Overview
Quark is how most people use IdonAI. Open the app to chat, write code, or work through a hard problem. Under the hood it runs Origin for speed and NextGen for deeper reasoning — the same models you call from the API.
In Quark
Preview · opens the live app
Open QuarkHow to use Quark
Open Quark
Start in the web app — ask questions, write code, or reason through a problem.
Open Quark →Pick a model
Origin for speed. NextGen for deeper reasoning. Quark routes when you don’t choose.
View models →Build with the API
Same models behind Quark. The API contract and SDK sketches are ready — keys under development.
Read docs →Capabilities
Conversation
Multi-turn dialogue that stays coherent — questions, drafts, and follow-ups in one thread.
Code assistance
Explain, debug, and iterate on code across languages without leaving the chat.
Model routing
Origin when latency matters; NextGen when the task needs longer context or deeper reasoning.
API parity
What you try in Quark maps to the same model IDs and endpoints in production.
Architecture
- 01
Model layer
Origin and NextGen with routing by task complexity — or pick a model explicitly.
- 02
Safety layer
Filtering, misuse detection, and clear limits at inference time.
- 03
Session memory
Context across turns so follow-ups stay grounded in the same thread.
Use Cases
Software engineering
Debug, explain algorithms, and review implementations.
Research & writing
Summarize sources, outline arguments, and stress-test ideas.
Learning
Step-through explanations across math, science, and code.
Prototyping for API
Validate prompts in Quark, then prepare the same model IDs for the API when keys ship.
Technical Features
- Web assistant at quark.idonai.com
- Origin / NextGen routing
- API contract ready — keys under development
- SDK sketches (Python, TypeScript, and more)
- Enterprise deployment options
API
Call the same models that power Quark — Origin and NextGen — through the IdonAI API.
FAQ
Related Products
[ Start ]
Open Quark, then ship with the API.
Try the assistant first. When you’re ready to integrate, the same models are one endpoint away.