- Python 98.8%
- Jupyter Notebook 1.2%
| .streamlit | ||
| agents | ||
| data | ||
| data_processing | ||
| docs | ||
| models | ||
| notebooks | ||
| outputs | ||
| tests | ||
| .gitignore | ||
| app.py | ||
| config.py | ||
| demo.png | ||
| main.py | ||
| README.md | ||
| requirements.txt | ||
Market Entry Agent
Authors / 作者:Lanting Lu & Yixin Zhang
一个基于 AI 的多智能体系统,帮助跨境卖家评估某一产品是否适合进入目标海外市场,并自动生成中英双语商业分析报告。
An AI-powered multi-agent system that helps cross-border sellers evaluate whether a product is worth entering a target overseas market.
在线演示 / Live Demo
Live demo: https://marketpilot-ai-xmsqvrsxc6dl3ycgv9mmib.streamlit.app/
Demo Preview / 演示截图
Local Demo / 本地运行
本地运行 Streamlit:
streamlit run app.py
然后打开:
http://localhost:8501
Demo 网站支持:
- 快速演示场景
- 目标国家和行业选择
- 产品自由输入
- 平台推荐
- 中英双语市场进入报告生成
- Markdown 报告下载
Deploy as a Public Live Demo / 部署为公开演示网站
推荐使用 Streamlit Community Cloud 部署本项目。部署后,你会得到一个类似下面这样的公开链接:
https://your-app-name.streamlit.app
部署步骤:
- 将项目 push 到 GitHub。
- 打开 Streamlit Community Cloud。
- 使用 GitHub 登录。
- 点击
New app。 - 选择该项目仓库。
Main file path填写:
app.py
- 点击
Deploy。 - 复制生成的公开 URL,并替换 README 顶部的
Live demo链接。
部署前请确认仓库中包含:
app.py
requirements.txt
data/processed/market_entry_dataset.csv
data/processed/market_entry_dataset.json
data/sample/sample_market_entry_dataset.csv
demo.png
项目速览 / Project Snapshot
| 项目 | 说明 |
|---|---|
| 项目类型 | AI Agent / Market intelligence / Decision support |
| 核心问题 | 这个产品是否值得进入该海外市场? |
| 输入 | 目标国家、行业、产品、平台 |
| 输出 | 市场进入评分、需求预测、风险评估、中英双语报告 |
| 当前市场 | United States, Hong Kong, Japan, South Korea, Vietnam, Germany, Netherlands, Malaysia |
| 当前行业 | Beauty, Home & Kitchen, Consumer Electronics, Pet Supplies, Sports & Outdoors |
| 数据层 | Excel 来源数据 + benchmark estimates + processed CSV/JSON/SQLite |
| 界面 | Streamlit Web App + command-line demo |
中文版本
1. 项目介绍
Market Entry Agent 是一个面向跨境业务场景的 AI 多智能体系统,用于评估某个产品是否适合进入目标海外市场。
用户选择目标国家和行业,输入任意产品名称后,系统会自动完成:
- 输入标准化;
- 结构化市场数据检索;
- 行业基准数据匹配;
- 市场需求预测;
- 市场进入风险评估;
- Market Entry Score 市场进入评分;
- 中英双语商业报告生成。
产品字段支持自由输入。如果系统没有该具体产品的精确数据,会自动使用所选国家和行业的 benchmark 指标进行分析,并在报告中明确标注为“行业基准分析”。
2. 项目价值
海外市场进入判断通常需要综合大量分散信息,例如:
- 宏观经济条件
- 电商成熟度
- 平台覆盖情况
- 行业增长趋势
- 产品价格区间
- 评论量与评分
- 竞争强度
- 物流复杂度
- 法规风险
- 文化匹配度
本项目展示了如何用 AI Agent 工作流,将这些碎片化信息整理成一套结构化、可解释、可复现、可展示的市场进入决策流程。
3. Demo 演示方式
推荐演示故事线:
一个跨境卖家想把产品卖到海外。
他选择目标国家和行业,输入产品名称。
Agent 自动生成中英双语市场进入报告。
报告包含评分、需求预测、风险评估和进入策略建议。
示例输入:
国家:United States
行业:Beauty
产品:Mineral Sunscreen Stick
平台:Amazon
系统输出:
市场进入评分
需求等级
预计月销量
综合风险
推荐进入策略
中英双语市场进入报告
4. Agent 工作流程
flowchart LR
A[用户输入] --> B[Planner Agent]
B --> C[Retriever Agent]
C --> D[Risk Model]
C --> E[Scoring Model]
C --> F[Demand Prediction Model]
D --> G[Report Agent]
E --> G
F --> G
G --> H[中英双语 Markdown 报告]
5. 核心模块
| 模块 | 作用 |
|---|---|
| Planner Agent | 标准化国家、行业、产品和平台输入 |
| Retriever Agent | 检索精确产品数据,或匹配国家 + 行业 benchmark 数据 |
| Risk Model | 评估市场饱和、价格压力、法规、物流和文化匹配风险 |
| Scoring Model | 计算 0-100 的 Market Entry Score |
| Demand Prediction Model | 预测需求指数和预计月销量区间 |
| Report Agent | 生成完整中文报告和完整英文报告 |
| Streamlit App | 提供网站式 demo、快速场景、报告预览和下载 |
6. 数据层说明
项目支持从 Excel 导入平台和来源数据。原始 Excel 文件放在:
data/raw/
数据处理流程会读取所有工作簿和工作表,标准化字段,保留来源信息,追加 benchmark estimate 数据,并导出:
data/processed/market_entry_dataset.csv
data/processed/market_entry_dataset.json
data/processed/market_entry_dataset.sqlite
统一字段结构:
country
industry
product
platform
metric_name
metric_value
metric_unit
source_name
source_url
source_type
updated_date
notes
当前数据集中包含 8 个目标市场和 5 个行业的 benchmark 场景,适合用于 demo、初筛和方案比较。正式用于客户决策前,仍建议接入实时平台数据、搜索趋势、竞品价格、广告成本和法规认证信息。
7. 技术栈
- Python
- pandas
- openpyxl
- SQLite
- Streamlit
- pytest
- Markdown 报告生成
- 基于规则的评分与预测模型
8. 运行方式
安装依赖:
pip install -r requirements.txt
构建结构化数据集:
python data_processing/build_dataset.py
如果 Windows 上 python 命令不可用,可以使用 Anaconda Python:
D:\Anaconda\anaconda3\python.exe data_processing\build_dataset.py
运行命令行版本:
python main.py
运行 Streamlit Web 应用:
streamlit run app.py
运行测试:
pytest
9. 项目结构
market-entry-agent/
├── agents/
│ ├── planner_agent.py
│ ├── retriever_agent.py
│ └── report_agent.py
├── data/
│ ├── raw/
│ ├── processed/
│ └── sample/
├── data_processing/
│ ├── build_dataset.py
│ ├── data_cleaner.py
│ ├── data_schema.py
│ └── excel_loader.py
├── models/
│ ├── demand_prediction_model.py
│ ├── risk_model.py
│ └── scoring_model.py
├── docs/
├── outputs/
├── tests/
├── app.py
├── main.py
├── config.py
└── requirements.txt
10. 未来改进方向
- 接入 Amazon、Rakuten、Walmart、Shopee、Lazada、eBay 等平台 API
- 增加 Google Trends 和社交媒体趋势信号
- 接入 World Bank、IMF、OECD 和各国统计局数据
- 增加 PDF 导出
- 支持多个国家/行业/产品方案对比
- 增加历史报告存储
- 将部分规则模型升级为机器学习模型
- 增加 LLM 分析师层,用于更丰富的商业解读
- 部署为可公开访问的 Streamlit demo 网站
English Version
1. What It Does
Market Entry Agent turns fragmented market signals into a structured market-entry decision report.
Given a target country, industry, product, and optional platform, the system:
- normalizes the user request,
- retrieves structured market and platform context,
- estimates demand,
- evaluates business and market-entry risk,
- calculates a weighted Market Entry Score,
- generates a client-ready bilingual Markdown report.
The product field is intentionally flexible. Users can enter any product name. If exact product-level data is unavailable, the agent uses benchmark metrics from the selected country and industry, and clearly marks the report as an industry-benchmark analysis.
2. Why This Matters
International market-entry decisions usually require combining signals from many places:
- macroeconomic conditions
- ecommerce readiness
- marketplace/platform availability
- category growth
- product pricing
- review volume
- competition level
- logistics complexity
- regulatory risk
- cultural fit
This project demonstrates how an AI Agent workflow can organize these signals into a reproducible, explainable, and presentation-ready decision process.
3. Demo Flow
A typical demo story:
A seller wants to launch a product overseas.
They select a target market and industry.
They enter any product name.
The agent generates a bilingual market-entry report with score, demand, risk, and strategy.
Example:
Market: United States
Industry: Beauty
Product: Mineral Sunscreen Stick
Platform: Amazon
The app returns:
Market Entry Score
Demand Level
Estimated Monthly Sales
Overall Risk
Recommended Entry Strategy
Bilingual Market Entry Report
4. Agent Workflow
flowchart LR
A[User Input] --> B[Planner Agent]
B --> C[Retriever Agent]
C --> D[Risk Model]
C --> E[Scoring Model]
C --> F[Demand Prediction Model]
D --> G[Report Agent]
E --> G
F --> G
G --> H[Bilingual Markdown Report]
5. System Components
| Component | Responsibility |
|---|---|
| Planner Agent | Normalizes country, industry, product, and platform input |
| Retriever Agent | Retrieves exact product data or country-industry benchmark data |
| Risk Model | Evaluates saturation, pricing, regulatory, logistics, and cultural-fit risks |
| Scoring Model | Calculates a weighted 0-100 Market Entry Score |
| Demand Prediction Model | Estimates demand index and monthly sales range |
| Report Agent | Generates full Chinese and English reports |
| Streamlit App | Provides the demo website, scenario builder, report preview, and downloads |
6. Data Layer
The project supports Excel-based source ingestion. Raw Excel files are placed in:
data/raw/
The data processing pipeline reads every workbook and sheet, standardizes fields, preserves source metadata, adds benchmark estimate records, and exports:
data/processed/market_entry_dataset.csv
data/processed/market_entry_dataset.json
data/processed/market_entry_dataset.sqlite
Canonical schema:
country
industry
product
platform
metric_name
metric_value
metric_unit
source_name
source_url
source_type
updated_date
notes
The current dataset includes structured benchmark scenarios for 8 target markets and 5 industries. These benchmark records are suitable for demo screening and scenario comparison. They should be validated with live marketplace data before real client decisions.
7. Tech Stack
- Python
- pandas
- openpyxl
- SQLite
- Streamlit
- pytest
- Markdown report generation
- Rule-based scoring and forecasting models
8. How To Run
Install dependencies:
pip install -r requirements.txt
Build the processed dataset:
python data_processing/build_dataset.py
If the python command is unavailable on Windows, use the Anaconda interpreter:
D:\Anaconda\anaconda3\python.exe data_processing\build_dataset.py
Run the command-line demo:
python main.py
Run the Streamlit web app:
streamlit run app.py
Run tests:
pytest
9. Repository Structure
market-entry-agent/
├── agents/
│ ├── planner_agent.py
│ ├── retriever_agent.py
│ └── report_agent.py
├── data/
│ ├── raw/
│ ├── processed/
│ └── sample/
├── data_processing/
│ ├── build_dataset.py
│ ├── data_cleaner.py
│ ├── data_schema.py
│ └── excel_loader.py
├── models/
│ ├── demand_prediction_model.py
│ ├── risk_model.py
│ └── scoring_model.py
├── docs/
├── outputs/
├── tests/
├── app.py
├── main.py
├── config.py
└── requirements.txt
10. Future Improvements
- integrate live marketplace APIs such as Amazon, Rakuten, Walmart, Shopee, Lazada, and eBay
- add Google Trends and social media listening signals
- connect World Bank, IMF, OECD, and national statistics APIs
- add PDF export
- support multi-scenario comparison
- add historical report storage
- replace some rule-based formulas with trained ML models
- add an LLM analyst layer for richer narrative reasoning
- deploy the Streamlit app as a public demo website
数据说明 / Disclaimer
本项目是一个适合 portfolio 和 AI Agent 演示的原型系统。部分数据为 benchmark estimate,用于早期市场筛选和场景比较,不应被直接视为实时平台数据。
This project is a portfolio-ready AI Agent demo. Some records are benchmark estimates designed for early-stage screening and scenario comparison. They should not be treated as live marketplace data.
