DEV Community

Iceberg DB
Iceberg DB

Posted on

Introducing Iceberg DB: India's Serverless Vector Database & AI Agent Memory Engine (<20ms Latency)

#ai #database #llm #serverless

While building RAG pipelines and autonomous AI agents in India, I kept hitting the same three walls with US-hosted vector databases:

  1. 200ms+ Network Latency: Every vector search from India to us-east-1 adds ~200ms of round-trip latency before the LLM even starts generating.
  2. Paused Free Tiers: Free tier pods pause after a few days of inactivity, breaking live projects and demos.
  3. Boilerplate Overload: Raw vector databases only store float arraysβ€”you still have to wire up document chunking, external embedding APIs, BM25 keyword + dense vector Reciprocal Rank Fusion (RRF), and TTL jobs for agent session memory.

Today, I'm excited to launch Iceberg DB β€” a serverless vector database and cognitive memory platform hosted in Mumbai (ap-south-1).

⚑ What Iceberg DB Gives You Out of the Box

  • <20ms Query Latency in India: Hosted in Mumbai with 100% Indian data residency (DPDP Act 2023 compliant).
  • Built-in Hybrid Search (RRF): Combines dense semantic vectors + sparse BM25 keyword search in a single API call (alpha slider from 0.0 to 1.0).
  • Native 4-Tier AI Agent Memory: Give your AI agents persistent memory across sessions (short_term with 1hr TTL, episodic with 30-day TTL, long_term, and semantic).
  • Built-in Neural Embeddings: Send raw text or upload PDFs directlyβ€”no separate OpenAI or Cohere embedding API key required.
  • Forever Free Tier (Never Pauses): 100,000 vectors free forever, zero cold-start pausing, no credit card required.

πŸš€ 30-Second Quickstart

All 5 official SDKs are live on global package registries today:

  • Python: pip install icebergdb (PyPI)
  • JavaScript / TypeScript: npm install icebergdb (NPM)
  • Rust: cargo add icebergdb (Crates.io)
  • Go: go get github.com/anku5265/iceberg-database/iceberg-sdk-go
  • .NET / C#: dotnet add package IcebergDb (NuGet) 
from iceberg import Client

client = Client(api_key="ib_your_key")

# 1. Create a collection & index raw text (embeddings generated automatically)
client.create_collection("my_docs")
client.index_text("my_docs", "Iceberg DB is hosted in Mumbai with sub-20ms latency.")

# 2. Hybrid Search (Dense + BM25)
results = client.search("my_docs", "where is Iceberg hosted?", search_type="hybrid")
for r in results:
    print(f"{r.score:.2f} β€” {r.text}")

# 3. Persistent Agent Memory
client.memory.remember("agent_01", "User prefers concise Python code", memory_type="long_term")
memories = client.memory.recall("agent_01", "how should I format answers?")
Enter fullscreen mode Exit fullscreen mode

πŸ”— Try It Live

I'd love for you to test it out and share your feedback!

Top comments (0)