Code-Buddy — Backend

A Spring Boot REST API that powers the Code-Buddy Chrome extension. It receives selected code, builds a structured prompt, and returns an AI-generated response from either a local Ollama model or HuggingFace cloud inference.

---

Tech Stack

Java 17+	Core language
Spring Boot 3	REST API framework
Spring Profiles	Dev/prod provider switching
RestClient	Outbound HTTP to AI providers
Lombok	Boilerplate reduction
Ollama	Local LLM for development
HuggingFace Inference API	Cloud LLM for production

---

Project Structure

src/main/java/com/buddy/Code_Buddy/
├── Controller/
│   └── ResearchController.java          # POST /research/process
├── Clients/
│   ├── AiService.java                   # Interface: generate()
│   ├── OlamaAiServiceImpl.java          # Active on dev profile
│   ├── HuggingFaceAiServiceImpl.java    # Active on prod profile
│   └── PromptBuilder.java               # Builds prompts per operation
├── DTO/
│   ├── ResearchRequest.java             # content + operation
│   ├── OlamaModelRequest/Response.java
│   └── HuggingFaceRequest/Response.java
├── Advice/
│   ├── ApiError.java
│   ├── ApiResponse.java
│   └── GlobalExceptionHandler.java
└── Configurations/
    └── RestConfig.java                  # RestClient bean

---

Getting Started

Prerequisites

• Java 17+, Maven
• Ollama for dev, or a HuggingFace token for prod

Setup

git clone https://github.com/your-username/code-buddy.git
cd code-buddy

# Dev only — pull the local model
ollama pull qwen2.5-coder:3b

Set environment variables in application.properties or your shell:

BASE_URL=http://localhost:11434            # dev (Ollama)
# BASE_URL=https://router.huggingface.co  # prod (HuggingFace)
HF_TOKEN=your_token_here                  # prod only

Start the server:

# Dev — local Ollama
./mvnw spring-boot:run -Dspring-boot.run.profiles=dev

# Prod — HuggingFace
./mvnw spring-boot:run -Dspring-boot.run.profiles=prod

Server runs at http://localhost:8080.

---

API

POST /research/process

Request

{
  "content": "public String generate(...) { ... }",
  "operation": "EXPLAIN"
}

operation accepts: SUMMARIZE, EXPLAIN, REFACTOR

Success — 200 OK, Content-Type: text/plain, markdown-formatted response body.

Error

{
  "error": {
    "message": "Content is empty",
    "status": "INTERNAL_SERVER_ERROR",
    "timestamp": "2024-06-01T10:30:00Z"
  }
}

---

AI Providers

	Dev (Ollama)	Prod (HuggingFace)
Model	qwen2.5-coder:3b	Qwen/Qwen3.5-397B-A17B
Endpoint	POST /api/generate	/v1/chat/completions
Data	Stays local	Sent to cloud
Streaming	Disabled	Temperature 0.3, max 800 tokens

Spring Profiles handle the switch — no code changes needed between environments.

---

Configuration

Variable	Description	Required
BASE_URL	AI provider base URL	Always
HF_TOKEN	HuggingFace Bearer token	Prod only
spring.profiles.active	dev or prod	Always

---

Error Handling

All exceptions flow through GlobalExceptionHandler and return a consistent error shape.

Situation	Status
Empty or null AI response	404
Missing content or operation	500
Unknown operation value	500
Unexpected failure	500

Link for Extension(Chrome) Repository