IndexPulse

Real-time vector index manager — a lightweight, in-memory Python library for managing, monitoring, and optimizing vector indexes for similarity search.

Inspired by vector DB trends (Milvus, Qdrant, Chroma), IndexPulse strips away the server overhead and gives you a pure-Python index you can embed anywhere.

---

Architecture

graph TD
    A[Client Code] -->|add / search / delete| B[VectorIndex]
    B --> C[Distance Engine]
    C --> D[Cosine Similarity]
    C --> E[Euclidean Distance]
    C --> F[Dot Product]
    B --> G[Storage Layer]
    G -->|save| H[Disk — .npz + JSON]
    G -->|load| B
    B --> I[Stats & Monitoring]
    I --> J[Index Health]
    I --> K[Dimensionality Report]
    I --> L[Memory Usage]
    B --> M[Optimizer]
    M -->|normalize + reindex| B

Loading

Features

• Multiple distance metrics — cosine similarity, euclidean distance, dot product
• Metadata filtering — attach arbitrary JSON metadata to every vector
• Persistence — save/load indexes to disk (numpy .npz + JSON sidecar)
• Index stats & monitoring — vector count, dimensionality, memory estimates
• Optimization — normalize vectors, compact storage
• Zero infrastructure — no server, no containers, just pip install

Quickstart

pip install indexpulse

import numpy as np
from indexpulse import VectorIndex

# Create an index
index = VectorIndex(dimension=128)

# Add vectors with metadata
for i in range(1000):
    vec = np.random.randn(128).astype(np.float32)
    index.add(id=f"doc-{i}", vector=vec, metadata={"source": "wiki", "page": i})

# Search
query = np.random.randn(128).astype(np.float32)
results = index.search(query, k=5, metric="cosine")

for result in results:
    print(f"{result.id}  score={result.score:.4f}  meta={result.metadata}")

# Persist to disk
index.save("my_index")

# Reload later
index2 = VectorIndex.load("my_index")

# Monitor health
print(index.get_stats())

Configuration

from indexpulse import VectorIndex, IndexConfig

config = IndexConfig(
    dimension=256,
    default_metric="cosine",
    normalize_on_add=True,
    max_vectors=100_000,
)

index = VectorIndex.from_config(config)

Development

git clone https://github.com/officethree/IndexPulse.git
cd IndexPulse
make install    # editable install + dev deps
make test       # run pytest
make lint       # ruff + black --check

License

MIT — see LICENSE for details.

---

Built by Officethree Technologies | Made with ❤️ and AI