Use Case Guide

Natural Language Semantic Search

Build search that understands meaning, not just keywords. Vector Panda auto-tunes its indexing on a sample of your real vectors so the ranking holds up on the queries your users actually run.

From a folder of files to search by meaning

Real Wikipedia articles, one embedding model, runnable as-is

import pathlib, tarfile, urllib.request
from sentence_transformers import SentenceTransformer
from veep import Client

# 40 short Wikipedia landmark articles — swap in your own folder of text files
urllib.request.urlretrieve(
    "https://next.vectorpanda.com/data/example-landmarks.tar.gz", "landmarks.tar.gz")
with tarfile.open("landmarks.tar.gz") as tf:
    tf.extractall(".")

model = SentenceTransformer("all-MiniLM-L6-v2")
files = sorted(pathlib.Path("landmarks").glob("*.md"))

vp = Client.login()
vp.collections.create("landmark-articles", tier="hot")
vp.vectors.upsert("landmark-articles", vectors=[
    {"id": f.stem, "vector": model.encode(f.read_text()).tolist(),
     "metadata": {"title": f.stem}}
    for f in files
])

q = model.encode("an iron lattice tower built for a world's fair").tolist()
for r in vp.vectors.query("landmark-articles", q, top_k=3):
    print(f"{r.metadata['title']}  score={r.score:.3f}")

The full walkthrough — every code block tested against the live service — is the Search your own data tutorial.

Traditional vs Semantic Search

Why semantic search delivers better results

Traditional Keyword Search

❌ Requires exact keyword matches
❌ Misses synonyms and related concepts
❌ Poor with typos and misspellings
❌ Language-specific implementations
❌ Complex query syntax for users

Vector Panda Semantic Search

✅ Understands meaning and intent
✅ Finds synonyms automatically
✅ Handles typos gracefully
✅ Multi-language support built-in
✅ Natural language queries

Semantic Search Features

Everything you need for modern search

🎯

Accuracy You Can Audit

Index accuracy is measured against exact ground truth on your own vectors, not assumed — and the benchmark numbers are published, raw output included.

🌍

Multi-Language

Search across languages seamlessly. A query in English can find results in Spanish, French, or any language your model supports.

🏷️

Metadata Filtering

Combine semantic search with traditional filters. Search by meaning while filtering by price, category, date, or any metadata.

⚡

Real-Time Updates

Add, update, or remove items instantly. No re-indexing needed. Your search results are always up-to-date.

📊

Relevance Scoring

Get similarity scores for each result. Fine-tune thresholds and implement custom ranking logic based on your needs.

🚀

Grows With You

The quickstart's five thousand vectors and the benchmarked million use the same six-line pattern — no re-architecture as your data grows, and storage-only pricing as it does.