Menu
基于 LLM 的深度迭代搜索与推理服务。擅长处理复杂问题，通过自动分解查询、多轮迭代检索、信息评估与验证，最终生成全面且结构化的深度分析报告。
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Writes, edits, and creates dbt models following best practices. Use when user needs to create new dbt SQL models, update existing models, or convert raw SQL to dbt format. Handles staging, intermediate, and mart models with proper config blocks, CTEs, and documentation.
Apply Domain-Driven Design, Clean Architecture, CQRS, and command/query patterns to code reviews and feature design. Use when analyzing or designing code in Application, Service, Infrastructure, DataAccess, Validation, Domain, or Functions projects, or when addressing architectural concerns, layering, mapping, entities, value objects, repositories, or validators in the Rome Repair Order Service.
Analyzes and refactors code using Domain-Driven Design principles. Use when refactoring domain models, identifying DDD anti-patterns, improving domain clarity, or applying tactical/strategic DDD patterns.
Guide for DDD strategic design - analyzing domains through structured questioning, conducting stakeholder interviews (PM/domain experts/users), and producing Bounded Context analysis, Context Maps, and Ubiquitous Language. Use when user needs help understanding domain boundaries, planning domain interviews, or structuring DDD strategic artifacts.
Win competitive rounds: run a clean process, deliver value previews before asking, coordinate partners, and manage timelines. Use when you're trying to close a 'must win' deal against other funds.
End-to-end associate workflow with time-boxed gates: thesis -> sourcing -> meetings -> diligence -> memo, ending with either IC-ready memo or explicit kill decision. Use when you need to run the full pipeline for a sector or a specific deal.
| name | deepsearch-service |
| description | 基于 LLM 的深度迭代搜索与推理服务。擅长处理复杂问题，通过自动分解查询、多轮迭代检索、信息评估与验证，最终生成全面且结构化的深度分析报告。 |
基于 LLM 的迭代式深度搜索服务，能够：
from services.deepsearch_service.client import DeepSearchClient
client = DeepSearchClient()
# 健康检查
status = client.health_check()
# 深度搜索（默认参数）
result = client.search("Python异步编程的最佳实践有哪些？")
# 自定义参数
result = client.search(
query="如何设计一个高可用的微服务架构？",
max_iterations=3, # 最大迭代次数 (1-5)
queries_per_iteration=3, # 每轮查询数 (1-5)
depth_level="deep" # 搜索深度: quick/normal/deep
)
# 获取报告和来源
print(result["report"])
for source in result["sources"]:
print(f"- {source['title']}: {source['url']}")
{
"query": "Python异步编程的最佳实践有哪些？",
"report": "# Python异步编程最佳实践\n\n## 1. 核心概念...",
"sources": [
{
"title": "Python官方asyncio文档",
"url": "https://docs.python.org/...",
"relevance": 0.95,
"snippet": "asyncio是Python标准库中的异步I/O框架..."
}
],
"iterations": [
{
"iteration": 1,
"queries": ["Python asyncio 教程", "async await 用法"],
"results_count": 6,
"key_findings": ["asyncio是标准库组件"]
}
],
"total_iterations": 2,
"total_sources": 8,
"elapsed_seconds": 45.32,
"search_timestamp": "2025-12-28T18:30:00"
}
| 特性 | WebSearch | DeepSearch |
|---|---|---|
| 搜索模式 | 单轮搜索 | 迭代多轮搜索 |
| 查询策略 | 直接使用输入 | LLM 分解/优化 |
| 输出格式 | 结构化列表 | 综合报告 + 来源 |
| 响应时间 | 10-30s | 30-120s |