Ask Orionfold privately.
A governed local AI advisor over the Orionfold corpus, running on one small desktop. Every answer names the exact source it came from. Questions the corpus cannot support get a refusal, not a guess. The first Advisor model is trained, promoted, and free to download.
Corpus
A sealed pack of 182 sources: books, story posts, model cards, product notes, and field lab material, with a fingerprint and its own recall checks.
Private first
Runs on one small desktop. Questions about private state never leave the machine, and hosted help is allow-listed under a dollar cap.
Inspectable
Every answer names the exact source it came from, and every gate the Advisor passed is a saved receipt a script can re-verify.
Live
The first Advisor shipped in June 2026. The trained model and its test set are free on Hugging Face, and the receipts are public.
Built on the NVIDIA stack
Orionfold is a member of the NVIDIA Inception program. Advisor targets customer-controlled hardware, with NVIDIA GPUs used for accelerated embedding, reranking, and local model inference where available.
See the DGX Spark field lab →© 2025 NVIDIA, the NVIDIA logo, NVIDIA DGX Spark, NVIDIA Inception, NVIDIA NeMo, NVIDIA NIM, and TensorRT-LLM are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries.
