Vector Databases (ChromaDB, Qdrant) Indexing

1Concept

Vector databases index millions of high-dimensional embeddings using Approximate Nearest Neighbor (ANN) algorithms (HNSW - Hierarchical Navigable Small World graphs), executing sub-millisecond similarity queries.

2Architecture Diagram

Query Vector ---> [ HNSW Index Graph ] ---> Instant Top-K Nearest Neighbors retrieved in 2ms!

3Code Example

vector_db_sample = '''
# Conceptual Chroma / Qdrant vector database query pipeline
class MockVectorStore:
    def __init__(self):
        self.vectors = {}

    def insert(self, doc_id: str, vector: list[float], metadata: dict):
        self.vectors[doc_id] = (vector, metadata)

    def query(self, query_vector: list[float], top_k: int = 2):
        # HNSW graph query simulation
        return ["doc_101", "doc_104"]

store = MockVectorStore()
store.insert("doc_101", [0.1, 0.2], {"title": "Auth Service"})
print(f"Top matches: {store.query([0.1, 0.2])}")
'''
print("=== Vector Database Architecture ===")
print(vector_db_sample.strip())

4Expected Output

=== Vector Database Architecture ===
# Conceptual Chroma / Qdrant vector database query pipeline
class MockVectorStore:
    def __init__(self):
        self.vectors = {}

    def insert(self, doc_id: str, vector: list[float], metadata: dict):
        self.vectors[doc_id] = (vector, metadata)

    def query(self, query_vector: list[float], top_k: int = 2):
        # HNSW graph query simulation
        return ["doc_101", "doc_104"]

store = MockVectorStore()
store.insert("doc_101", [0.1, 0.2], {"title": "Auth Service"})
print(f"Top matches: {store.query([0.1, 0.2])}")

5Key Takeaways

• ✓HNSW graphs balance recall accuracy with logarithmic search latency.
• ✓Store document text and metadata alongside vector embeddings for fast retrieval.
• ✓Supports metadata filtering (e.g. `where={'tenant_id': 'corp_12'}`).