qdrant_27.py

from qdrant_client import QdrantClient
from qdrant_client.models import PointStruct, PointsBatch, Batch, Filter, FieldCondition, MatchValue, VectorParams, Distance
from openai import OpenAI
from langchain_core.documents import Document
from langchain_community.document_loaders import TextLoader
from langchain_openai.embeddings import OpenAIEmbeddings
from dotenv import load_dotenv
import os
import uuid

MEMORY_PATH = "memory.md"
COLLECTION_NAME = "ai_devs"

qdrant = QdrantClient(url=os.getenv("QDRANT_URL"))
embeddings = OpenAIEmbeddings()
query = "Do you know the name of Adam's dog?"
result = qdrant.get_collections()
indexed = next((element for element in result.collections if element.name == COLLECTION_NAME), None)

print(result)

if not indexed:
  qdrant.create_collection(COLLECTION_NAME, vectors_config=VectorParams(size=1536, distance=Distance.COSINE, on_disk=True))

collectionInfo = qdrant.get_collection(COLLECTION_NAME)

if collectionInfo.points_count == 0:
  loader = TextLoader(MEMORY_PATH)
  memory = loader.load()
  documents = list()

  for mem in memory:
    lines = mem.page_content.split("\n\n")

    for line in lines:
      documents.append(Document(line))
  
  for document in documents:
    document.metadata["source"] = COLLECTION_NAME
    document.metadata["content"] = document.page_content
    document.metadata["uuid"] = str(uuid.uuid4())

  points = list()

  for document in documents:
    embedding = embeddings.embed_documents(list(document.page_content), chunk_size=1)
    
    points.append(PointStruct(
      id=document.metadata["uuid"],
      payload=document.metadata,
      vector=embedding[0]
    ))

  batchPoints = PointsBatch(batch=Batch(
    ids=list(map(lambda p: p.id, points)),
    payloads=list(map(lambda p: p.payload, points)),
    vectors=list(map(lambda p: p.vector, points))
  ))
  qdrant.upsert(COLLECTION_NAME, points=batchPoints, wait=True)

queryEmbedding = embeddings.embed_query(text=query)
search = qdrant.search(
  COLLECTION_NAME, 
  query_vector=queryEmbedding, 
  limit=1, 
  query_filter=Filter(must=[
    FieldCondition(
      key="source", 
      match=MatchValue(value=COLLECTION_NAME)
    )
  ]))

print(search)