Mar 31, 2026

From Sandbox to Launchpad: The Free Tier on Aiven for OpenSearch

A free plan with enough RAM for real vector search, k-NN indexing, and long-term memory for your AI projects.

Dima Kan
|RSS Feed
Dima is the Product Director of Search at Aiven, where he spearheads the strategy for managed OpenSearch services. With a focus on developer experience, he leads cross-functional teams to deliver robust, production-ready search solutions that solve complex data infrastructure challenges.

🚀 From Sandbox to Launchpad: The Free Tier on Aiven for OpenSearch

For the creative minds, the tinkerers, and the AI architects: the wait for a high-performance playground is over. Aiven's OpenSearch service provides a truly powerful and versatile tool for real-time search, logging, and analytics. We’ve reimagined what a "Free Tier" looks like by providing a true launchpad for your most ambitious ideas, leveraging the full capabilities of OpenSearch.

While most free services give you just enough to get started, we’re giving you enough to launch.

💎 The Specs: Pro-Level Power

Build your production-ready prototypes with hardware that actually keeps up with your imagination:

4 GB RAM: A robust engine for k-NN (k-Nearest Neighbors) and real-time vector indexing.
20 GB Disk: Generous space for high-density document storage and complex metadata.
Full Feature Access: Unlock the entire OpenSearch toolbox, from hybrid scoring to advanced analytics.

🤖 The Brain for Your AI Agents

This is where your agentic workflows find their home. With 4 GB of RAM, you have the specialized memory needed to build:

Persistent Agentic Memory: Move beyond ephemeral prompts. Give your agents a "long-term memory" where they can store past experiences, learn from user feedback, and retrieve context across sessions.
High-Precision RAG: Create sophisticated Retrieval-Augmented Generation pipelines that blend the nuance of Semantic Search with the surgical accuracy of Keyword Filtering.
Self-Evolving Knowledge Bases: Index codebases, research, or personal data and watch your applications discover connections that stay sharp as your data grows.

📊Search and Analytics Powered Up

Index anything from documents to products to logs and use OpenSearch’s powerful query options, or OpenSearch Dashboards’s flexible UI, accessible directly from your Aiven console.

📈 Mastery in Motion

This tier is designed for those ready to bridge the gap between "tutorial" and "tool." You are moving straight into the architect's chair—building systems that are scalable, fast, and intelligent from day one.

When your project moves towards production workloads, you’re already on the platform built to scale. Transition to a distributed cluster with a single click, keeping your focus exactly where it belongs: on the code.

The infrastructure is ready. Are you?

Get Building

⚡ Quick Start: Index Your First Vector

Initialize a k-NN index in seconds and start feeding your agent's memory:

from opensearchpy import OpenSearch
from sentence_transformers import SentenceTransformer

# 1. Load a lightweight, open-source embedding model
# all-MiniLM-L6-v2 is fast, accurate, and uses only 384 dimensions
model = SentenceTransformer('all-MiniLM-L6-v2')

# 2. Connect to your Aiven OpenSearch launchpad
client = OpenSearch(
   hosts=[{'host': 'os-free-cincinnat-playground.f.aivencloud.com', 'port': 27785}],
   http_auth=('avnadmin', 'your_opensearch_password'),
   use_ssl=True
)

# 3. Create a FAISS-powered Index
index_name = "agent_memory"
index_body = {
   "settings": {"index": {"knn": True}},
   "mappings": {
       "properties": {
           "embedding": {
               "type": "knn_vector",
               "dimension": 384,
               "method": {
                   "name": "hnsw",
                   "engine": "faiss", # High-performance vector engine
                   "space_type": "l2",
                   "parameters": {"ef_construction": 128, "m": 16}
               }
           },
           "context": {"type": "text"},
           "timestamp": {"type": "date"}
       }
   }
}

client.indices.create(index=index_name, body=index_body)

# 4. Upsert a "Memory" (Vector + Metadata)
# Transform text into a vector and Upsert
text_to_remember = "The user prefers Python for backend development."
sample_vector = model.encode(text_to_remember).tolist()

response = client.index(
   index=index_name,
   body={
       "embedding": sample_vector,
       "context": "The user prefers Python for backend development.",
       "timestamp": "2024-03-25T10:00:00Z"
   },
   refresh=True
)

print(f"Memory stored! Document ID: {response['_id']}")
# sample output:  
# Memory stored! Document ID: k4J7B50BgqP7WnAngPAT
Loading code...

Go Further and Faster with Aiven’s Free Tier

Table of contents

🚀 From Sandbox to Launchpad: The Free Tier on Aiven for OpenSearch
💎 The Specs: Pro-Level Power
🤖 The Brain for Your AI Agents
📊Search and Analytics Powered Up
📈 Mastery in Motion
⚡ Quick Start: Index Your First Vector

Stay updated with Aiven

Subscribe for the latest news and insights on open source, Aiven offerings, and more.

Subscribe to RSS