A PROJECT FOR CURIOUS MINDS
A question is
a good beginning.
The web has plenty to say. Nuelexity explores how search can become a readable answer with sources you can inspect.
Built for learning, one question at a time.
Follow your curiosityFROM QUESTION TO EVIDENCE
Start with a question. Follow what makes you curious.
Explore an answer, check its sources and let what you learn shape the next question.
An illustrative Nuelexity search: “Help me understand Quantum Computing”, then “How does superposition allow qubits to process multiple states at once?”. Explore sources and an explanation of superposition, interference and measurement. This is a prepared example, not a live search or a response generated by the selected models.
Ask a question
Start with what you want to understand. The example begins with quantum computing.
Inspect the sources
Use the numbered references to open the original material and check the explanation.
Continue with a follow-up
Ask about one part of the answer. The conversation supplies context for the next question.
THE LEARNING BEHIND IT
Making the pieces
work together.
Building Nuelexity meant learning how to give a model useful context: the question, relevant passages from the web and enough of the conversation to keep a follow-up on topic. That is the context engineering behind the answer.
Retrieval brings in sources you can check. Streaming lets you begin reading as the answer arrives. Application development connects those pieces with sign-in, saved threads and shared search limits.
The Return Cat is its reminder to keep asking. We also like cats. Nuelexity is an independent project and is not affiliated with Perplexity.
WHAT IT IS BUILT WITH
A practical, inspectable stack.
- React + TypeScriptReact Router, Vite, Tailwind CSS, Radix controls and safe Markdown rendering.
- Node.js + ExpressThe server handles search requests and streams answers to the interface.
- SupabaseGoogle/GitHub sign-in and PostgreSQL storage for saved conversations.
- Tavily · Groq · TurnstileTavily retrieves web sources. Groq serves the answer models. Cloudflare Turnstile helps protect guest search from bots.
Useful today.
Room to grow.
Available now
Guest research, streamed answers, numbered sources, suggested follow-up questions, Google/GitHub sign-in, private saved threads, conversation navigation and a choice of supported models for signed-in users.
Real limits
Search grounding makes evidence easier to inspect. It cannot guarantee accuracy, complete coverage, reliable sources or a citation for every sentence. Answers may be wrong or incomplete; important decisions need independent verification.
Free provider allowances and shared quotas limit availability. Conversation context and answer length are limited. Guest threads are temporary. There is no public sharing, file upload, subscription or autonomous browsing agent.
BEHIND THE PROJECT
S. Adnan
Built with curiosity, an appetite for learning and a desire to understand how things work.