How It Works

From a question to a verdict you can trust.

Every OMNYA answer runs through a five-stage pipeline — Plan, Search, Verify, Score, Explain — grounded in evidence and scored before it ever reaches you.

STEP 01Planner LLM

Plan

Expands the query into sub-questions and selects the trusted domains worth searching.

STEP 02Retriever + Source-Rating DB

Search

Fetches live evidence and weights it against a graph of 200k+ rated sources.

STEP 03Answerer + Critic LLMs

Verify

Grounds a cited answer, then a critic network checks it for contradiction and bias.

STEP 04Scoring Engine

Score

Five sub-scores (SCS · FAS · TVS · BBI · CRS) collapse into one Credibility Score.

STEP 05Output Layer

Explain

Returns a JSON verdict, a dashboard view, and an auditable PDF report with citations.

Every Drax answer arrives with a score, its sources, and an audit trail.

That is not a feature — it is the whole point of verifying before you speak.

Credibility Scoring Engine

One score. Five dimensions of truth.

Every Drax answer collapses into a single Credibility Score — composed from five independent sub-scores you can inspect and audit.

SCS

Source Credibility Score

How authoritative and well-rated the cited sources are, drawn from the Source-Rating Graph.

FAS

Factual Alignment Score

How tightly the answer aligns with the retrieved, verified evidence.

TVS

Temporal Validity Score

Whether the claim still holds at the time of asking — no stale facts.

BBI

Bias Balance Index

Ideological balance across sources, surfacing one-sided evidence.

CRS

Contradiction Risk Score

How much the source set disagrees — lower is better.

Pre-Series A · Raising $25–30M

Building the Bloomberg Terminal of Truth.

AI changed how we access information. OMNYA — through Drax and Veritas — will change how we trust it. Talk to us about the round, a pilot, or a partnership.