Spring AI Interview Agent

Intent

This project is a playground, learning, and showcase project with the intent to learn while building something useful for myself. One day, I realized that my CV was not very up to date and I had to update it. I experimented with n8n to build an interview loop that would update my CV with additional information. I found it to be a bit too complex and not very flexible and started experimenting in this repo. Now (a few month later), that I am writing this I'd do a few things differently, and I might even do so in the future. But I also learned a lot and I want to share some of my learnings with you.

Learnings

• Start with a larger model, make things work, tune for smaller model
• Saving costs through context caching in system message can be effective
• Split logic whenever possible. Allows for easier testing, failure isolation, parallelization and smaller models

Improvements

• Use graph RAG to steer the interview
• Add STT/TTS voice control

🚀 Features

• AI-Powered Interviews: Uses Spring AI to run an interview loop via an LLM
• Structured Data Collection: Follows a predefined interview plan with topics and threads
• Persistence: Stores career data and interview progress in MongoDB
• Preprocessing: Includes natural language preprocessing to interpret user inputs
• Observability: Spring AI client observations can be enabled/adjusted via config

🏗️ Architecture

The application follows a Spring Boot architecture with the following key components:

• InterviewApplication: Main entry point for the Spring Boot application
• InterviewEngine: Orchestrates message handling, preprocessing, planning and persistence updates
• Interviewer: Uses Spring AI ChatClient to ask/continue questions
• InterviewPlanner: Manages topics/threads for the interview flow
• CareerData: Domain model stored in MongoDB

🛠️ Technical Stack

• Spring Boot 3.4.4: Application framework
• Spring AI 1.0.0-M8: AI capabilities integration
• MongoDB: Data persistence
• JUnit 5: Testing framework
• Testcontainers: Integration testing with containerized dependencies
• Lombok: Reduces boilerplate code

⚙️ Setup & Configuration

Prerequisites

• Java 17+
• Maven
• MongoDB (or Docker, if you only want to run tests with Testcontainers)

Running the Application

# Build the project
mvn clean install

# Run the application
mvn spring-boot:run

Configuration

The application is configured via application.yml:

spring:
  ai:
    chat:
      client:
        observations:
          include-input: false
  data:
    mongodb:
      host: localhost
      port: 27018
      database: interview
      username: mongouser
      password: mongopw

API Key Setup

To run LLM-backed parts (OpenAI), set the API key as an environment variable:

export SPRING_AI_OPENAI_API_KEY=your-api-key-here

Without this environment variable, calls to the OpenAI model will fail.

🧪 Testing

The project includes both unit and integration tests:

# Run all tests
mvn test

# Run a specific test
mvn test -Dtest=InterviewerTest#methodName

🧠 Interview Flow

1. The system follows a structured interview plan with topics and threads
2. Each thread represents a specific area of information to collect
3. The interviewer asks one question at a time and processes the response
4. Responses are stored in MongoDB for later analysis

📝 License

MIT License

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.