hgzero/rag/config.yaml
2025-10-29 05:54:08 +09:00

96 lines
1.9 KiB
YAML

# Vector DB 통합 시스템 설정
# PostgreSQL (용어집)
postgres:
host: ${POSTGRES_HOST}
port: ${POSTGRES_PORT}
database: ${POSTGRES_DATABASE}
user: ${POSTGRES_USER}
password: ${POSTGRES_PASSWORD}
pool_size: 10
max_overflow: 20
# Azure OpenAI (Embedding)
azure_openai:
api_key: ${AZURE_OPENAI_API_KEY}
endpoint: ${AZURE_OPENAI_ENDPOINT}
embedding_model: text-embedding-ada-002
embedding_dimension: 1536
api_version: "2023-05-15"
# Azure AI Search (관련자료)
azure_search:
endpoint: ${AZURE_SEARCH_ENDPOINT}
api_key: ${AZURE_SEARCH_API_KEY}
index_name: meeting-minutes-index
api_version: "2023-11-01"
# Claude AI
claude:
api_key: ${CLAUDE_API_KEY}
model: claude-3-5-sonnet-20241022
max_tokens: 1024
temperature: 0.3
# Redis (캐싱)
redis:
host: redis
port: 6379
db: 0
password: ${REDIS_PASSWORD}
decode_responses: true
# Azure Event Hub
eventhub:
connection_string: ${EVENTHUB_CONNECTION_STRING}
name: ${EVENTHUB_NAME}
consumer_group: ${AZURE_EVENTHUB_CONSUMER_GROUP}
storage:
connection_string: ${AZURE_STORAGE_CONNECTION_STRING}
container_name: ${AZURE_STORAGE_CONTAINER_NAME}
# Application Settings
app:
name: "Vector DB Service"
version: "1.0.0"
debug: true
log_level: INFO
# 용어집 설정
term_glossary:
# 검색 설정
search:
top_k: 5
confidence_threshold: 0.7
keyword_weight: 0.6
vector_weight: 0.4
# 캐싱 설정
cache:
ttl: 3600 # 1시간
prefix: "term:"
# 관련자료 설정
related_documents:
# 검색 설정
search:
top_k: 3
relevance_threshold: 0.70
folder_weight_boost: 0.20
semantic_ranking: true
# 캐싱 설정
cache:
ttl: 3600 # 1시간
prefix: "doc:"
# 데이터 로딩
data:
terms_dir: design/aidata
terms_files:
- terms-01.json
- terms-02.json
- terms-03.json
- terms-04.json
documents_file: design/aidata/meet-ref.json