pip install "langchain-google-community[bigquery]" langchain langchain-google-genai langgraph

import os
from datetime import datetime

from langchain.agents import create_agent
from langchain.messages import HumanMessage
from langchain.tools import tool
from langchain_google_community.callbacks.bigquery_callback import (
    BigQueryCallbackHandler,
    BigQueryLoggerConfig,
)
from langchain_google_genai import ChatGoogleGenerativeAI

# 1. 为智能体定义工具
@tool
def get_current_time() -> str:
    """返回当前本地时间。"""
    now = datetime.now()
    return f"当前时间：{now.strftime('%I:%M:%S %p')}，日期：{now.strftime('%B %d, %Y')}"

@tool
def get_weather(city: str) -> str:
    """返回特定城市的当前天气。"""
    # 模拟天气数据（生产环境中使用真实 API）
    weather_data = {
        "new york": {"temp": 22, "condition": "晴朗"},
        "tokyo": {"temp": 24, "condition": "晴天"},
        "london": {"temp": 14, "condition": "阴天"},
    }
    city_lower = city.lower()
    if city_lower in weather_data:
        data = weather_data[city_lower]
        return f"{city.title()} 的天气：{data['temp']}°C，{data['condition']}"
    return f"'{city}' 的天气数据不可用。"

@tool
def convert_currency(amount: float, from_currency: str, to_currency: str) -> str:
    """在不同货币之间转换金额。"""
    rates = {"USD": 1.0, "EUR": 1.08, "GBP": 1.27, "JPY": 0.0067}
    from_curr, to_curr = from_currency.upper(), to_currency.upper()
    if from_curr not in rates or to_curr not in rates:
        return f"未知货币"
    result = amount * rates[from_curr] / rates[to_curr]
    return f"{amount} {from_curr} = {result:.2f} {to_curr}"

def run_agent_example(project_id: str):
    """运行带有 BigQuery 日志记录的 LangGraph 智能体。"""

    dataset_id = "agent_analytics"
    table_id = "agent_events_v2"

    # 2. 配置回调处理器
    config = BigQueryLoggerConfig(
        batch_size=1,
        batch_flush_interval=0.5,
    )

    handler = BigQueryCallbackHandler(
        project_id=project_id,
        dataset_id=dataset_id,
        table_id=table_id,
        config=config,
        graph_name="travel_assistant",  # 启用 LangGraph 跟踪
    )

    # 3. 创建 LLM 和智能体
    # 对于 Vertex AI 使用 project 参数，对于 Gemini Developer API 使用 api_key
    llm = ChatGoogleGenerativeAI(
        model="gemini-2.5-flash",
        project=project_id,  # 用于 Vertex AI
    )
    tools = [get_current_time, get_weather, convert_currency]
    agent = create_agent(llm, tools)

    # 4. 使用 graph_context 运行以获取 GRAPH_START/GRAPH_END 事件
    query = "现在几点了？东京的天气怎么样？100 美元等于多少欧元？"

    run_metadata = {
        "session_id": "session-001",
        "user_id": "user-123",
        "agent": "travel_assistant",
    }

    with handler.graph_context("travel_assistant", metadata=run_metadata):
        result = agent.invoke(
            {"messages": [HumanMessage(content=query)]},
            config={
                "callbacks": [handler],
                "metadata": run_metadata,
            },
        )

    print(f"响应：{result['messages'][-1].content}")

    # 5. 清理
    handler.shutdown()

if __name__ == "__main__":
    project_id = os.environ.get("GCP_PROJECT_ID", "your-project-id")
    run_agent_example(project_id)

from langchain_google_community.callbacks.bigquery_callback import (
    BigQueryCallbackHandler,
    BigQueryLoggerConfig,
)

# 1. 配置 BigQueryLoggerConfig
config = BigQueryLoggerConfig(
    enabled=True,
    event_allowlist=["LLM_REQUEST", "LLM_RESPONSE"],  # 仅记录这些特定事件
    shutdown_timeout=10.0,  # 退出时最多等待 10 秒让日志刷新
    max_content_length=500,  # 将内容截断为 500 个字符
)

# 2. 初始化回调处理器
handler = BigQueryCallbackHandler(
    project_id="your-project-id",
    dataset_id="your_dataset",
    table_id="your_table",
    config=config,
)

CREATE TABLE `your-gcp-project-id.adk_agent_logs.agent_events_v2`
(
  timestamp TIMESTAMP NOT NULL OPTIONS(description="事件发生的 UTC 时间戳。"),
  event_type STRING OPTIONS(description="事件的类别。"),
  agent STRING OPTIONS(description="智能体的名称。"),
  session_id STRING OPTIONS(description="对话会话的唯一标识符。"),
  invocation_id STRING OPTIONS(description="单轮调用的唯一标识符。"),
  user_id STRING OPTIONS(description="最终用户的标识符。"),
  trace_id STRING OPTIONS(description="OpenTelemetry 跟踪 ID。"),
  span_id STRING OPTIONS(description="OpenTelemetry 跨度 ID。"),
  parent_span_id STRING OPTIONS(description="OpenTelemetry 父跨度 ID。"),
  content JSON OPTIONS(description="事件的主要负载。"),
  content_parts ARRAY<STRUCT<
    mime_type STRING,
    uri STRING,
    object_ref STRUCT<
      uri STRING,
      version STRING,
      authorizer STRING,
      details JSON
    >,
    text STRING,
    part_index INT64,
    part_attributes STRING,
    storage_mode STRING
  >> OPTIONS(description="对于多模态事件，包含内容部分的列表。"),
  attributes JSON OPTIONS(description="任意键值对。"),
  latency_ms JSON OPTIONS(description="延迟测量值。"),
  status STRING OPTIONS(description="事件的结果。"),
  error_message STRING OPTIONS(description="详细的错误消息。"),
  is_truncated BOOLEAN OPTIONS(description="指示内容是否被截断的标志。")
)
PARTITION BY DATE(timestamp)
CLUSTER BY event_type, agent, user_id;

SELECT
  timestamp,
  event_type,
  span_id,
  parent_span_id,
  -- 根据事件类型提取摘要或特定内容
  COALESCE(
    JSON_VALUE(content, '$.messages[0].content'),
    JSON_VALUE(content, '$.summary'),
    JSON_VALUE(content)
  ) AS summary,
  JSON_VALUE(latency_ms, '$.total_ms') AS duration_ms
FROM
  `your-gcp-project-id.adk_agent_logs.agent_events_v2`
WHERE
    -- 替换为您的日志中的特定 trace_id
  trace_id = '019bb986-a0db-7da1-802d-2725795ab340'
ORDER BY
  timestamp ASC;

SELECT
  JSON_VALUE(attributes, '$.model') AS model,
  COUNT(*) AS total_calls,
  AVG(CAST(JSON_VALUE(latency_ms, '$.total_ms') AS FLOAT64)) AS avg_latency_ms,
  SUM(CAST(JSON_VALUE(attributes, '$.usage.total_tokens') AS INT64)) AS total_tokens
FROM
  `your-gcp-project-id.adk_agent_logs.agent_events_v2`
WHERE
  event_type = 'LLM_RESPONSE'
GROUP BY
  1;

SELECT
  logs.session_id,
  -- 获取图像的签名 URL（可选，用于查看）
  STRING(OBJ.GET_ACCESS_URL(parts.object_ref, "r").access_urls.read_url) as signed_url,
  -- 使用远程模型（例如 gemini-2.5-flash）分析图像
  AI.GENERATE(
    ('简要描述此图像。是什么公司标志？', parts.object_ref)
  ) AS generated_result
FROM
  `your-gcp-project-id.adk_agent_logs.agent_events_v2` logs,
  UNNEST(logs.content_parts) AS parts
WHERE
  parts.mime_type LIKE 'image/%'
ORDER BY logs.timestamp DESC
LIMIT 1;

SELECT
  span_id,
  parent_span_id,
  event_type,
  timestamp,
  -- 从 latency_ms 中提取已完成操作的持续时间
  CAST(JSON_VALUE(latency_ms, '$.total_ms') AS INT64) as duration_ms,
  -- 识别特定的工具或操作
  COALESCE(
    JSON_VALUE(content, '$.tool'),
    'LLM_CALL'
  ) as operation
FROM `your-gcp-project-id.adk_agent_logs.agent_events_v2`
WHERE trace_id = 'your-trace-id'
  AND event_type IN ('LLM_RESPONSE', 'TOOL_COMPLETED')
ORDER BY timestamp ASC;

SELECT
  timestamp,
  event_type,
  part.mime_type,
  part.storage_mode,
  part.object_ref.uri AS gcs_uri,
  -- 生成签名 URL 以直接读取内容（需要 connection_id 配置）
  STRING(OBJ.GET_ACCESS_URL(part.object_ref, 'r').access_urls.read_url) AS signed_url
FROM `your-gcp-project-id.adk_agent_logs.agent_events_v2`,
UNNEST(content_parts) AS part
WHERE part.storage_mode = 'GCS_REFERENCE'
ORDER BY timestamp DESC
LIMIT 10;

-- 1. 会话化对话历史记录（创建视图）
-- 将所有事件合并为每个会话一行，并带有格式化的历史记录。
CREATE OR REPLACE VIEW `your-project.your-dataset.agent_sessions` AS
SELECT
  session_id,
  user_id,
  MIN(timestamp) AS session_start,
  MAX(timestamp) AS session_end,
  ARRAY_AGG(
    STRUCT(timestamp, event_type, TO_JSON_STRING(content) as content, error_message)
    ORDER BY timestamp ASC
  ) AS events,
  STRING_AGG(
      CASE
          WHEN event_type = 'USER_MESSAGE_RECEIVED' THEN CONCAT('用户：', JSON_VALUE(content, '$.input'))
          WHEN event_type = 'LLM_RESPONSE' THEN CONCAT('智能体：', JSON_VALUE(content, '$.text'))
          WHEN event_type = 'TOOL_STARTING' THEN CONCAT('系统：正在调用 ', JSON_VALUE(content, '$.tool_name'))
          WHEN event_type = 'TOOL_COMPLETED' THEN CONCAT('系统：来自 ', JSON_VALUE(content, '$.tool_name'), ' 的结果')
          WHEN event_type = 'TOOL_ERROR' THEN CONCAT('系统：', JSON_VALUE(content, '$.tool_name'), ' 中的错误')
          ELSE NULL
      END,
      '\n' ORDER BY timestamp ASC
  ) AS full_conversation
FROM
  `your-project.your-dataset.agent_events_v2`
GROUP BY
  session_id, user_id;

-- 2. 工具使用情况分析
-- 提取工具名称并统计执行状态
SELECT
  JSON_VALUE(content, '$.tool_name') AS tool_name,
  event_type,
  COUNT(*) as count
FROM `your-project.your-dataset.agent_events_v2`
WHERE event_type IN ('TOOL_STARTING', 'TOOL_COMPLETED', 'TOOL_ERROR')
GROUP BY 1, 2
ORDER BY tool_name, event_type;

-- 3. 细粒度成本和令牌估算
-- 根据内容字符长度估算令牌（大约 4 个字符/令牌）
SELECT
  session_id,
  COUNT(*) as interaction_count,
  SUM(LENGTH(TO_JSON_STRING(content))) / 4 AS estimated_tokens,
  -- 示例成本：每 1k 令牌 $0.0001
  ROUND((SUM(LENGTH(TO_JSON_STRING(content))) / 4) / 1000 * 0.0001, 6) AS estimated_cost_usd
FROM `your-project.your-dataset.agent_events_v2`
GROUP BY session_id
ORDER BY estimated_cost_usd DESC
LIMIT 5;

-- 4. AI 驱动的根本原因分析（需要 BigQuery ML）
-- 使用 Gemini 分析失败的会话
SELECT
  session_id,
  AI.GENERATE(
    ('分析此对话并解释失败的根本原因。日志：', full_conversation),
    connection_id => 'your-project.us.bqml_connection',
    endpoint => 'gemini-2.5-flash'
  ).result AS root_cause_explanation
FROM `your-project.your-dataset.agent_sessions`
WHERE error_message IS NOT NULL
LIMIT 5;

https://lookerstudio.google.com/reporting/create?c.reportId=f1c5b513-3095-44f8-90a2-54953d41b125&ds.ds3.connector=bigQuery&ds.ds3.type=TABLE&ds.ds3.projectId=<your-project-id>&ds.ds3.datasetId=<your-dataset-id>&ds.ds3.tableId=<your-table-id>

from langchain.agents import create_agent
from langchain.messages import HumanMessage
from langchain_google_community.callbacks.bigquery_callback import (
    BigQueryCallbackHandler,
    BigQueryLoggerConfig,
)

# 使用图名称初始化处理器
handler = BigQueryCallbackHandler(
    project_id="your-project-id",
    dataset_id="agent_analytics",
    table_id="agent_events_v2",
    graph_name="my_agent",
)

# 创建您的智能体
agent = create_agent(llm, tools)

# 使用图上下文管理器以获取正确的 GRAPH_START/GRAPH_END 事件
run_metadata = {
    "session_id": "session-123",
    "user_id": "user-456",
    "agent": "my_agent",
}

with handler.graph_context("my_agent", metadata=run_metadata):
    result = agent.invoke(
        {"messages": [HumanMessage(content="东京的天气怎么样？")]},
        config={
            "callbacks": [handler],
            "metadata": run_metadata,
        },
    )

-- 按事件类型查询延迟
SELECT
    event_type,
    agent,
    COUNT(*) as count,
    ROUND(AVG(CAST(JSON_EXTRACT_SCALAR(latency_ms, '$.total_ms') AS FLOAT64)), 2) as avg_latency_ms,
    ROUND(APPROX_QUANTILES(CAST(JSON_EXTRACT_SCALAR(latency_ms, '$.total_ms') AS FLOAT64), 100)[OFFSET(95)], 2) as p95_latency_ms
FROM `your-project.your-dataset.agent_events_v2`
WHERE DATE(timestamp) = CURRENT_DATE()
  AND event_type IN ('LLM_RESPONSE', 'TOOL_COMPLETED', 'GRAPH_END')
GROUP BY event_type, agent
ORDER BY avg_latency_ms DESC;

from langchain_google_community.callbacks.bigquery_callback import (
    BigQueryCallbackHandler,
    BigQueryLoggerConfig,
)

# 生产配置：仅记录重要事件
config = BigQueryLoggerConfig(
    event_allowlist=[
        "LLM_RESPONSE",
        "LLM_ERROR",
        "TOOL_COMPLETED",
        "TOOL_ERROR",
        "GRAPH_END",
        "GRAPH_ERROR",
    ],
)

handler = BigQueryCallbackHandler(
    project_id="your-project-id",
    dataset_id="agent_analytics",
    config=config,
)

# 排除链事件但记录其他所有事件
config = BigQueryLoggerConfig(
    event_denylist=["CHAIN_START", "CHAIN_END"],
)

# 运行仪表板
cd libs/community/examples/bigquery_callback/webapp
pip install -r requirements.txt
uvicorn main:app --port 8001
# 打开 http://localhost:8001