Serverless vector memory for AI applications.

import requests

# Execute sub-50ms cosine similarity query
res = requests.post(
    "https://lumozai.com/api/v1/memory/search",
    headers={"Authorization": f"Bearer {LUMOZ_API_KEY}"},
    json={
        "project": "customer-knowledge-base",
        "query": "What uptime guarantee is provided for enterprise customers?",
        "top_k": 3
    }
).json()

for item in res["results"]:
    print(f"[{item['score']:.4f}] {item['text']}")

from openai import OpenAI

# Use the standard OpenAI SDK configured with your Lumoz base_url
client = OpenAI(
    api_key=LUMOZ_API_KEY,
    base_url="https://lumozai.com/api/v1"
)

# Pass "memory" in extra_body — Lumoz retrieves relevant chunks & injects context automatically
completion = client.chat.completions.create(
    model="llama-3.3-70b-instruct",
    messages=[{"role": "user", "content": "What is our enterprise uptime SLA?"}],
    extra_body={
        "memory": {"project": "customer-knowledge-base", "top_k": 3}
    }
)

print(completion.choices[0].message.content)

// HTTP 200 OK — Sub-50ms Cosine Dot-Product Retrieval
{
  "status": "success",
  "query": "What uptime guarantee is provided for enterprise customers?",
  "latency_ms": 31.8,
  "results_count": 1,
  "results": [
    {
      "id": "mem_9a8f4c21",
      "score": 0.9412,
      "text": "Enterprise plans guarantee 99.99% uptime and dedicated 24/7 Slack channel support.",
      "metadata": {
        "category": "sla_terms",
        "tier": "enterprise"
      }
    }
  ]
}