A smart, role-based AI assistant built with Java 21, Spring Boot 3, Spring AI, and locally hosted LLM (via Ollama). It extracts user intent from chat input and routes it to appropriate backend services (e.g., weather, stock price, book DB, etc.). The front-end is built with Angular: Angular Frontend Repo: https://github.com/eminyuce/angular-chatbot
- 🧠 Intent Extraction with vector similarity and LLM JSON parsing
- 🔐 Role-based Access using Spring Security + JWT (Angular/Admin roles)
- 📡 Third-party API Integration via FeignClient
- ⚙️ Pluggable Services dynamically executed based on intent
- 🖥️ Locally Hosted LLM using Ollama for fast and secure inference
Angular Frontend
↓
REST API (Spring Boot)
↓
Intent Extraction Service (LLM + Embedding Similarity)
↓
Resolved Service Layer (Weather, Books, Stocks, etc.)
↓
Third Party APIs / Internal Logic
| Layer | Tech |
|---|---|
| Language | Java 21 |
| Framework | Spring Boot 3 |
| LLM Runtime | Ollama (e.g., LLaMA2, Mistral) |
| Security | Spring Security + JWT |
| REST API | Spring Web, Spring Actuator |
| Database | H2 (in-memory) |
| Vector Matching | Custom Embedding Service |
| External APIs | FeignClient |
| Documentation | Swagger/OpenAPI |
- Java 21
- Docker
- Ollama installed
- A running LLM model (e.g.
llama2,mistral, etc.)
ollama run llama2./mvnw spring-boot:runhttp://localhost:8080/swagger-ui/index.html
src/
├── controller/ // REST endpoints
├── service/ // Core services (intent, embedding, execution)
├── security/ // JWT & role-based access
├── llm/ // Spring AI LLM integration
├── config/ // App and Feign configuration
└── model/ // DTOs and domain models
| Intent | Example Input |
|---|---|
| Weather | “What is the weather in Istanbul?” |
| Stock Price | “What is the stock price of AMD?” |
| Book CRUD | “Add book titled ‘AI Revolution’ by John” |
| Recipe | “Give me the recipe of New York Pizza” |
| Drug Info | “Tell me about aspirin” |
| Admin Users | “Get all users of chatbot” (admin only) |
| Service | Example Input | Roles Allowed |
|---|---|---|
| Weather | "What's the weather in Istanbul?" | Angular, Admin |
| Stock Prices | "Get AMD's stock price." | Angular, Admin |
| Book Database | "Add a book titled 'AI Revolution'." | Angular, Admin |
| Food Recipe | "Give me the recipe for New York pizza" | Angular, Admin |
| Drug Info | "Tell me about Aspirin." | Angular, Admin |
| Chatbot Users | "List all chatbot users." | Admin only |
- Receive user input via API
- Generate embedding vector of input
- Match it to stored intents
- Use LLM to extract structured intent JSON
- Execute appropriate backend service
You can use tools like Postman or the Swagger UI to test:
- Integrate real Drug APIs via GraphQL
- Support image generation with multimodal models
- Improve Aspect-based role validation messages