Questions, answered plainly.
Do I have to adopt the whole platform?
No. ORIS stands alone; most teams land there and add VERI when each source has to be defensible on its own, and MESH when the answer depends on how the sources relate to each other. Everything is usage-based — start where the problem is and compose as you go.
Does Orves lock me into its intelligence?
The opposite. VERI takes the output of whatever parser you trust, and what the chain produces is an asset you own — model-ready for any model you choose: retrieval, fine-tuning, training data. Being served by Orves is a choice, never a requirement.
What do you mean by “verifiable”?
You can verify the origin of the information, the evidence behind it, where it sits in the source document, its version and its history. Certification of every transformation on an append-only ledger — with public, account-free verification — is the designed direction and ships with the rollout.
What does Orves VERI produce?
One verifiable result per source: every claim it contains, tied to the exact span it came from, with authorship, qualifiers, within-source relationships and an explicit ABSTAINED wherever the source does not support the claim. Relationships across sources are Orves MESH, which runs on these results.
Where do benchmark numbers come from?
From pre-registered evaluations against versioned gold sets, with n and confusion matrices reported. A number appears on this site only after it survives that process — until then you see the honest state.
How is Orves delivered?
Exclusively through the API. Nothing to install, nothing to host. Thin SDKs are provided for convenience — the platform lives server-side.
Is Orves a RAG stack?
No. RAG is one thing you can build on top. Orves governs the verifiable knowledge layer that RAG stacks improvise.