Open source · Apache-2.0

dynavec

The serverless vector database that lives in your AWS account. Here's the 90-second tour.

The problem

Every vector DB forces a trade-off.

PineconeManaged — your data leaves your account.

OpenSearchCostly, always-on capacity even when idle.

Qdrant · Weaviate · MilvusYou run and scale the servers.

At a billion vectors, that's tens of thousands of dollars a month.

The goal

What if your vector DB was just… AWS?

Serverless — you pay only when you use it
In your account — your region, your compliance
Billions of vectors, no servers to run

How it works

Two AWS primitives, joined by a key.

Embedyour key → S3 Vectorsbillion-scale ANN → DynamoDBdocs in ~ms → Results

S3 Vectors finds the nearest keys. DynamoDB hydrates the documents in single-digit milliseconds. dynavec joins them.

The proof · 1B vectors, 1536-d

Same scale. A fraction of the cost.

$469dynavec / month vs $80k+managed clusters / month

Storage priced like S3, not RAM. (Cost-model estimate, public list prices.)

What's inside

Batteries included.

Pluggable embedders · metrics & rerank · namespaces
Knowledge graph · caching · product quantization
MCP ingestion · LangChain · LlamaIndex · CrewAI

Contribute

It's open source — and we need you.

pip install dynavec — try it in your account
100+ issues, labeled by area & difficulty
Plenty of good first issues waiting

Get started

Star it. Read the docs. Pick an issue.

github.com/codeforstartups/dynavec

Star on GitHub Read the docs Browse issues

Want a recorded walkthrough instead?

The full narration script + storyboard is in VIDEO_SCRIPT.md. Record it (Loom, OBS, or a screen capture of this page), upload to YouTube/Loom, then paste the embed URL into explainer.html where marked to replace this animation with your video.