Search the live web.
Synthesize answers.
Oryn is an open-source, citation-first search assistant that crawls references dynamically, routing requests to powerful LLMs for clear, verified answers.
Experience Real-Time Web RAG
Select a question to inspect Oryn's background crawling and synthesis pipeline
Engineered for Accuracy
We combine deep indexing algorithms with rapid inference pipelines for verified knowledge generation.
Live Web Crawling
Queries search engines dynamically, fetches target pages, and parses DOM trees instantly to pull latest context.
Supabase Sync
Connects directly to your own Supabase PostgreSQL database for persistent, encrypted message storing and query caching.
Multi-Model Routing
Routes queries to Groq-hosted open-weights models like Llama 3.3 for high-speed, cost-effective inference.
How Oryn Answers Questions
A four-step pipeline structured to avoid LLM hallucinations
Query Analysis
The engine identifies query intent, extracts keywords, determines whether it needs web search, and reformulates query variations.
Parallel Crawling
Fetches direct search links. We spin up asynchronous page parsers to download raw text and meta structures from the top-ranking web domains.
Context Compacting
Parsed documents are ranked. Only relevant text fragments are packed into the LLM system prompt context, discarding useless markup and ads.
Synthesized Output
The LLM crafts the answer while forcing strict inline referencing. The result is a readable response backed by clickable references.
Why Choose Oryn?
How we compare to traditional search engines and static LLMs
| Feature | Oryn AI | Traditional Search | Static LLM |
|---|---|---|---|
| Answers Complex Queries | Yes (Synthesized) | No (List of links) | Yes (But cuts off) |
| Live Web Access | Yes (Real-time crawling) | Yes (Indexed sites) | No (Static cutoff) |
| Inline Citations | Yes (Clickable sources) | No | Rarely (Unreliable) |
| Privacy Control | Yes (Own DB config) | No (Ad-tracked profiles) | No (Data training) |
Flexible Access Tiers
Get started for free or upgrade for professional search limits
Hobby
For casual everyday queries
- Llama 3.1 Inference models
- Standard Live Web Search
- 100 Queries per day
Creator
For power users and developers
- Access to Llama 3.3 70B models
- Deep multi-page site crawling
- Unlimited chat history storage
- Priority Groq API bandwidth
Enterprise
For corporate internal query lines
- Self-hosted Docker configurations
- Dedicated API key integrations
- SLA response uptime metrics
Frequently Asked Questions
Everything you need to know about the engine operations