Spaces:
Sleeping
Sleeping
File size: 1,142 Bytes
0df7243 c20fa4f 0df7243 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 |
# vectordb_utils.py
from qdrant_client import QdrantClient
from qdrant_client.models import VectorParams, Distance, PointStruct
from sentence_transformers import SentenceTransformer
import uuid
cache_dir = os.environ.get("MODEL_CACHE_DIR", "/app/cache") # Fallback to /app/cache
os.makedirs(cache_dir, exist_ok=True)
encoder =SentenceTransformer("all-MiniLM-L6-v2", cache_folder=cache_dir)
qdrant = QdrantClient(":memory:")
collection_name = "customer_support_docsv1"
def init_qdrant_collection():
qdrant.recreate_collection(
collection_name=collection_name,
vectors_config=VectorParams(size=384, distance=Distance.COSINE)
)
def add_to_vectordb(query, response):
vector = encoder.encode(query).tolist()
qdrant.upload_points(
collection_name=collection_name,
points=[PointStruct(
id=str(uuid.uuid4()),
vector=vector,
payload={"query": query, "response": response}
)]
)
def search_vectordb(query, limit=3):
vector = encoder.encode(query).tolist()
return qdrant.search(collection_name=collection_name, query_vector=vector, limit=limit)
|