Spaces:
Running
Running
File size: 3,398 Bytes
21c1bff | 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 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 | # embedding_engine.py
import uuid, time
from qdrant_client import QdrantClient, models
from qdrant_client.http.models import Distance, VectorParams
from sentence_transformers import SentenceTransformer
from app.core.config import QDRANT_URL, QDRANT_API_KEY
MODEL_PATH = "app/core/models/bge-base-en-v1.5"
embedder = SentenceTransformer(MODEL_PATH)
qdrant = QdrantClient(
url=QDRANT_URL,
api_key=QDRANT_API_KEY,
check_compatibility=False
)
COLLECTION_NAME = "smartnotes"
BATCH_SIZE = 5 # ✅ reduced for free tier
def ensure_collection():
collections = qdrant.get_collections().collections
if COLLECTION_NAME not in [c.name for c in collections]:
qdrant.create_collection(
collection_name=COLLECTION_NAME,
vectors_config=VectorParams(size=768, distance=Distance.COSINE),
)
qdrant.create_payload_index(
collection_name=COLLECTION_NAME,
field_name="doc_id",
field_schema="keyword"
)
def embed_and_store(text_chunks, doc_id):
print(f"📊 Final chunks being embedded: {len(text_chunks)}")
ensure_collection()
vectors = embed_documents(text_chunks) # ✅ now uses correct doc prefix
points = [
models.PointStruct(
id=str(uuid.uuid4()),
vector=vectors[i],
payload={
"doc_id": doc_id,
"text": text_chunks[i],
"chunk_id": i,
"length": len(text_chunks[i])
},
)
for i in range(len(vectors))
]
failed_batches = []
for i in range(0, len(points), BATCH_SIZE):
batch = points[i:i + BATCH_SIZE]
batch_num = i // BATCH_SIZE + 1
success = False
for attempt in range(4): # ✅ 4 attempts with exponential backoff
try:
qdrant.upsert(collection_name=COLLECTION_NAME, points=batch)
success = True
print(f" → Batch {batch_num} uploaded")
break
except Exception as e:
wait = 2 ** attempt # 1s, 2s, 4s, 8s
print(f" ⚠️ Batch {batch_num} attempt {attempt+1} failed: {e} | retrying in {wait}s")
time.sleep(wait)
if not success:
failed_batches.append(batch_num)
print(f" ❌ Batch {batch_num} permanently failed")
time.sleep(0.6) # ✅ throttle between successful batches
if failed_batches:
# ✅ raise so the caller (routes.py) knows something went wrong
raise RuntimeError(f"Failed to upload batches: {failed_batches}")
print(f"✅ All batches uploaded for doc_id={doc_id}")
def embed_documents(texts):
"""Embed document chunks with correct BGE prefix and normalization."""
prefixed = [f"Represent this sentence: {t}" for t in texts] # ✅ correct BGE doc prefix
vectors = []
for i in range(0, len(prefixed), 32):
batch = prefixed[i:i + 32]
batch_vectors = embedder.encode(
batch, normalize_embeddings=True, show_progress_bar=False)
vectors.extend(batch_vectors.tolist())
return vectors
def embed_query(text):
"""Embed a search query — BGE uses 'query:' prefix for retrieval."""
return embedder.encode(
f"query: {text}",
normalize_embeddings=True
).tolist() # ✅ always return list, not numpy array
|