Add web topology viewer with hot path visualization

- FastAPI backend (server.py) with REST API + WebSocket for event streaming
- SVG-based topology viewer (index.html) with SIP/CUBE/PE drill-down views
- Event logging infrastructure (event_log.py) generating events from real
  probe cases (H2D/D2H/PE-DMA) and bench workloads (QKV GEMM single/multi-PE)
- Timeline replay engine with play/pause, speed control, and scrubbing
- Workload selector dropdown grouped by category (Probe/Bench)
- CLI entry points: kernbench web, ./kernbench wrapper scripts

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

src/kernbench/web/server.py

@@ -0,0 +1,126 @@
+"""Phase B: FastAPI backend with WebSocket streaming for KernBench viewer."""
+
+import asyncio
+import json
+import pathlib
+import webbrowser
+import threading
+
+from fastapi import FastAPI, WebSocket, WebSocketDisconnect
+from fastapi.staticfiles import StaticFiles
+from fastapi.responses import FileResponse, JSONResponse
+import uvicorn
+
+STATIC_DIR = pathlib.Path(__file__).parent / "static"
+DEFAULT_PORT = 8765
+
+app = FastAPI(title="KernBench Viewer")
+
+
+# ── REST: workloads + events ───────────────────────────────────────
+@app.get("/api/workloads")
+async def list_workloads():
+    from kernbench.sim_engine.event_log import get_workload_list
+    return JSONResponse(content=get_workload_list())
+
+
+@app.get("/api/events/demo")
+async def get_demo_events():
+    from kernbench.sim_engine.event_log import generate_demo_events
+    return JSONResponse(content=generate_demo_events())
+
+
+@app.get("/api/events/{workload_id}")
+async def get_workload_events(workload_id: str):
+    from kernbench.sim_engine.event_log import generate_workload_events
+    try:
+        return JSONResponse(content=generate_workload_events(workload_id))
+    except ValueError as e:
+        return JSONResponse(content={"error": str(e)}, status_code=404)
+
+
+# ── WebSocket: stream events with timing ───────────────────────────
+@app.websocket("/ws")
+async def ws_endpoint(ws: WebSocket):
+    await ws.accept()
+    try:
+        while True:
+            msg = await ws.receive_text()
+            data = json.loads(msg)
+            cmd = data.get("cmd")
+
+            if cmd == "load_demo" or cmd == "load_workload":
+                from kernbench.sim_engine.event_log import generate_workload_events
+                wid = data.get("workload_id", "probe-h2d")
+                events = generate_workload_events(wid)
+                await ws.send_json({"type": "events", "events": events})
+
+            elif cmd == "stream_demo" or cmd == "stream_workload":
+                from kernbench.sim_engine.event_log import generate_workload_events
+                wid = data.get("workload_id", "probe-h2d")
+                events = generate_workload_events(wid)
+                speed = data.get("speed", 1.0)  # ns → real-time multiplier
+
+                # Send metadata first
+                duration = max(e["t_ns"] for e in events) if events else 0
+                await ws.send_json({
+                    "type": "stream_start",
+                    "total_events": len(events),
+                    "duration_ns": duration,
+                })
+
+                prev_t = 0.0
+                for ev in events:
+                    # Wait proportional to simulation time gap
+                    dt = ev["t_ns"] - prev_t
+                    if dt > 0 and speed > 0:
+                        # Scale: 1 ns sim → (1/speed) ms real
+                        await asyncio.sleep(dt / (speed * 1000.0))
+                    await ws.send_json({"type": "event", **ev})
+                    prev_t = ev["t_ns"]
+
+                await ws.send_json({"type": "stream_end"})
+
+            elif cmd == "ping":
+                await ws.send_json({"type": "pong"})
+
+    except WebSocketDisconnect:
+        pass
+    except Exception:
+        pass
+
+
+# ── Static files (must be last) ────────────────────────────────────
+@app.get("/")
+async def index():
+    return FileResponse(STATIC_DIR / "index.html")
+
+
+app.mount("/", StaticFiles(directory=str(STATIC_DIR)), name="static")
+
+
+# ── Entry point (called from CLI) ──────────────────────────────────
+def serve(port: int = DEFAULT_PORT, open_browser: bool = True) -> None:
+    import socket
+
+    # Check port availability before starting
+    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+        try:
+            s.bind(("127.0.0.1", port))
+        except OSError:
+            print(f"ERROR: Port {port} is already in use.")
+            print(f"  Try: kernbench web --port {port + 1}")
+            print(f"  Or kill the existing process using port {port}")
+            return
+
+    url = f"http://127.0.0.1:{port}"
+    print(f"KernBench Viewer: {url}")
+
+    if open_browser:
+        threading.Timer(0.5, webbrowser.open, args=(url,)).start()
+
+    uvicorn.run(app, host="127.0.0.1", port=port, log_level="warning")
+
+
+if __name__ == "__main__":
+    serve()