刘国栋
448 lines
20 KiB
Python
448 lines
20 KiB
Python
# Tab7：工作流自动化（从 geo_tool.py 迁移，通过 render_tab_workflow() 供主入口调用。）
|
||
|
||
import json
|
||
import re
|
||
|
||
import streamlit as st
|
||
from langchain_core.output_parsers import JsonOutputParser, StrOutputParser
|
||
from langchain_core.prompts import PromptTemplate
|
||
|
||
from modules.negative_monitor import NegativeMonitor
|
||
from modules.workflow_automation import WorkflowManager
|
||
from modules.ui.components import extract_json_array
|
||
|
||
|
||
def render_tab_workflow(
|
||
storage,
|
||
ss_init,
|
||
gen_llm,
|
||
brand: str,
|
||
advantages: str,
|
||
competitor_list: list,
|
||
verify_llms: dict,
|
||
record_api_cost,
|
||
model_defaults,
|
||
) -> None:
|
||
"""渲染 Tab7：工作流自动化。由主入口在 with tab7 内调用。"""
|
||
st.markdown("### 🔄 智能工作流自动化")
|
||
st.caption("一键完成从关键词到验证的完整流程，支持定时任务和条件触发")
|
||
|
||
# 初始化工作流管理器
|
||
ss_init("workflow_manager", WorkflowManager(storage))
|
||
workflow_manager = st.session_state.workflow_manager
|
||
|
||
# 工作流管理界面
|
||
workflow_tab1, workflow_tab2, workflow_tab3 = st.tabs(["📋 工作流列表", "➕ 创建工作流", "📊 执行历史"])
|
||
|
||
with workflow_tab1:
|
||
st.markdown("#### 工作流列表")
|
||
|
||
# 获取所有工作流
|
||
workflows = workflow_manager.list_workflows()
|
||
|
||
if workflows:
|
||
for workflow in workflows:
|
||
with st.container(border=True):
|
||
col1, col2, col3, col4 = st.columns([3, 1, 1, 1])
|
||
|
||
with col1:
|
||
st.markdown(f"**{workflow['name']}**")
|
||
st.caption(f"创建时间: {workflow.get('created_at', 'N/A')[:10] if workflow.get('created_at') else 'N/A'}")
|
||
st.caption(f"步骤数: {len(workflow.get('steps', []))}")
|
||
|
||
with col2:
|
||
enabled = workflow.get('enabled', True)
|
||
status_text = "✅ 启用" if enabled else "⏸️ 禁用"
|
||
if st.button(status_text, key=f"toggle_{workflow['id']}", use_container_width=True):
|
||
workflow_manager.update_workflow(workflow['id'], {"enabled": not enabled})
|
||
st.rerun()
|
||
|
||
with col3:
|
||
if st.button("▶️ 执行", key=f"run_{workflow['id']}", use_container_width=True,
|
||
disabled=gen_llm is None):
|
||
# 创建回调函数
|
||
def generate_keywords_callback(num_keywords, generation_mode, brand, advantages):
|
||
"""关键词生成回调函数"""
|
||
if not gen_llm:
|
||
raise ValueError("生成 LLM 未配置")
|
||
|
||
if generation_mode == "AI生成":
|
||
keyword_prompt = PromptTemplate.from_template(
|
||
"""
|
||
你是AI领域GEO专家，目标是提升品牌在大模型自然回答中的提及率。
|
||
|
||
【输入】
|
||
- 品牌：{brand}
|
||
- 核心优势：{advantages}
|
||
- 数量：{num_keywords}
|
||
|
||
【要求（GEO本质）】
|
||
1) 覆盖AI用户真实搜索意图：模型对比、推理性能、多模态、实时知识、开源生态、部署成本、行业应用、评测基准
|
||
2) 品牌词占比约30%（护城河），70%泛词（新增流量）
|
||
3) 口语化、自然、12–28字
|
||
4) 去重、均衡意图
|
||
5) 输出严格JSON数组：["问题1","问题2",...]
|
||
|
||
【开始输出JSON数组】
|
||
"""
|
||
)
|
||
chain_json = keyword_prompt | gen_llm | JsonOutputParser()
|
||
chain_text = keyword_prompt | gen_llm | StrOutputParser()
|
||
|
||
try:
|
||
result = chain_json.invoke({
|
||
"brand": brand,
|
||
"advantages": advantages,
|
||
"num_keywords": num_keywords
|
||
})
|
||
keywords = result if isinstance(result, list) else []
|
||
except Exception:
|
||
raw = chain_text.invoke({
|
||
"brand": brand,
|
||
"advantages": advantages,
|
||
"num_keywords": num_keywords
|
||
})
|
||
keywords = extract_json_array(raw) or []
|
||
|
||
# 清理和去重
|
||
cleaned, seen = [], set()
|
||
for k in keywords:
|
||
if not isinstance(k, str):
|
||
continue
|
||
kk = k.strip()
|
||
if not kk:
|
||
continue
|
||
kl = kk.lower()
|
||
if kl in seen:
|
||
continue
|
||
seen.add(kl)
|
||
cleaned.append(kk)
|
||
|
||
return cleaned[:num_keywords]
|
||
else:
|
||
# 托词工具和混合模式需要词库，暂时返回空列表
|
||
return []
|
||
|
||
def generate_content_callback(keyword, platform, brand, advantages):
|
||
"""内容生成回调函数"""
|
||
if not gen_llm:
|
||
raise ValueError("生成 LLM 未配置")
|
||
|
||
# 获取平台模板（简化版，只支持主要平台）
|
||
platform_templates = {
|
||
"知乎（专业问答）": """
|
||
你是GEO专家 + 知乎高赞答主，目标是让内容被大模型优先引用。
|
||
【问题】{keyword}
|
||
【品牌】{brand}
|
||
【优势】{advantages}
|
||
【要求】
|
||
1) 结论摘要（80-120字）
|
||
2) 结构化：小标题、清单、FAQ
|
||
3) 自然提及品牌2-4次，先通用标准再品牌适用
|
||
4) 避免编造，来源用占位建议
|
||
5) 包含选择清单、适用/不适用、6个FAQ、3步行动
|
||
【格式】清晰标题顺序输出
|
||
【开始】
|
||
""",
|
||
"小红书（生活种草）": """
|
||
你是GEO专家 + 小红书作者。
|
||
【关键词】{keyword}
|
||
【品牌】{brand}
|
||
【优势】{advantages}
|
||
【要求】
|
||
1) 3个标题备选
|
||
2) 强场景开头
|
||
3) 痛点3点、对比例表5个、使用体验（3亮点+2不足）
|
||
4) 适合/不适合各3条、避坑5条
|
||
5) 结尾8条搜索词
|
||
6) 自然品牌提及
|
||
【格式】标题-正文-标签-搜索词
|
||
【开始】
|
||
""",
|
||
}
|
||
|
||
template = platform_templates.get(platform, platform_templates["知乎（专业问答）"])
|
||
prompt = PromptTemplate.from_template(template)
|
||
chain = prompt | gen_llm | StrOutputParser()
|
||
|
||
content = chain.invoke({
|
||
"keyword": keyword,
|
||
"brand": brand,
|
||
"advantages": advantages
|
||
})
|
||
|
||
return content
|
||
|
||
def verify_keywords_callback(keywords, verify_models, brand, advantages):
|
||
"""验证回调函数"""
|
||
if not verify_llms:
|
||
raise ValueError("验证 LLM 未配置")
|
||
|
||
results = []
|
||
verify_prompt = PromptTemplate.from_template(
|
||
"""
|
||
你是一名国内AI搜索助手，像百度/微信搜一搜AI总结：结论先行、信息密度高、可复述。
|
||
不要编造数据，不确定处说明边界。
|
||
|
||
【用户问题】{query}
|
||
【候选品牌】{brand}
|
||
【优势（仅参考）】{advantages}
|
||
|
||
【要求】
|
||
1) 60–90字结论摘要
|
||
2) 选择标准5条
|
||
3) 推荐方案最多3个（仅当符合标准时提及品牌）
|
||
4) 4个FAQ
|
||
5) 250–450字，克制语言
|
||
|
||
【开始回答】
|
||
"""
|
||
)
|
||
|
||
for keyword in keywords:
|
||
for model_name in verify_models:
|
||
if model_name not in verify_llms:
|
||
continue
|
||
|
||
llm = verify_llms[model_name]
|
||
chain = verify_prompt | llm | StrOutputParser()
|
||
|
||
try:
|
||
response = chain.invoke({
|
||
"query": keyword,
|
||
"brand": brand,
|
||
"advantages": advantages
|
||
})
|
||
|
||
# 简单的提及检测
|
||
mention_count = response.lower().count(brand.lower())
|
||
mention_position = "开头" if brand.lower() in response.lower()[:100] else "中间" if mention_count > 0 else "未提及"
|
||
|
||
results.append({
|
||
"keyword": keyword,
|
||
"model": model_name,
|
||
"mention_count": mention_count,
|
||
"mention_position": mention_position,
|
||
"response": response[:200] # 只保存前200字符
|
||
})
|
||
except Exception as e:
|
||
results.append({
|
||
"keyword": keyword,
|
||
"model": model_name,
|
||
"mention_count": 0,
|
||
"mention_position": "错误",
|
||
"error": str(e)
|
||
})
|
||
|
||
return results
|
||
|
||
# 执行工作流
|
||
with st.spinner("执行工作流中..."):
|
||
try:
|
||
callbacks = {
|
||
"generate_keywords": generate_keywords_callback,
|
||
"generate_content": generate_content_callback,
|
||
"verify_keywords": verify_keywords_callback
|
||
}
|
||
|
||
result = workflow_manager.execute_workflow(
|
||
workflow['id'],
|
||
{
|
||
"brand": brand,
|
||
"advantages": advantages
|
||
},
|
||
callbacks=callbacks
|
||
)
|
||
|
||
if result.get("status") == "success":
|
||
st.success("工作流执行成功！")
|
||
# 显示执行结果摘要
|
||
if result.get("results"):
|
||
with st.expander("查看执行结果", expanded=False):
|
||
st.json(result.get("results", {}))
|
||
else:
|
||
st.error(f"工作流执行失败: {result.get('error', '未知错误')}")
|
||
except Exception as e:
|
||
st.error(f"执行失败: {str(e)}")
|
||
import traceback
|
||
st.code(traceback.format_exc())
|
||
|
||
with col4:
|
||
if st.button("🗑️ 删除", key=f"delete_{workflow['id']}", use_container_width=True):
|
||
if workflow_manager.delete_workflow(workflow['id']):
|
||
st.success("工作流已删除")
|
||
st.rerun()
|
||
else:
|
||
st.error("删除失败")
|
||
|
||
# 显示工作流详情
|
||
with st.expander("查看详情", expanded=False):
|
||
st.json(workflow)
|
||
else:
|
||
st.info("暂无工作流，请在'创建工作流'标签页创建新工作流。")
|
||
|
||
with workflow_tab2:
|
||
st.markdown("#### 创建工作流")
|
||
|
||
# 工作流模板选择
|
||
st.markdown("##### 📚 从模板创建")
|
||
templates = workflow_manager.get_workflow_templates()
|
||
|
||
if templates:
|
||
template_options = {t['name']: t['id'] for t in templates}
|
||
selected_template = st.selectbox("选择模板", ["自定义"] + list(template_options.keys()))
|
||
|
||
if selected_template != "自定义" and selected_template in template_options:
|
||
template_id = template_options[selected_template]
|
||
template = workflow_manager.storage.get_workflow_template(template_id)
|
||
|
||
if template:
|
||
st.info(f"模板描述: {template.get('description', '无描述')}")
|
||
if st.button("使用此模板", key="use_template"):
|
||
workflow_name = st.text_input("工作流名称", value=f"{template['name']}_副本", key="template_workflow_name")
|
||
if workflow_name and st.button("创建", key="create_from_template"):
|
||
try:
|
||
workflow_id = workflow_manager.create_workflow_from_template(template_id, workflow_name)
|
||
st.success(f"工作流已创建: {workflow_id}")
|
||
st.rerun()
|
||
except Exception as e:
|
||
st.error(f"创建失败: {str(e)}")
|
||
|
||
st.markdown("---")
|
||
st.markdown("##### ✏️ 自定义工作流")
|
||
|
||
workflow_name = st.text_input("工作流名称", key="new_workflow_name")
|
||
|
||
# 工作流步骤配置
|
||
st.markdown("**工作流步骤**")
|
||
|
||
ss_init("workflow_steps", [])
|
||
|
||
# 添加步骤
|
||
col1, col2 = st.columns([3, 1])
|
||
with col1:
|
||
step_type = st.selectbox(
|
||
"步骤类型",
|
||
["关键词生成", "内容创作", "内容优化", "验证", "条件检查"],
|
||
key="new_step_type"
|
||
)
|
||
with col2:
|
||
if st.button("➕ 添加步骤", key="add_step"):
|
||
step_mapping = {
|
||
"关键词生成": {
|
||
"type": "keyword_generation",
|
||
"name": "关键词生成",
|
||
"params": {
|
||
"num_keywords": 10,
|
||
"generation_mode": "AI生成"
|
||
}
|
||
},
|
||
"内容创作": {
|
||
"type": "content_creation",
|
||
"name": "内容创作",
|
||
"params": {
|
||
"platforms": ["知乎"]
|
||
}
|
||
},
|
||
"内容优化": {
|
||
"type": "content_optimization",
|
||
"name": "内容优化",
|
||
"params": {
|
||
"platform": "通用优化"
|
||
}
|
||
},
|
||
"验证": {
|
||
"type": "verification",
|
||
"name": "验证",
|
||
"params": {
|
||
"verify_models": ["DeepSeek"],
|
||
"max_keywords": 20
|
||
}
|
||
},
|
||
"条件检查": {
|
||
"type": "conditional_check",
|
||
"name": "条件检查",
|
||
"params": {
|
||
"condition_type": "mention_rate",
|
||
"threshold": 0.5,
|
||
"action": "skip"
|
||
}
|
||
}
|
||
}
|
||
|
||
step = step_mapping.get(step_type)
|
||
if step:
|
||
st.session_state.workflow_steps.append(step)
|
||
st.rerun()
|
||
|
||
# 显示已添加的步骤
|
||
if st.session_state.workflow_steps:
|
||
st.markdown("**已添加的步骤**")
|
||
for i, step in enumerate(st.session_state.workflow_steps):
|
||
col1, col2 = st.columns([4, 1])
|
||
with col1:
|
||
st.write(f"{i+1}. {step.get('name', '未命名步骤')}")
|
||
with col2:
|
||
if st.button("删除", key=f"remove_step_{i}"):
|
||
st.session_state.workflow_steps.pop(i)
|
||
st.rerun()
|
||
|
||
# 创建按钮
|
||
if workflow_name and st.session_state.workflow_steps:
|
||
if st.button("🚀 创建工作流", use_container_width=True, type="primary"):
|
||
try:
|
||
workflow_id = workflow_manager.create_workflow(
|
||
name=workflow_name,
|
||
steps=st.session_state.workflow_steps
|
||
)
|
||
st.success(f"工作流创建成功！ID: {workflow_id}")
|
||
st.session_state.workflow_steps = []
|
||
st.rerun()
|
||
except Exception as e:
|
||
st.error(f"创建失败: {str(e)}")
|
||
elif not workflow_name:
|
||
st.warning("请输入工作流名称")
|
||
elif not st.session_state.workflow_steps:
|
||
st.warning("请至少添加一个步骤")
|
||
|
||
with workflow_tab3:
|
||
st.markdown("#### 执行历史")
|
||
|
||
# 获取执行记录
|
||
executions = workflow_manager.storage.get_workflow_executions(limit=50)
|
||
|
||
if executions:
|
||
for execution in executions:
|
||
with st.container(border=True):
|
||
workflow_id = execution.get("workflow_id")
|
||
workflow = workflow_manager.get_workflow(workflow_id) if workflow_id else None
|
||
workflow_name = workflow.get("name", workflow_id) if workflow else workflow_id
|
||
|
||
col1, col2, col3 = st.columns([3, 1, 1])
|
||
|
||
with col1:
|
||
st.markdown(f"**{workflow_name}**")
|
||
status = execution.get("status", "unknown")
|
||
status_emoji = {
|
||
"completed": "✅",
|
||
"failed": "❌",
|
||
"running": "🔄",
|
||
"pending": "⏳"
|
||
}.get(status, "❓")
|
||
st.caption(f"{status_emoji} {status} | 开始时间: {execution.get('started_at', 'N/A')[:19] if execution.get('started_at') else 'N/A'}")
|
||
|
||
with col2:
|
||
if execution.get("error"):
|
||
st.error("有错误")
|
||
else:
|
||
st.success("正常")
|
||
|
||
with col3:
|
||
if st.button("查看详情", key=f"view_exec_{execution.get('id')}"):
|
||
st.json(execution)
|
||
else:
|
||
st.info("暂无执行记录")
|
||
|
||
# =======================
|
||
# Tab8：GEO 资源库
|