Make entity, goal, and audience first-class inputs

Three equal inputs define the SGO spec:
- Entity: what you're putting out there
- Goal: what outcome you want
- Audience: who evaluates it

Goal and audience are auto-inferred from entity via LLM if left blank,
but visible and editable so the user always sees and controls the spec.
Audience context now drives Nemotron demographic filtering directly.

- Add /api/infer-spec endpoint
- Promote audience from hidden Advanced Options to main form
- Auto-fill on Evaluate if blank, show in progress log

web/app.py

@@ -235,6 +235,52 @@ async def get_session(sid: str):
     }
 
 
+class InferSpecInput(BaseModel):
+    entity_text: str
+
+
+@app.post("/api/infer-spec")
+async def infer_spec(input: InferSpecInput):
+    """Infer goal and audience from entity text."""
+    client = get_client()
+    model = get_model()
+
+    prompt = f"""Read this entity and infer two things:
+1. What is the most likely GOAL the author has? (what outcome they want)
+2. Who is the intended AUDIENCE? (who evaluates or decides)
+
+Entity:
+{input.entity_text[:2000]}
+
+Return JSON:
+{{
+    "goal": "<1 sentence — the outcome they're optimizing for>",
+    "audience": "<1 sentence — who should evaluate this, with demographics if obvious>"
+}}
+
+Examples:
+- Product landing page → goal: "Convert visitors to paying customers", audience: "Software developers evaluating dev tools"
+- Resume → goal: "Get interview callbacks from target companies", audience: "Engineering hiring managers at mid-stage startups"
+- Profile → goal: "Attract compatible connections", audience: "Professionals aged 28-40 in the same metro area"
+- Pitch deck → goal: "Secure Series A funding", audience: "VCs and angels focused on B2B SaaS"
+
+Be specific to THIS entity, not generic."""
+
+    try:
+        resp = client.chat.completions.create(
+            model=model,
+            messages=[{"role": "user", "content": prompt}],
+            response_format={"type": "json_object"},
+            max_tokens=256,
+            temperature=0.5,
+        )
+        content = resp.choices[0].message.content
+        content = re.sub(r'<think>[\s\S]*?</think>', '', content).strip()
+        return json.loads(content)
+    except Exception as e:
+        raise HTTPException(500, f"Failed to infer spec: {e}")
+
+
 @app.post("/api/suggest-changes")
 async def suggest_changes(input: SuggestChangesInput):
     """Generate candidate changes from evaluation concerns and goal."""

web/static/index.html

@@ -345,21 +345,23 @@
     </div>
 
     <div class="field">
-      <textarea id="entityText" placeholder="Paste your entity here..."></textarea>
+      <label>Entity — what are you putting out there?</label>
+      <textarea id="entityText" placeholder="Paste your landing page, resume, pitch, profile, policy..."></textarea>
     </div>
 
     <div class="field">
-      <label>What's your goal?</label>
-      <input type="text" id="goalText" placeholder="e.g. 'Get hired at a Series B startup' or 'Close enterprise deals'">
+      <label>Goal — what outcome do you want?</label>
+      <input type="text" id="goalText" placeholder="Will be auto-suggested from your entity">
+    </div>
+
+    <div class="field">
+      <label>Audience — who are you trying to reach?</label>
+      <input type="text" id="cohortDesc" placeholder="Will be auto-suggested from your entity">
     </div>
 
     <details class="mb-8">
       <summary style="cursor:pointer;color:var(--text2);font-size:0.85rem">Advanced options</summary>
       <div style="padding:12px 0">
-        <div class="field">
-          <label>Audience context (optional — auto-detected if blank)</label>
-          <input type="text" id="cohortDesc" placeholder="e.g. 'Would customers buy this product?'">
-        </div>
         <div class="field">
           <label>Panel size</label>
           <input type="number" id="panelSize" value="30" min="5" max="80"
@@ -656,6 +658,30 @@ function logStep(msg, cls = '') {
 
 // ── Step 1: Full pipeline (one click) ──
 
+async function inferSpec() {
+  const text = document.getElementById('entityText').value.trim();
+  if (!text) return;
+
+  const goalField = document.getElementById('goalText');
+  const audienceField = document.getElementById('cohortDesc');
+
+  // Only auto-fill if both are empty
+  if (goalField.value.trim() && audienceField.value.trim()) return;
+
+  try {
+    const resp = await fetch('/api/infer-spec', {
+      method: 'POST',
+      headers: {'Content-Type': 'application/json'},
+      body: JSON.stringify({entity_text: text}),
+    });
+    const data = await resp.json();
+    if (!goalField.value.trim() && data.goal) goalField.value = data.goal;
+    if (!audienceField.value.trim() && data.audience) audienceField.value = data.audience;
+  } catch (e) {
+    // Silent fail — user can fill in manually
+  }
+}
+
 async function runFullPipeline() {
   const text = document.getElementById('entityText').value.trim();
   if (!text) return alert('Please enter your entity text.');
@@ -669,9 +695,20 @@ async function runFullPipeline() {
   document.getElementById('evalLog').innerHTML = '';
   document.getElementById('pipelineProgressBar').style.width = '5%';
 
+  // Auto-infer goal + audience if not provided
+  const goalField = document.getElementById('goalText');
+  const audienceField = document.getElementById('cohortDesc');
+  if (!goalField.value.trim() || !audienceField.value.trim()) {
+    document.getElementById('pipelineProgressText').textContent = 'Inferring goal and audience from entity...';
+    logStep('Inferring goal and audience from entity...');
+    await inferSpec();
+    if (goalField.value) logStep(`Goal: ${goalField.value}`, 'pos');
+    if (audienceField.value) logStep(`Audience: ${audienceField.value}`, 'pos');
+  }
+
   const biasCal = document.getElementById('biasCalibration').checked;
   const panelSize = parseInt(document.getElementById('panelSize').value) || 30;
-  const audienceCtx = document.getElementById('cohortDesc').value.trim();
+  const audienceCtx = audienceField.value.trim();
 
   try {
     // Phase 1: Create session