Verification API
Verify convenes a panel of independent reasoners over a symbol, measures their agreement, surfaces explicit contradictions, and returns an evidence-weighted consensus whose confidence falls when the panel disagrees.
What it returns
The response returns consensus (stance, agreement 0 to 100, confidence, weighted_vote), a contradictions array of explicit directional conflicts, the full reasoners panel (each with reasoner, provider, stance, confidence, rationale), a models_live flag, data_mode, and the trust envelope.
Stances are constructive, cautious, or neutral about what the data shows. Agreement is 100 when the panel votes together and falls toward 0 as it splits.
When to use it
Use Verify when you want the strength of a read, not just the read: how much an independent panel agrees and where it conflicts. It feeds the Readiness gate's contradiction factor directly, so calling Readiness already runs it, but Verify exposes the raw panel on its own.
When there is no data it returns available:false with a neutral consensus and an empty panel.
Notes and honesty
Offline, the reasoners are deterministic methodology lenses derived from the Market Brain (trend, mean-reversion, participation, regime), each labeled by its methodology. Live foundation-model reasoners plug into the identical contract when their keys are configured, and the tool never labels a deterministic lens as a model it isn't. The panel judges the read, never buy or sell.
- consensus: stance, agreement, confidence, weighted_vote
- contradictions: explicit constructive-vs-cautious conflicts
- reasoners: the full panel with per-reasoner stance and rationale
- models_live: true only when live model reasoners joined
- data_mode: live or sample
- trust envelope with evidence-weighted confidence
curl -s 'https://api.lyramindos.com/v1/verify?symbol=MSFT' -H 'x-api-key: YOUR_KEY'
Questions
Are the reasoners real AI models?
Offline they are deterministic methodology lenses (trend, mean-reversion, participation, regime), each honestly labeled. Live foundation-model reasoners plug into the same contract when configured, and models_live tells you which case applies.
Why does confidence drop when reasoners disagree?
That is the point. Consensus confidence is the panel's mean conviction discounted by its agreement, so a split panel yields lower confidence, which is what an agent should act on.
How does this relate to Readiness?
Verify feeds the Readiness gate's contradiction factor. Readiness runs it internally over the same read; Verify exposes the panel and contradictions on their own.