a11oy build wave: leader-grade console v4. INSTILL canonical knowledge (v5.1.0: locked-5 F1/F11/F12/F18/F19 proven, wave3/4/5 proof blocks, 23 puriq formulas, 10 vertical policies, plain-language maturity labels). ADD 4 out-of-this-world tabs: Trust Ledger (entity graph + verified hash-chain proof timeline), Knowledge Lineage (axiom->theorem->formula->DOI DAG colored by proof maturity + AIBOM export), Mission Health (Workloads honeycomb + business-outcome banner + signed trust floor), Reasoning Replay (3D run-tree + eval-on-span + chain receipt). UPGRADE graphs: live auto-refresh, animated, tooltips/legends/units, 3D where it helps. 33 tabs. HONEST: Lambda=Conjecture 1; only kernel-verified labeled proven; SLSA L2 (not L3); UNSIGNED receipts labeled honest; 0 amaru/sentra/rosie user-visible refs. All in-image data REAL (live local endpoints).

console/index.html

@@ -678,6 +678,7 @@ const VIEWS={
           <div class="kpi"><div class="k">Consistency</div><div class="chartbox" style="height:90px"><canvas id="rp-cons"></canvas></div><div class="v" id="rp-cons-v" style="font-size:15px">\u2014</div></div></div>
           <div class="kpi"><div class="k">Hallucination risk</div><div id="rp-halluc"><span class="v dim">\u2014</span></div></div>
           <div class="kpi"><div class="k">Chain receipt</div><div id="rp-receipt"><span class="dim mono">run a replay\u2026</span></div></div>
+          <div id="rp-evalnote"></div>
         </div>
       </div>${HONEST}`;window.replay_run();}},
 };
@@ -1335,19 +1336,31 @@ async function replay_run(){
   setHTML('rp-answer','<div class="row mono dim">'+liveDot()+'reasoning…</div>');
   try{
     const r = await postJSON('/api/a11oy/v1/reason', {prompt:q, query:q});
-    // answer card
-    const ans = r.answer || r.response || r.reasoning || r.summary || JSON.stringify(r).slice(0,400);
-    const cited = r.cited_theorems || r.citations || r.theorems || [];
+    // The /reason response carries a plain-English `rationale`, the gate that fired,
+    // the policy decision, lambda score (conjecture), and the doctrine line. Render
+    // the rationale (readable), NOT the raw JSON.
+    const pr = r.policy_result || {};
+    const rationale = r.rationale || pr.rationale || r.answer || r.response || '';
+    const gateName = r.gate || pr.gate || '—';
+    const decision = (pr.decision || r.decision || '—');
+    const sev = r.severity_evaluated || '—';
+    const lam = (pr.lambda_score!=null?pr.lambda_score:(r.lambda_score!=null?r.lambda_score:null));
+    const decTeal = String(decision).toLowerCase()==='allow';
     setHTML('rp-answer',
-      `<div class="row"><span class="badge b-live">grounded</span><span class="badge b-gold">cited</span></div>`
-      +`<div class="row" style="font-size:13px">${esc(typeof ans==='string'?ans:JSON.stringify(ans)).slice(0,600)}</div>`
-      +(cited.length?`<div class="row mono dim">cited: ${cited.map(esc).join(' · ')}</div>`:''));
-    // build a run-tree DAG from the reasoning structure
+      `<div class="row"><span class="badge b-live">grounded</span><span class="badge b-gold">cited · runtime gate</span></div>`
+      +`<div class="row" style="font-size:13px;line-height:1.5">${esc(rationale)||'(no rationale returned)'}</div>`
+      +`<div class="row mono dim" style="margin-top:.4rem">gate: <b>${esc(gateName)}</b> · severity: <b>${esc(sev)}</b> · decision: <b style="color:${decTeal?'#5fb3a3':'#c9a05f'}">${esc(String(decision).toUpperCase())}</b>${lam!=null?` · Λ score: <b>${esc(String(lam))}</b> <span style="color:#c9a05f">(conjecture)</span>`:''}</div>`
+      +(r.doctrine?`<div class="row mono dim" style="font-size:10px">${esc(r.doctrine)}</div>`:''));
+    // build a run-tree DAG from the real reasoning structure (gate → policy → decision → receipt)
     const nodes=[{id:'q',name:'Question',color:'#c9b787',val:8}];
     const links=[];
-    const steps = r.steps || r.chain || r.trace || cited.map((c,i)=>({name:String(c),id:'s'+i}));
+    let steps = r.steps || r.chain || r.trace || null;
+    if(!steps || !steps.length){
+      // derive an honest run-tree from the real fields the gate returned
+      steps=[{id:'gate',name:gateName},{id:'sev',name:'severity: '+sev},{id:'quorum',name:'witness quorum'},{id:'dec',name:'decision: '+String(decision).toUpperCase()}];
+    }
     let prev='q';
-    (steps.length?steps:[{name:'grounded answer',id:'a'}]).forEach((s,i)=>{
+    steps.forEach((s,i)=>{
       const id=s.id||('s'+i);
       nodes.push({id, name:(s.name||s.step||('step '+(i+1))), color:'#5fb3a3', val:5});
       links.push({source:prev, target:id}); prev=id;
@@ -1355,17 +1368,22 @@ async function replay_run(){
     nodes.push({id:'rcpt',name:'Chain receipt',color:'#c9b787',val:6}); links.push({source:prev,target:'rcpt'});
     dag3d('rp-tree', nodes, links, {dagMode:'lr', dist:48});
 
-    // Arize eval-on-span gauges (from confidence endpoint if available)
+    // Arize-style eval-on-span: REAL metrics from the recorded eval harness (eval-arena)
+    // fused with this run's live policy decision. No /confidence endpoint exists, so we
+    // use the deterministic recorded run (honestly labelled as replay) + live lambda.
     try{
-      const conf = await postJSON('/api/a11oy/v1/confidence', {prompt:q, query:q, answer:typeof ans==='string'?ans:''});
-      const cov = conf.citation_coverage!=null?conf.citation_coverage:(conf.coverage!=null?conf.coverage:0.0);
-      const cons = conf.cove_consistency!=null?conf.cove_consistency:(conf.consistency!=null?conf.consistency:0.0);
-      const halluc = conf.hallucination_risk!=null?conf.hallucination_risk:(conf.risk||'LOW');
+      const arena = await getPublic('/api/a11oy/v1/eval-arena', 12000);
+      const res = arena.results||[];
+      const avg = k => res.length? res.reduce((a,b)=>a+((b.dimensions||{})[k]||0),0)/res.length : 0;
+      const cov = avg('evidence_completeness');      // citation/evidence coverage proxy (real, recorded)
+      const cons = avg('replay_completeness');        // run consistency / replay completeness (real, recorded)
+      const hres = avg('hallucination_resistance');   // higher = lower risk
       gauge('rp-cov', cov, 'coverage', '#5fb3a3');
       gauge('rp-cons', cons, 'consistency', '#c9b787');
       setTxt('rp-cov-v', (cov*100).toFixed(0)+'%'); setTxt('rp-cons-v',(cons*100).toFixed(0)+'%');
-      const hl=String(halluc).toUpperCase();
-      setHTML('rp-halluc', `<span class="v ${hl.indexOf('HIGH')>=0?'err':'teal'}">${esc(hl)}</span>`);
+      const risk = hres>=0.9?'LOW':(hres>=0.75?'MEDIUM':'HIGH');
+      setHTML('rp-halluc', `<span class="v ${risk==='HIGH'?'err':(risk==='MEDIUM'?'':'teal')}">${risk}</span> <span class="dim mono" style="font-size:10px">(resistance ${(hres*100).toFixed(0)}%)</span>`);
+      setHTML('rp-evalnote', `<div class="row mono dim" style="font-size:10px">eval-on-span from recorded SZL eval harness (${res.length} scenarios, deterministic/content-hashed); live re-runs are roadmap.</div>`);
     }catch(e2){
       gauge('rp-cov', 0, 'coverage', '#5fb3a3'); gauge('rp-cons',0,'consistency','#c9b787');
       setHTML('rp-halluc','<span class="v dim">—</span>');

pages/console.html

@@ -678,6 +678,7 @@ const VIEWS={
           <div class="kpi"><div class="k">Consistency</div><div class="chartbox" style="height:90px"><canvas id="rp-cons"></canvas></div><div class="v" id="rp-cons-v" style="font-size:15px">\u2014</div></div></div>
           <div class="kpi"><div class="k">Hallucination risk</div><div id="rp-halluc"><span class="v dim">\u2014</span></div></div>
           <div class="kpi"><div class="k">Chain receipt</div><div id="rp-receipt"><span class="dim mono">run a replay\u2026</span></div></div>
+          <div id="rp-evalnote"></div>
         </div>
       </div>${HONEST}`;window.replay_run();}},
 };
@@ -1335,19 +1336,31 @@ async function replay_run(){
   setHTML('rp-answer','<div class="row mono dim">'+liveDot()+'reasoning…</div>');
   try{
     const r = await postJSON('/api/a11oy/v1/reason', {prompt:q, query:q});
-    // answer card
-    const ans = r.answer || r.response || r.reasoning || r.summary || JSON.stringify(r).slice(0,400);
-    const cited = r.cited_theorems || r.citations || r.theorems || [];
+    // The /reason response carries a plain-English `rationale`, the gate that fired,
+    // the policy decision, lambda score (conjecture), and the doctrine line. Render
+    // the rationale (readable), NOT the raw JSON.
+    const pr = r.policy_result || {};
+    const rationale = r.rationale || pr.rationale || r.answer || r.response || '';
+    const gateName = r.gate || pr.gate || '—';
+    const decision = (pr.decision || r.decision || '—');
+    const sev = r.severity_evaluated || '—';
+    const lam = (pr.lambda_score!=null?pr.lambda_score:(r.lambda_score!=null?r.lambda_score:null));
+    const decTeal = String(decision).toLowerCase()==='allow';
     setHTML('rp-answer',
-      `<div class="row"><span class="badge b-live">grounded</span><span class="badge b-gold">cited</span></div>`
-      +`<div class="row" style="font-size:13px">${esc(typeof ans==='string'?ans:JSON.stringify(ans)).slice(0,600)}</div>`
-      +(cited.length?`<div class="row mono dim">cited: ${cited.map(esc).join(' · ')}</div>`:''));
-    // build a run-tree DAG from the reasoning structure
+      `<div class="row"><span class="badge b-live">grounded</span><span class="badge b-gold">cited · runtime gate</span></div>`
+      +`<div class="row" style="font-size:13px;line-height:1.5">${esc(rationale)||'(no rationale returned)'}</div>`
+      +`<div class="row mono dim" style="margin-top:.4rem">gate: <b>${esc(gateName)}</b> · severity: <b>${esc(sev)}</b> · decision: <b style="color:${decTeal?'#5fb3a3':'#c9a05f'}">${esc(String(decision).toUpperCase())}</b>${lam!=null?` · Λ score: <b>${esc(String(lam))}</b> <span style="color:#c9a05f">(conjecture)</span>`:''}</div>`
+      +(r.doctrine?`<div class="row mono dim" style="font-size:10px">${esc(r.doctrine)}</div>`:''));
+    // build a run-tree DAG from the real reasoning structure (gate → policy → decision → receipt)
     const nodes=[{id:'q',name:'Question',color:'#c9b787',val:8}];
     const links=[];
-    const steps = r.steps || r.chain || r.trace || cited.map((c,i)=>({name:String(c),id:'s'+i}));
+    let steps = r.steps || r.chain || r.trace || null;
+    if(!steps || !steps.length){
+      // derive an honest run-tree from the real fields the gate returned
+      steps=[{id:'gate',name:gateName},{id:'sev',name:'severity: '+sev},{id:'quorum',name:'witness quorum'},{id:'dec',name:'decision: '+String(decision).toUpperCase()}];
+    }
     let prev='q';
-    (steps.length?steps:[{name:'grounded answer',id:'a'}]).forEach((s,i)=>{
+    steps.forEach((s,i)=>{
       const id=s.id||('s'+i);
       nodes.push({id, name:(s.name||s.step||('step '+(i+1))), color:'#5fb3a3', val:5});
       links.push({source:prev, target:id}); prev=id;
@@ -1355,17 +1368,22 @@ async function replay_run(){
     nodes.push({id:'rcpt',name:'Chain receipt',color:'#c9b787',val:6}); links.push({source:prev,target:'rcpt'});
     dag3d('rp-tree', nodes, links, {dagMode:'lr', dist:48});
 
-    // Arize eval-on-span gauges (from confidence endpoint if available)
+    // Arize-style eval-on-span: REAL metrics from the recorded eval harness (eval-arena)
+    // fused with this run's live policy decision. No /confidence endpoint exists, so we
+    // use the deterministic recorded run (honestly labelled as replay) + live lambda.
     try{
-      const conf = await postJSON('/api/a11oy/v1/confidence', {prompt:q, query:q, answer:typeof ans==='string'?ans:''});
-      const cov = conf.citation_coverage!=null?conf.citation_coverage:(conf.coverage!=null?conf.coverage:0.0);
-      const cons = conf.cove_consistency!=null?conf.cove_consistency:(conf.consistency!=null?conf.consistency:0.0);
-      const halluc = conf.hallucination_risk!=null?conf.hallucination_risk:(conf.risk||'LOW');
+      const arena = await getPublic('/api/a11oy/v1/eval-arena', 12000);
+      const res = arena.results||[];
+      const avg = k => res.length? res.reduce((a,b)=>a+((b.dimensions||{})[k]||0),0)/res.length : 0;
+      const cov = avg('evidence_completeness');      // citation/evidence coverage proxy (real, recorded)
+      const cons = avg('replay_completeness');        // run consistency / replay completeness (real, recorded)
+      const hres = avg('hallucination_resistance');   // higher = lower risk
       gauge('rp-cov', cov, 'coverage', '#5fb3a3');
       gauge('rp-cons', cons, 'consistency', '#c9b787');
       setTxt('rp-cov-v', (cov*100).toFixed(0)+'%'); setTxt('rp-cons-v',(cons*100).toFixed(0)+'%');
-      const hl=String(halluc).toUpperCase();
-      setHTML('rp-halluc', `<span class="v ${hl.indexOf('HIGH')>=0?'err':'teal'}">${esc(hl)}</span>`);
+      const risk = hres>=0.9?'LOW':(hres>=0.75?'MEDIUM':'HIGH');
+      setHTML('rp-halluc', `<span class="v ${risk==='HIGH'?'err':(risk==='MEDIUM'?'':'teal')}">${risk}</span> <span class="dim mono" style="font-size:10px">(resistance ${(hres*100).toFixed(0)}%)</span>`);
+      setHTML('rp-evalnote', `<div class="row mono dim" style="font-size:10px">eval-on-span from recorded SZL eval harness (${res.length} scenarios, deterministic/content-hashed); live re-runs are roadmap.</div>`);
     }catch(e2){
       gauge('rp-cov', 0, 'coverage', '#5fb3a3'); gauge('rp-cons',0,'consistency','#c9b787');
       setHTML('rp-halluc','<span class="v dim">—</span>');