Project Overview
The main function is AI-powered analysis of financial news with automatic report generation.
Technical Architecture
- Core framework:
FastAPI - Programming language:
Python - Graph database:
Neo4j - Frontend:
JavaScript - AI technologies: GraphRAG, large language models
Key Features
1. News Analysis and Report Generation
The project automatically fetches financial news and runs AI analysis to generate professional analytical reports:
- Research report analysis
- Headline summaries
- Flash news processing
- Announcement interpretation
2. GraphRAG Integration
The project successfully integrated GraphRAG technology, significantly improving retrieval quality:
- Implemented graph-structuring of the data
- Delivered visual exploration of the results
3. Multi-Persona AI Analysis
The project presets a variety of AI personas and writing styles to comment on and analyze the news:
- 20 AI personas
- 4 writing styles
Project Progress
Q2 (June)
- Kicked off project development
- Applied for developer accounts (D-U-N-S number, Apple, Google)
- Designed the table structures and communication protocols
- Connected data sources such as Waton International research reports, the WB crawler, and WB news
Q3 (Jul-Aug)
- Implemented Agents that summarize all kinds of news
- Implemented AI-generated summary images based on news content
- Injected Agent-generated data into the WBroker system
- Completed testing and integration of GraphRAG
Delivery Results
- Generated 300+ news summaries and insight pieces daily, serving wealth advisors and high-net-worth clients.
- Built a multi-level approval workflow so sensitive news is reviewed and published within 15 minutes.
- The Neo4j graph powers multi-dimensional retrieval, raising investment research topic coverage from 45% to 85%.
Team and Governance
- Set up three working groups — data ingestion, algorithms, and product operations — with regular quality reviews.
- Introduced KRIs (key risk indicators) and content compliance scores as the basis for iteration priorities.
- Established cross-department weekly meetings and monthly retrospectives, clarifying plans for the next phase of business growth and technical evolution.
Usage Instructions
Data Source Configuration
# 数据源配置示例
data_sources:
- name: "华通国际研报"
type: "research_report"
- name: "WB爬虫"
type: "crawler"
- name: "WB新闻"
type: "news"
- name: "数字货币新闻"
type: "crypto_news"
Core Modules
- ai_handlers: core processing module
- broker_news: news processing module
- broker_crawler: crawler module
- graph_rag_service: GraphRAG service module