192 lines
4.0 KiB
Markdown
192 lines
4.0 KiB
Markdown
## Getting Started
|
||
|
||
[READ THIS IN ENGLISH](README_en.md)
|
||
|
||
**LangChain4j-AIDeepin**
|
||
基于 ChatGPT 等大语言模型与 Langchain4j 等应用框架实现，开源、可离线部署的检索增强生成(RAG)项目。
|
||
|
||
> **该项目如对您有帮助，欢迎点赞**
|
||
|
||
代码仓库地址：[github](https://github.com/moyangzhan/langchain4j-aideepin) [gitee](https://gitee.com/moyangzhan/langchain4j-aideepin)
|
||
|
||
## 体验网址
|
||
|
||
[http://www.aideepin.com](http://www.aideepin.com/)
|
||
|
||
## 功能点
|
||
|
||
* 注册&登录
|
||
* 多会话（多角色）
|
||
* 图片生成（文生图、修图、图生图）
|
||
* 提示词
|
||
* 额度控制
|
||
* 基于大模型的知识库（RAG）
|
||
* 基于大模型的搜索（RAG）
|
||
* 多模型随意切换
|
||
* 多搜索引擎随意切换
|
||
|
||
## 接入的模型：
|
||
|
||
* ChatGPT 3.5
|
||
* 通义千问
|
||
* 文心一言
|
||
* ollama
|
||
* DALL-E 2
|
||
|
||
## 接入的搜索引擎
|
||
|
||
Google
|
||
|
||
Bing (TODO)
|
||
|
||
百度 (TODO)
|
||
|
||
## 技术栈
|
||
|
||
该仓库为后端服务，前端项目见[langchain4j-aideepin-web](https://github.com/moyangzhan/langchain4j-aideepin-web)
|
||
|
||
后端：
|
||
|
||
jdk17
|
||
|
||
springboot3.0.5
|
||
|
||
[langchain4j(Java version of LangChain)](https://github.com/langchain4j/langchain4j)
|
||
|
||
**Postgresql(需要安装[pgvector](https://github.com/pgvector/pgvector)扩展)**
|
||
|
||
前端：
|
||
|
||
vue3+typescript+pnpm
|
||
|
||
## 如何部署
|
||
|
||
### 初始化
|
||
|
||
**a. 初始化数据库**
|
||
|
||
* 创建数据库aideepin
|
||
* 执行docs/create.sql
|
||
* 填充各AI平台的配置(至少设置一个)
|
||
|
||
openai的secretKey
|
||
|
||
```plaintext
|
||
update adi_sys_config set value = '{"secret_key":"my_openai_secret_key"}' where name = 'openai_setting';
|
||
```
|
||
|
||
灵积大模型平台的apiKey
|
||
|
||
```plaintext
|
||
update adi_sys_config set value = '{"api_key":"my_dashcope_api_key"}' where name = 'dashscope_setting';
|
||
```
|
||
|
||
千帆大模型平台的配置
|
||
|
||
```plaintext
|
||
update adi_sys_config set value = '{"api_key":"my_qianfan_api_key","secret_key":"my_qianfan_secret_key"}' where name = 'qianfan_setting';
|
||
```
|
||
|
||
ollama的配置
|
||
|
||
```
|
||
update adi_sys_config set value = '{"base_url":"my_ollama_base_url"}' where name = 'ollama_setting';
|
||
```
|
||
|
||
* 启用模型或新增模型
|
||
|
||
```
|
||
-- Enable model
|
||
update adi_ai_model set is_enable = true where name = 'gpt-3.5-turbo';
|
||
update adi_ai_model set is_enable = true where name = 'dall-e-2';
|
||
update adi_ai_model set is_enable = true where name = 'qwen-turbo';
|
||
update adi_ai_model set is_enable = true where name = 'ernie-3.5-8k-0205';
|
||
update adi_ai_model set is_enable = true where name = 'tinydolphin';
|
||
|
||
-- Add new model
|
||
INSERT INTO adi_ai_model (name, type, platform, is_enable) VALUES ('vicuna', 'text', 'ollama', true);
|
||
```
|
||
|
||
* 填充搜索引擎的配置
|
||
|
||
Google的配置
|
||
|
||
```
|
||
update adi_sys_config set value = '{"url":"https://www.googleapis.com/customsearch/v1","key":"my key from cloud.google.com","cx":"my cx from programmablesearchengine.google.com"}' where name = 'google_setting';
|
||
```
|
||
|
||
**b. 修改配置文件**
|
||
|
||
* postgresql: application-[dev|prod].xml中的spring.datasource
|
||
* redis: application-[dev|prod].xml中的spring.data.redis
|
||
* mail: application.xml中的spring.mail
|
||
|
||
### 编译及运行
|
||
|
||
* 进入项目
|
||
|
||
```plaintext
|
||
cd langchain4j-aideepin
|
||
```
|
||
|
||
* 打包：
|
||
|
||
```
|
||
mvn clean package -Dmaven.test.skip=true
|
||
```
|
||
|
||
* 运行
|
||
|
||
a. jar包启动：
|
||
|
||
```plaintext
|
||
cd adi-bootstrap/target
|
||
nohup java -jar -Xms768m -Xmx1024m -XX:+HeapDumpOnOutOfMemoryError adi-bootstrap-0.0.1-SNAPSHOT.jar --spring.profiles.active=[dev|prod] dev/null 2>&1 &
|
||
```
|
||
|
||
b. docker启动
|
||
|
||
```plaintext
|
||
cd adi-bootstrap
|
||
docker build . -t aideepin:0.0.1
|
||
docker run -d \
|
||
--name=aideepin \
|
||
-e APP_PROFILE=[dev|prod] \
|
||
-v="/data/aideepin/logs:/data/logs" \
|
||
aideepin:0.0.1
|
||
```
|
||
|
||
## 待办：
|
||
|
||
高级RAG
|
||
|
||
* 查询压缩 √
|
||
* 查询路由
|
||
* Re-rank
|
||
|
||
增加搜索引擎（BING、百度）
|
||
|
||
## 截图
|
||
|
||
**AI聊天：**
|
||
|
||
|
||
**AI画图：**
|
||
|
||
|
||
|
||
**知识库：**
|
||
|
||
|
||
|
||
|
||
**向量化：**
|
||
|
||
|
||
|
||
|
||
|
||
**额度统计：**
|
||
|
||
!
|