# 🤖 simsimi-ai-agent-data

A curated dataset of ~20K Korean youth counseling dialogues designed for building empathetic AI agents. The data originates from AI Hub's public dataset and has been refined and structured for use in Retrieval-Augmented Generation (RAG) pipelines and dialogue modeling tasks.

## 📂 Dataset Details

- **Source**: [AI Hub 청소년상담 데이터](https://aihub.or.kr)
- **Size**: ~20,000 dialogue pairs
- **Language**: Korean
- **Format**: JSONL

### Each record includes:
| Field    | Description                                       |
|----------|---------------------------------------------------|
| `id`     | Unique identifier                                 |
| `user`   | Utterance from the adolescent                     |
| `agent`  | Counselor or chatbot response                     |
| `embedding` | (Optional) KoSBERT vector for similarity search |
| `topic`  | (Optional) Manually assigned topic tag            |

## 💡 Key Features

- Preprocessed and cleaned for safe, domain-specific use  
- Embeddings generated using **KoSBERT** for efficient vector search  
- Ideal for **RAG-based chatbots**, **counseling agents**, and **prompt-tuning**

## 🔧 Usage Example

```python
from datasets import load_dataset

dataset = load_dataset("youdie006/simsimi-ai-agent-data")
print(dataset["train"][0])