🎯 Vector Database

AI vector database for embeddings and similarity search. 30× faster than Pinecone.

Quick Start

Create a Collection

curl -X POST https://api.44s.io:9002/collections \
  -H "Content-Type: application/json" \
  -H "X-API-Key: 44s_your_api_key" \
  -d '{
    "name": "documents",
    "dimension": 1536,
    "metric": "cosine"
  }'

# Response
{
  "name": "documents",
  "dimension": 1536,
  "metric": "cosine"
}

Upsert Vectors

curl -X POST https://api.44s.io:9002/collections/documents/vectors \
  -H "Content-Type: application/json" \
  -H "X-API-Key: 44s_your_api_key" \
  -d '{
    "vectors": [
      {
        "id": "doc-1",
        "values": [0.1, 0.2, 0.3, ...],
        "metadata": {"title": "Introduction to AI", "category": "tech"}
      },
      {
        "id": "doc-2", 
        "values": [0.4, 0.5, 0.6, ...],
        "metadata": {"title": "Machine Learning Basics", "category": "tech"}
      }
    ]
  }'

# Response
{"upserted_count": 2}

Query Similar Vectors

curl -X POST https://api.44s.io:9002/collections/documents/query \
  -H "Content-Type: application/json" \
  -H "X-API-Key: 44s_your_api_key" \
  -d '{
    "vector": [0.15, 0.25, 0.35, ...],
    "top_k": 5,
    "include_metadata": true
  }'

# Response
{
  "matches": [
    {"id": "doc-1", "score": 0.95, "metadata": {"title": "Introduction to AI"}},
    {"id": "doc-2", "score": 0.87, "metadata": {"title": "Machine Learning Basics"}}
  ],
  "namespace": "default"
}

Python Client

import requests
import numpy as np

BASE_URL = "https://api.44s.io:9002"
HEADERS = {
    "Content-Type": "application/json",
    "X-API-Key": "44s_your_api_key"
}

# Create collection
requests.post(f"{BASE_URL}/collections", headers=HEADERS, json={
    "name": "embeddings",
    "dimension": 1536,
    "metric": "cosine"
})

# Upsert with OpenAI embeddings
from openai import OpenAI
client = OpenAI()

texts = ["Hello world", "How are you?", "Machine learning is cool"]
embeddings = client.embeddings.create(input=texts, model="text-embedding-3-small")

vectors = [
    {"id": f"text-{i}", "values": e.embedding, "metadata": {"text": texts[i]}}
    for i, e in enumerate(embeddings.data)
]

requests.post(f"{BASE_URL}/collections/embeddings/vectors", headers=HEADERS, json={
    "vectors": vectors
})

# Query
query_embedding = client.embeddings.create(
    input=["What is ML?"],
    model="text-embedding-3-small"
).data[0].embedding

response = requests.post(f"{BASE_URL}/collections/embeddings/query", headers=HEADERS, json={
    "vector": query_embedding,
    "top_k": 3,
    "include_metadata": True
})

for match in response.json()["matches"]:
    print(f"{match['score']:.3f}: {match['metadata']['text']}")

API Reference

Collections

# Request body
{
  "name": "string",       // Collection name
  "dimension": 1536,      // Vector dimension (must match your embeddings)
  "metric": "cosine"      // cosine | euclidean | dotproduct
}

Vectors

# Request body
{
  "vectors": [
    {
      "id": "string",           // Unique ID
      "values": [0.1, 0.2, ...], // Vector values
      "metadata": {}            // Optional metadata
    }
  ],
  "namespace": "default"        // Optional namespace
}

# Request body
{
  "vector": [0.1, 0.2, ...],    // Query vector
  "top_k": 10,                   // Number of results
  "include_values": false,       // Return vector values?
  "include_metadata": true,      // Return metadata?
  "filter": {}                   // Metadata filter (optional)
}

Distance Metrics

Common Embedding Dimensions

Performance