zepdb

zepdb / zeppelin

Public

Open-source, S3-native vector and full-text search engine. Fast, cheap, self-hostable.

bm25 database fulltextsearch objectstorage s3
38
5
100% credibility
Found Feb 20, 2026 at 17 stars 2x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

Zeppelin is a server that stores and searches AI embeddings directly in cloud object storage like S3, supporting similarity search, filters, and text matching.

How It Works

1
🔍 Discover Zeppelin

You hear about Zeppelin, a simple tool that lets you quickly find similar items, images, or texts using AI embeddings stored in cloud storage.

2
🚀 Start Locally

With one easy command, you launch a local version on your computer to test it out right away.

3
📦 Create Your Collection

You make a new space for your data, telling it the size of your embeddings, and get a unique name to use.

4
➕ Add Your Data

You upload batches of your embeddings with labels and extra details, and they save securely.

5
🔎 Find Matches Instantly

You search with a new embedding and get the closest matches ranked by similarity in seconds.

6
Add Smarts
🔤
Text Search

Find items by keywords in descriptions using smart ranking.

📊
Filter Results

Narrow down by numbers, lists, or conditions like price ranges.

7
☁️ Go Live on Cloud

Switch to real cloud storage like S3, and your search engine handles big data effortlessly.

🎉 Your Search Shines

Now you have a fast, scalable AI-powered search that works on millions of items, ready for your app.

Sign up to see the full architecture

6 more

Sign Up Free

Star Growth

See how this repo grew from 17 to 38 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is zeppelin?

Zeppelin is a Rust-built, open-source vector and full-text search engine that runs directly on S3-compatible storage like MinIO, treating objects as the source of truth for stateless, self-hostable nodes. You create namespaces via REST API, upsert vectors with metadata attributes, query for nearest neighbors using IVF or hierarchical indexes with bitmap filters, and run BM25 text searches across fields—all without managing a separate database. Docker Compose spins it up locally in seconds, with Python and TypeScript SDKs handling the full API including deletes and strong/eventual consistency.

Why is it gaining traction?

Unlike managed vector DBs charging per query or GB, Zeppelin's S3-native approach leverages your existing cheap object storage for infinite scale, sub-second filtered searches, and hybrid vector+text ranking via simple rank-by expressions. As an Apache-2.0 licensed open source GitHub tool, it skips vendor lock-in while offering Prometheus metrics, rate limiting, and concurrency controls—ideal for devs eyeing self-hosted alternatives to cloud-heavy stacks.

Who should use this?

Backend engineers building RAG apps on S3/MinIO who want fast ANN with attribute filters and text search without infra overhead. AI teams prototyping semantic search over docs or logs, or ops folks self-hosting vector indexes alongside existing object workflows.

Verdict

With 15 stars and 1.0% credibility score, it's early but impressively mature: full docs, Docker quickstart, client SDKs, and extensive benchmarks/fuzzing. Try it for cheap, self-hosted vector needs if you tolerate namespace UUIDs—fork from the zepdb/zeppelin GitHub repo today.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.