+        logging.error(
+            f"[Token Logger] Network error while logging tokens for user {display_name} ({user_id}): {e}, "
+            f"elapsed time: {elapsed_time:.3f}s"
+        )
+    except Exception as e:
+        elapsed_time = time.time() - start_time
+        logging.error(
+            f"[Token Logger] Unexpected error while logging tokens for user {display_name} ({user_id}): {e}, "
+            f"elapsed time: {elapsed_time:.3f}s"
+        )
+        traceback.print_exc()
+
+
 def get_prompt_from_google_doc(tag_name: str, fallback_prompt: str = ""):
     """
     從 Google 文件中提取指定標籤的內容
@@ -736,13 +846,37 @@ async def handle_message(
     _processing_users.add(user_id)
 
     try:
-        reply_text, related_questions = await get_reply_text(
+        reply_text, related_questions, usage_metadata = await get_reply_text(
             user_message, user_id, display_name, request
         )
         logging.info(
             f"[Line Bot Webhook: handle_message] Total response length: {len(reply_text)}"
         )
 
+        # 記錄 token 使用量到 BigQuery (非阻塞式)
+        if usage_metadata:
+            # 把 user_message 跟 reply_text 合併成 json 格式紀錄
+            log_content = json.dumps(
+                {
+                    "user_message": user_message,
+                    "reply_text": reply_text
+                },
+                ensure_ascii=False
+            )
+            asyncio.create_task(
+                log_tokens_to_bigquery(
+                    user_id=user_id,
+                    display_name=display_name,
+                    log_content=log_content,
+                    model=usage_metadata.get("model", SUBSIDY_LINE_BOT_MODEL_NAME),
+                    input_tokens=usage_metadata.get("promptTokenCount", None),
+                    output_tokens=usage_metadata.get("candidatesTokenCount", None),
+                    total_tokens=usage_metadata.get("totalTokenCount", None),
+                    request=request,
+                    session_id=user_id,
+                )
+            )
+
         # 將長訊息分段，每段不超過 LINE_MAX_MESSAGE_LENGTH
         message_chunks = []
         remaining_text = reply_text
@@ -934,7 +1068,7 @@ async def get_reply_text(
     user_id: str,
     display_name: str,
     request: Request,
-) -> tuple[str, list]:
+) -> tuple[str, list, dict]:
     """
     使用外部 API 處理使用者訊息並回傳回覆內容
 
@@ -945,7 +1079,7 @@ async def get_reply_text(
         request (Request): FastAPI request 物件，用於記錄到 BigQuery
 
     Returns:
-        tuple[str, list]: 包含回覆訊息和相關問題的元組
+        tuple[str, list, dict]: 包含回覆訊息、相關問題和 token 使用量的元組
     """
     start_time = time.time()
 
@@ -961,6 +1095,15 @@ async def get_reply_text(
             system_instruction=system_instruction
         )
 
+        # 提取 token 使用量資訊
+        usage_metadata = api_response.get("usageMetadata", {})
+        logging.info(
+            f"[Line Bot Webhook: get_reply_text] Token usage: "
+            f"input={usage_metadata.get('promptTokenCount', 0)}, "
+            f"output={usage_metadata.get('candidatesTokenCount', 0)}, "
+            f"total={usage_metadata.get('totalTokenCount', 0)}"
+        )
+
         # 從 API 回應中提取文字內容
         full_response = ""
         if "candidates" in api_response and len(api_response["candidates"]) > 0:
@@ -1003,7 +1146,7 @@ async def get_reply_text(
             f"[Line Bot Webhook: get_reply_text] total took {time.time() - start_time:.3f}s"
         )
 
-        return full_response, related_questions
+        return full_response, related_questions, usage_metadata
 
     except Exception as e:
         import traceback
@@ -1012,7 +1155,7 @@ async def get_reply_text(
 
         # 返回錯誤訊息
         error_message = "抱歉，處理您的訊息時遇到問題，請稍後再試。"
-        return error_message, []
+        return error_message, [], {}
 
 
 async def handle_feedback(

botrun_flow_lang/langgraph_agents/agents/langgraph_react_agent.py

@@ -142,11 +142,42 @@ def get_react_agent_model_name(model_name: str = ""):
 
 ANTHROPIC_MAX_TOKENS = 64000
 GEMINI_MAX_TOKENS = 32000
+TAIDE_MAX_TOKENS = 8192
 
 
 def get_react_agent_model(model_name: str = ""):
     final_model_name = get_react_agent_model_name(model_name).strip()
 
+    # 處理 taide/ 前綴的模型
+    if final_model_name.startswith("taide/"):
+        taide_api_key = os.getenv("TAIDE_API_KEY", "")
+        taide_base_url = os.getenv("TAIDE_BASE_URL", "")
+
+        if not taide_api_key or not taide_base_url:
+            raise ValueError(
+                f"Model name starts with 'taide/' but TAIDE_API_KEY or TAIDE_BASE_URL not set. "
+                f"Both environment variables are required for: {final_model_name}"
+            )
+
+        # 取得 taide/ 後面的模型名稱
+        taide_model_name = final_model_name[len("taide/"):]
+
+        if not taide_model_name:
+            raise ValueError(
+                f"Invalid taide model format: {final_model_name}. "
+                "Expected format: taide/<model_name>"
+            )
+
+        model = ChatOpenAI(
+            openai_api_key=taide_api_key,
+            openai_api_base=taide_base_url,
+            model_name=taide_model_name,
+            temperature=0,
+            max_tokens=TAIDE_MAX_TOKENS,
+        )
+        logger.info(f"model ChatOpenAI (TAIDE) {taide_model_name} @ {taide_base_url}")
+        return model
+
     # 處理 vertexai/ 前綴的模型
     if final_model_name.startswith("vertex-ai/"):
         vertex_project = os.getenv("VERTEX_AI_LANGCHAIN_PROJECT", "")

{botrun_flow_lang-5.11.11.dist-info → botrun_flow_lang-5.11.281.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: botrun-flow-lang
-Version: 5.11.11
+Version: 5.11.281
 Summary: A flow language for botrun
 Author-email: sebastian-hsu <sebastian.hsu@gmail.com>
 License: MIT

{botrun_flow_lang-5.11.11.dist-info → botrun_flow_lang-5.11.281.dist-info}/RECORD

@@ -10,7 +10,7 @@ botrun_flow_lang/api/botrun_back_api.py,sha256=mE2NSejaYIiE0L9GmNJbLc_FRWCy6BXlc
 botrun_flow_lang/api/flow_api.py,sha256=DcxuoGE1OcbTgLSYKZ2SO9IdcH3UB5Ik3cVmX3v3-Po,108
 botrun_flow_lang/api/hatch_api.py,sha256=qZG-Wwi_8SHPuWNfbt-dhz-O41VYetTxrJzcVjHbJCo,15913
 botrun_flow_lang/api/langgraph_api.py,sha256=zqu0xeTiy2Pr4UL6vvGqVVAy2KX3ZUn1uzcq-Tfb_aM,29291
-botrun_flow_lang/api/line_bot_api.py,sha256=ECCGgl-rbac4GwJ7O39jzrvTPR-M_tShQjQFm1esuag,49153
+botrun_flow_lang/api/line_bot_api.py,sha256=__Rul_JWd7KwtgGIyziOKxD9PSconkjuN0dmI8JEwjg,54930
 botrun_flow_lang/api/model_api.py,sha256=bXemey_XUUdylZwh7Z10eksoBWe9xSa8I9TEL7jIBtE,9483
 botrun_flow_lang/api/rate_limit_api.py,sha256=SkpjfvShHRdP5XJzy3DdrH4jLtdYAEHROGBMBkC9OIY,948
 botrun_flow_lang/api/routes.py,sha256=rd0IoMsteJT9BO3MQuyXirhPQbas6OeiKaEC8Yf2SZs,1570
@@ -24,7 +24,7 @@ botrun_flow_lang/api/youtube_api.py,sha256=R384jNRheMKnDyzvlLnbzackipZhiLYTZl4w4
 botrun_flow_lang/langgraph_agents/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 botrun_flow_lang/langgraph_agents/agents/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 botrun_flow_lang/langgraph_agents/agents/agent_runner.py,sha256=fOZgHDsCA_EDTTGQFBmhGUhpfLB3m_N6YW2UHgMpKBg,6241
-botrun_flow_lang/langgraph_agents/agents/langgraph_react_agent.py,sha256=Vwf2HbpOuulx1nS0ReoE4OkVndiLgK5aByusPvGAXA8,28827
+botrun_flow_lang/langgraph_agents/agents/langgraph_react_agent.py,sha256=4fK_hMoUAqcEYv7rrHbAx6PFsJ7UcvGI0G2OgWhVhnw,29972
 botrun_flow_lang/langgraph_agents/agents/search_agent_graph.py,sha256=6fz-ewLQGacEx-uqGfF3-go9FdiioiMzW_sfANzYTcI,31182
 botrun_flow_lang/langgraph_agents/agents/agent_tools/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 botrun_flow_lang/langgraph_agents/agents/agent_tools/step_planner.py,sha256=CgEhfGR28Rq7ui9cKxj_DczfNfjJNXIP9DXQNIwBLv0,2350
@@ -95,6 +95,6 @@ botrun_flow_lang/utils/yaml_utils.py,sha256=1A6PSEE8TM0HSD_6l-fhUsjYnXJcrEKuPgot
 botrun_flow_lang/utils/clients/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 botrun_flow_lang/utils/clients/rate_limit_client.py,sha256=TRpA56OKrfYsoLoJ-TPYlC7Znp9s267-u6CX6BLyVko,8349
 botrun_flow_lang/utils/clients/token_verify_client.py,sha256=BtrfLvMe-DtS8UKeDhaIkVKDZHphZVP7kyqXn9jhXEc,5740
-botrun_flow_lang-5.11.11.dist-info/METADATA,sha256=iaavVUSFaUN3cAkwoJYi1X8XPw0Xd_heaMtQM6WkBlY,6191
-botrun_flow_lang-5.11.11.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-botrun_flow_lang-5.11.11.dist-info/RECORD,,
+botrun_flow_lang-5.11.281.dist-info/METADATA,sha256=BN-Ufk4sSRSkuAr3c8mIbKfh78gZX6yOMMxzq4oRSMo,6192
+botrun_flow_lang-5.11.281.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+botrun_flow_lang-5.11.281.dist-info/RECORD,,

{botrun_flow_lang-5.11.11.dist-info → botrun_flow_lang-5.11.281.dist-info}/WHEEL

@@ -1,4 +1,4 @@
 Wheel-Version: 1.0
-Generator: hatchling 1.27.0
+Generator: hatchling 1.28.0
 Root-Is-Purelib: true
 Tag: py3-none-any