# 與資料對話：深入探索 WrenAI 的開源與架構 - 祥溥（Andy）

{%hackmd @JSDC2024/INuUpOqWS42gD1_pz5S4Bg %}

Slido：
https://app.sli.do/event/iJsYyZ4rdo77nm1KZ2JRNd

> 開始做筆記

## About Me

## Agenda

## Current Data Challenges
- Data Access
- Data dirty
- Dynamic Demands

## Wren AI
- https://getwren.ai/oss
- https://github.com/Canner/WrenAI

### Architecture
- Sementic Layer 語意層
- 土豆 = potato vs peanut
- 自行車 / 腳踏車
- "今年" - 財務 或 其他業務
- AI - 理解商業邏輯
- "Wren Engine"
- Flow
- Onboard
1. Connect Data Source
2. Modeling
3. Add sementic indexing
- Ask
1. Ask a question
2. Generate NC SQL
- Security
- Only schema and sementic layer

### Model
- A RAG pipeline
- General Model don't know your business model
- Pretrained model

### Wren Engine

Dialect SQL Transpiler Abis

UI - Next.js

> 聊天區
>