Javad

Posted on Jan 1

🚀 Atoqu — The Atomic Quantum Search Engine

#ai #discuss #opensource #programming

Yes! It's a Manifesto for the Next Generation of Search

(and yes… v1.2 is coming very soon)

Hey Dev Community 👋

This is not just another “we built a search engine” post.

This is a manifesto — a declaration of intent, a spark, a challenge to the old world of search.

Atoqu is the name.

Atomic + Quantum.

A search engine core built from scratch by OverLab Group — a young, creative, fearless team that believes the future of search should be fast, open, independent, and beautifully engineered.

Let’s dive in.

⚛️ Atomic — True Atomic Architecture

Atoqu is built on a simple philosophy:

“Every component should be small, independent, testable, and replaceable.”

This means:

Zero external dependencies
No legacy baggage
No heavy frameworks
No JVM, no Python, no Lucene, no FAISS
Pure C++17
GPU kernels written by hand
Every module is an atomic unit

This is something Elasticsearch, Solr, Vespa, and Lucene simply cannot offer.

🧬 Quantum — Multi‑Mode, Probabilistic, Future‑Ready Search

Quantum in Atoqu is not a buzzword.

It’s a design principle:

Multi‑Mode Ranking
Literal + Vector + Hybrid + BM25 + Recency + TagBoost
Probabilistic scoring
Collapse‑style ranking
Parallel evaluation
Embedding‑aware search
LLM‑ready architecture

Atoqu doesn’t just search — it interprets, evaluates, and adapts.

⚡ Speed — Beyond Google Core, Beyond Elasticsearch

Let’s be honest:

Google has optimized its search core for 20–30 years.

Elasticsearch has been tuned for more than a decade.

But they carry massive legacy weight.

Atoqu doesn’t.

Atoqu is:

Pure C++
Zero dependency
GPU‑accelerated
CUDA kernels
OpenCL backend
Zero‑copy flattening
Memory reuse
Atomic architecture

The result?

Atoqu is built like a Tesla Model X Plaid.

Google Core is a 30‑year‑old gasoline car with a turbo strapped on.

One is born fast.

The other is trying to stay fast.

🎮 GPU Support — Real, Native, Production‑Ready

Atoqu is the only modern open‑source search engine core with real GPU support:

✅ CUDA backend (production‑ready)
✅ OpenCL backend (production‑ready host side)
✅ GPU‑aware VectorStore
✅ Cosine similarity on GPU
✅ Batch processing
✅ Buffer reuse
✅ Safe stubs for Vulkan / Metal / SYCL / HIP

Elasticsearch? No.

Lucene? No.

DuckDuckGo? No.

Brave? No.

Google? Yes — but behind closed doors.

Atoqu brings GPU search to the open‑source world.

🧠 LLM‑Ready — Built for the Future

Atoqu includes:

HashEmbeddingProvider
LlmEmbeddingProvider
Semantic search
Hybrid ranking
Embedding‑aware scoring

Plug in any embedding model — local or cloud — and Atoqu adapts instantly.

🔮 OLFCE — OverLab Forward Compatibility Engine

This is one of Atoqu’s secret weapons.

Big companies struggle with:

Legacy APIs
Backward compatibility
Old systems they can’t remove
Technical debt that grows forever

Atoqu solves this with OLFCE:

Old APIs can be removed safely
New APIs automatically get aliases
Old code keeps working
New code stays clean
Zero legacy debt
Zero backward‑compatibility nightmares

This is something Google, Bing, Elasticsearch, and Brave simply cannot do.

🧱 Full Feature List (v1.2)

✅ Core Engine

DocumentStore
VectorStore (GPU‑aware)
Multi‑mode ranking
Embedding providers
Optimizer
Indexer

✅ GPU Backends

CUDA
OpenCL
Vulkan (stub)
Metal (stub)
SYCL (stub)
HIP (stub)

✅ Modes

NormalMode
LiteralMode
VectorMode
HybridMode
BM25Mode
RecencyMode
TagBoostMode

✅ Tooling

CMake
Makefile
Docker (CPU/GPU)
GitHub Actions
GitLab CI
Sanitizers
Static Analysis
Benchmarks

✅ Documentation

Doxygen
Sphinx
Breathe
Architecture docs
Full changelog

✅ License

Apache 2.0 (tentative — see community questions below)

🗣️ Community Questions

Atoqu is not released yet — and we want your voice in shaping it.

✅ 1) What license should Atoqu use?

We’re considering:

Apache v2.0
MPL 2.0
GPLv3
AGPLv3

Which one do you think fits a next‑generation search engine core?

✅ 2) How fast do you think Atoqu is compared to Google Core?

We know our internal benchmarks.

But we want to hear your predictions:

2× faster?
5× faster?
10× faster?
20× faster?
More?

Drop your guess in the comments — let’s see what the community thinks.

✅ 3) What features do you want in v2.0?

We’re planning:

Distributed indexing
Web crawler
Sharding
Query planner
Ranking pipeline
Caching layer
Quantum‑inspired search algorithms

But we want your ideas too.

🤝 Join the Atoqu Community

When Atoqu v1.2 is released, we invite you to:

✅ Explore the code

✅ Read the architecture

✅ Run the benchmarks

✅ Test GPU kernels

✅ Open issues

✅ Submit PRs

✅ Star the repo

✅ Join discussions

✅ Help shape the future of search

Atoqu is not just a project.

It’s a movement.

🙏 Thank You — and See You Soon

Thanks for reading this manifesto.

Thanks for being part of this journey.

Thanks for believing that search can be faster, cleaner, smarter, and more open.

Atoqu v1.2 is coming soon.

Stay atomic. Stay quantum. Stay curious.

— OverLab Group

— Atoqu Project

🚀💛