collections dict | aliases unknown |
|---|---|
{
"kr_news": {
"vectors": {
"": {
"size": 1024,
"distance": "Cosine",
"hnsw_config": null,
"quantization_config": null,
"on_disk": null,
"datatype": null,
"multivector_config": null
}
},
"shard_number": null,
"sharding_method": null,
... | {} |
Kanana Agent Data
Kanana Agent 멀티 에이전트 시스템에서 사용하는 사전 수집 데이터입니다.
프로젝트 루트의 ./data/ 경로에 그대로 두면 각 에이전트가 참조합니다.
Stock Agent, Report Agent는 런타임 크롤링 또는 별도 데이터가 필요 없으므로 이 데이터셋에 포함되지 않습니다.
디렉터리 구조
data/
├── legal_data/ # Legal Agent — 법률 RAG DB
│ └── LawDB
├── news_data/ # News Agent — 뉴스 Vector DB
│ └── collection/
└── trend_data/ # Trend Agent — 금융 트렌드 데이터
└── chroma_db_bge/
에이전트별 사용 데이터
| 에이전트 | 경로 | 설명 |
|---|---|---|
| Legal Agent | legal_data/LawDB |
법률 문서 벡터 DB |
| News Agent | news_data/collection |
뉴스 기사 벡터 DB |
| Trend Agent | trend_data/chroma_db_bge |
KCIF 기사 Chroma DB |
다운로드 방법
Hugging Face CLI
pip install huggingface_hub
# data/ 폴더 전체 다운로드 (Kanana_Agent 프로젝트 루트에서 실행)
huggingface-cli download munchkincat/Kanana_Agent-data --repo-type dataset --local-dir ./data
Python
from huggingface_hub import snapshot_download
snapshot_download(
repo_id="munchkincat/Kanana_Agent-data",
repo_type="dataset",
local_dir="./data",
local_dir_use_symlinks=False,
)
다운로드 후 ./data/ 아래에 위 디렉터리 구조가 생성되면 됩니다.
데이터 출처
| 데이터 | 출처 |
|---|---|
| Legal Agent | 국가법령정보센터 |
| News Agent | 한국경제, 머니투데이 |
| Trend Agent | KCIF 국제금융센터 |
업데이트
데이터는 크롤링·DB 구축 시점 기준이며, 이후 주기적으로 갱신될 수 있습니다.
- Downloads last month
- 10