https://github.com/albertllousas/virtual-customer-agent-with-ai
https://github.com/albertllousas/virtual-customer-agent-with-ai
Last synced: about 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/albertllousas/virtual-customer-agent-with-ai
- Owner: albertllousas
- Created: 2024-02-09T13:33:20.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-04-20T05:56:26.000Z (about 1 year ago)
- Last Synced: 2025-01-30T22:48:25.788Z (3 months ago)
- Language: Kotlin
- Size: 222 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: readme.md
Awesome Lists containing this project
README
# Next Gen Pets - Customer Service Chatbot - AI & RAG
Keywords: `kotlin`, `spring-boot`, `spring-AI`, `Open-AI`, `Vector Store`, `RAG`, `Redis`
Codebase for [this](https://medium.com/@albert.llousas/practical-ai-for-everyday-software-engineers-f2ac1fcf0063) medium article
## Description
This is a simple customer service chatbot for a company called **Next Gen Pets** that uses the OpenAI API to generate responses to customer queries.
All the FAQs and specific information about the company is loaded from external files and used to provide domain expertise in the answers.
Service will be able to:
- Load all specific FAQs from our imaginary company from files in [docs folder](/src/main/resources/docs), generate vector embeddings and store them in our InMemory Vector Store
- For each customer query:
- Retrieve the necessary specific Next Gen Pet related FAQs from the InMemory Vector Store
- Retrieve previous chats
- Call OpenAI API to generate a response with customer query, but also with the retrieved related DOCS and prevvious chats
## What is RAG (Retrieval Augmented Generation)?
RAG, is a technique for giving a pretrained generic AI model like ChatGPT additional context in order to get a more specific answer.
RAG involves supplying a model with data it hasn't been trained on and requesting it to incorporate this information into its response.
With RAG, there's no modification to the model itself; instead, asking the model to consider data provided along the question to fulfill the request.
## How to run it
```bash
docker-compose -f docker-compose-redis-only.yml up
export OPENAI_API_KEY=****YOUR_API_KEY****
./gradlew run
```
```bash
curl -X POST 'http://localhost:8080/customer-service/chats' \
-H "Content-Type: application/json" \
--data-raw '{
"message": "Hi, my name is Albert"
}'
```
```json
{"chatId":"{CHAT_ID}","message":"Hello Albert! How can I assist you today?"}
```
```bash
curl -X PUT 'http://localhost:8080/customer-service/chats/{CHAT_ID}' \
-H "Content-Type: application/json" \
--data-raw '{
"message": "Could you list all the aquatic series models that you have available?"
}'
```
```json
{
"chatId":"{CHAT_ID}",
"message":"Certainly! Here is a list of the aquatic series models available from NextGen Pets:\n\n1. EcoDolphin\n2. EcoShark\n3. EcoJelly\n\nThese models in the Aquatic Series are designed to provide an interactive and immersive aquatic experience. Each model comes with unique specifications and functionalities to enhance your aquatic companionship."
}
```
## Missing
- A proper **"Intent" mechanism** to automate actions, metrics ...
- A proper Vector Store
## Resources
- https://www.linkedin.com/pulse/rag-chatgpt-richard-reyes-mba-7heff/
- https://github.com/spring-tips/llm-rag-with-spring-ai/