feat: Pattern Lab Discover + fix history run display

Fix: history runs now always show results — create synthetic selected preset
from run data when no matching preset found (was broken for custom events).
Also force mode='events' and reset matchResults on history load.

Discover tab: new "Discover" panel in left sidebar (AI knowledge search).
- GPT-4o generates 6 matching events from a free-text query (date, assets, hint)
- Confidence score + category badge per event
- Click → pre-fills experiment form exactly like a preset → ready to Run
- Backend: POST /api/pattern-lab/discover (DiscoverRequest, sorted by confidence)
- Frontend: useDiscoverEvents hook + DiscoveredEvent type + Discover UI with
  Enter-to-search, spinner, empty states

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/routers/pattern_lab.py

@@ -49,6 +49,11 @@ class FindMatchingRequest(BaseModel):
     pattern_index: int
 
 
+class DiscoverRequest(BaseModel):
+    query: str
+    n: int = 6
+
+
 class InstrumentScanRequest(BaseModel):
     ticker: str
     instrument_name: str
@@ -270,6 +275,57 @@ def evaluate_instrument_run(run_id: str):
     return {"run_id": run_id, "outcomes": outcomes}
 
 
+@router.post("/discover")
+async def discover_events(req: DiscoverRequest):
+    """Use GPT-4o to generate historical/recent event suggestions matching the query."""
+    from openai import AsyncOpenAI
+    key = get_config("openai_api_key") or ""
+    if not key:
+        raise HTTPException(400, "OpenAI API key not configured")
+    client = AsyncOpenAI(api_key=key)
+
+    prompt = f"""You are a financial markets research expert specialized in geopolitical and macro events.
+
+Query: "{req.query}"
+
+Identify {req.n} specific real events (2000-2025) that match this query and are relevant for options/derivatives trading analysis.
+Prioritize events with clear market impact and dateable start points.
+
+Return ONLY valid JSON:
+{{
+  "events": [
+    {{
+      "label": "Concise event name",
+      "date": "YYYY-MM-DD",
+      "category": "geopolitical|macro|volatility|commodities|fx_macro|credit",
+      "horizon_days": 60,
+      "assets": ["TICKER1", "TICKER2", "TICKER3"],
+      "hint": "2-3 sentence market context for a trader: what happened, why it matters, key dynamics.",
+      "confidence": 90
+    }}
+  ]
+}}
+
+Rules:
+- Use valid Yahoo Finance tickers (SPY, GLD, USO, EURUSD=X, BTC-USD, etc.)
+- horizon_days: typical holding period to capture the event's full market impact
+- confidence: 0-100, how well-documented and dateable this event is
+- Diversify dates and instruments across the {req.n} results"""
+
+    resp = await client.chat.completions.create(
+        model="gpt-4o",
+        messages=[{"role": "user", "content": prompt}],
+        response_format={"type": "json_object"},
+        temperature=0.3,
+        timeout=30,
+    )
+    result = json.loads(resp.choices[0].message.content)
+    events = result.get("events", [])
+    # Sort by confidence desc
+    events.sort(key=lambda e: e.get("confidence", 0), reverse=True)
+    return {"events": events, "source": "ai", "query": req.query}
+
+
 @router.post("/find-matching")
 async def find_matching_pattern(req: FindMatchingRequest):
     """Ask GPT-4o-mini if this lab pattern matches / counter-scenario an existing library pattern."""