diff --git a/README.md b/README.md
index 45bacd1..8dca9cb 100644
--- a/README.md
+++ b/README.md
@@ -117,16 +117,30 @@ hit **Start**, and watch the AI think.
### Demo mode (no MT5 required)
-Don't have MT5 installed? Run the offline demo that feeds synthetic backtest results
-through the same AI loop and dashboard:
+Don't have MT5 installed? The offline demo feeds synthetic backtest results
+through the same AI loop and dashboard. **It auto-starts** — open the link
+the script prints, sit back, and watch APEX think:
```bash
python -m demo.run_demo
```
-This is the path to use if you're a hackathon judge — you'll see the full thinking feed,
-parameter‑change panel, validation phase, and verdict screen without needing a Windows
-machine with MT5.
+That's it. The browser opens to the live dashboard, an optimization kicks off
+automatically, and you'll see all three phases play out over ~3-4 minutes:
+exploration → AI iteration with reasoning streaming live → out-of-sample +
+sensitivity validation → final verdict. Every panel populates so you can see
+exactly what the system does.
+
+If you'd rather drive it yourself (configure your own EA, dates, targets), use:
+
+```bash
+python -m demo.run_demo --quick # boots the server, you click "New Run"
+python -m demo.run_demo --loop # auto-restart between runs (unattended recording)
+```
+
+This is the path to use if you're a hackathon judge — you'll see the full
+thinking feed, parameter-change panel, validation phase, and verdict screen
+without needing a Windows machine with MT5.
---
diff --git a/SUBMISSION.md b/SUBMISSION.md
index a8c9e4c..0caf38a 100644
--- a/SUBMISSION.md
+++ b/SUBMISSION.md
@@ -81,8 +81,9 @@ https://github.com/tonnylegacy/MT5_Optimizer
No hosted demo (the app runs locally to drive a local MT5 install). For judges:
- **Static**: open `screenshots/apex_demo.gif` in the repo (6‑frame timelapse of one autonomous run)
-- **Run it**: `git clone … && pip install -r requirements.txt && python -m demo.run_demo` → opens at `http://localhost:5000` with synthetic backtests, no MT5 required
-- **With API key**: set `ANTHROPIC_API_KEY` env var to see live Claude reasoning stream into the Thinking Feed
+- **Run it**: `git clone … && pip install -r requirements.txt && python -m demo.run_demo` → the browser opens to the dashboard, a ~3-4 minute optimization auto-starts, and every phase (exploration → AI iteration → validation → verdict) plays out without you touching anything
+- **With API key**: set `ANTHROPIC_API_KEY` env var (or fill `ai.anthropic_api_key` in `config.yaml`) to see live Claude reasoning stream into the Thinking Feed
+- **Faster / your-own-config**: `python -m demo.run_demo --quick` to skip the auto-run and drive the demo from `/setup` yourself
---
diff --git a/demo/run_demo.py b/demo/run_demo.py
index a32f29f..c34f503 100644
--- a/demo/run_demo.py
+++ b/demo/run_demo.py
@@ -7,16 +7,20 @@ generates synthetic backtest results instead of calling MT5. Lets judges
without a Windows + MT5 install see the full live AI loop, validation,
and verdict flow.
+By default, the demo **auto-runs**: it opens the dashboard in your browser
+and immediately starts a ~3-4 minute optimization showcasing every phase
+(exploration → AI iteration → validation → verdict). No manual setup needed.
+
Usage:
- python -m demo.run_demo
- # or
- python demo/run_demo.py
+ python -m demo.run_demo # auto-running showcase (default, ~3-4 min)
+ python -m demo.run_demo --quick # original fast/manual demo (you click "New Run")
+ python -m demo.run_demo --loop # auto-restart on completion (for unattended recording)
"""
from __future__ import annotations
+import argparse
import os
import sys
-import textwrap
from pathlib import Path
import yaml
@@ -83,24 +87,159 @@ def ensure_config() -> None:
print(f" [!] No config.yaml or config.example.yaml — app may fail to start")
-def banner() -> None:
+def banner(quick: bool = False, loop: bool = False) -> None:
bar = "=" * 72
- print(textwrap.dedent(f"""
- {bar}
- APEX -- DEMO MODE (offline / no MT5)
- {bar}
- * Backtests are synthetic (deterministic from params + jitter)
- * The AI loop, validation, and verdict flow are 100% real
- * Set ANTHROPIC_API_KEY to see live AI reasoning
+ if quick:
+ mode_line = "APEX -- DEMO MODE (quick / manual)"
+ body = (
+ " * Backtests are synthetic; AI reasoning is real if API key is set\n"
+ " * Per-run latency is short for fast UI testing\n\n"
+ ' Open http://localhost:5000 in your browser, hit "New Run", and\n'
+ " configure your own optimization."
+ )
+ else:
+ suffix = " [LOOP]" if loop else ""
+ mode_line = f"APEX -- DEMO MODE (auto-running showcase){suffix}"
+ body = (
+ " * Auto-starts a ~3-4 min optimization showing every phase\n"
+ " * AI reasoning streams live (API key loaded from env or config.yaml)\n"
+ " * The dashboard opens itself; just sit back or hit your recorder hotkey\n"
+ " * Heads up: AI API calls add ~12-15s each — actual run can stretch\n"
+ " to ~5 min depending on Claude latency.\n"
+ " * To skip the auto-start and configure your own run: --quick"
+ )
+ print(f"\n{bar}\n{mode_line.center(72)}\n{bar}\n{body}\n{bar}\n")
- Open http://localhost:5000 in your browser, hit "New Run", and
- watch the AI think.
- {bar}
- """))
+
+# ── Showcase auto-start config ──────────────────────────────────────────────
+# Tuned for ~2.5-3 min runtime that actually shows the AI loop iterating:
+# * Phase 1 (10 LHS samples × ~2.5s, no AI) ≈ 25-30s
+# * Phase 2 (5 AI iterations × ~15-18s incl. analyze + suggest_next_params)
+# ≈ 75-90s — targets are deliberately STIFF so the loop can't early-exit
+# at iteration 0; Phase 1's best typically lands around PF≈1.9 / Calmar≈1
+# which is below these targets, so the AI gets to actually do its job.
+# * Phase 3 (1 OOS + 3 sensitivity, each with AI analyze) ≈ 50-60s
+# AI call latency dominates the math; the synthetic backtest delay is short.
+CINEMATIC_PAYLOAD = {
+ "ea_name": "APEX_DEMO_EA",
+ "symbol": "XAUUSD",
+ "timeframe": "H1",
+ "train_start": "2022.01.01",
+ "train_end": "2023.12.31",
+ "val_start": "2024.01.01",
+ "val_end": "2024.06.30",
+ "objective": "balanced",
+ "budget_minutes": 15,
+ "autonomous_mode": True,
+ "autonomous_max_iterations": 5,
+ "target_profit_factor": 2.5,
+ "target_max_drawdown_pct": 5.0,
+ "target_min_calmar": 1.5,
+ "selected_params": [],
+}
+
+
+def _autostart_cinematic_run(loop: bool = False) -> None:
+ """
+ Background thread: waits for the server to be up, then POSTs /api/start
+ with cinematic settings. If --loop, polls /api/status and re-triggers
+ when the pipeline goes idle (so an unattended recording keeps producing
+ fresh footage).
+ """
+ import json
+ import time
+ import urllib.error
+ import urllib.request
+
+ base = "http://127.0.0.1:5000"
+
+ def _server_ready() -> bool:
+ try:
+ with urllib.request.urlopen(f"{base}/api/status", timeout=2) as r:
+ return r.status == 200
+ except Exception:
+ return False
+
+ def _is_running() -> bool:
+ try:
+ with urllib.request.urlopen(f"{base}/api/status", timeout=2) as r:
+ data = json.loads(r.read())
+ state = (data.get("state") or "").lower()
+ return state in ("running", "starting")
+ except Exception:
+ return False
+
+ def _post_start() -> bool:
+ body = json.dumps(CINEMATIC_PAYLOAD).encode("utf-8")
+ req = urllib.request.Request(
+ f"{base}/api/start", data=body,
+ headers={"Content-Type": "application/json"}, method="POST",
+ )
+ try:
+ with urllib.request.urlopen(req, timeout=5) as r:
+ resp = json.loads(r.read())
+ return bool(resp.get("ok"))
+ except Exception as e:
+ print(f" [cinematic] /api/start failed: {e}")
+ return False
+
+ # Wait for server to come up (~10 seconds max)
+ for _ in range(40):
+ if _server_ready():
+ break
+ time.sleep(0.25)
+ else:
+ print(" [cinematic] server didn't become ready — aborting auto-start")
+ return
+
+ # Small grace period so the dashboard tab finishes connecting via SocketIO
+ time.sleep(2.0)
+
+ while True:
+ print(" [cinematic] starting optimization run …")
+ if not _post_start():
+ print(" [cinematic] failed to start — retrying in 10s")
+ time.sleep(10)
+ continue
+
+ # Wait for run to complete (poll until idle for 3 consecutive checks)
+ idle_streak = 0
+ while idle_streak < 3:
+ time.sleep(4)
+ if _is_running():
+ idle_streak = 0
+ else:
+ idle_streak += 1
+
+ if not loop:
+ print(" [cinematic] run complete — staying on verdict screen")
+ return
+ print(" [cinematic] run complete — restarting in 12s for next loop iteration")
+ time.sleep(12)
def main() -> int:
- banner()
+ parser = argparse.ArgumentParser(
+ prog="demo.run_demo",
+ description="APEX offline demo runner.",
+ )
+ parser.add_argument(
+ "--quick", action="store_true",
+ help="Skip the auto-start showcase. Server boots with fast per-run latency "
+ "and you drive the demo manually from /setup.",
+ )
+ parser.add_argument(
+ "--loop", action="store_true",
+ help="Auto-restart a fresh showcase run when each one completes (for unattended "
+ "screen recording). Ignored with --quick.",
+ )
+ parser.add_argument(
+ "--per-run-seconds", type=float, default=None,
+ help="Override APEX_DEMO_RUN_SECONDS (default: 5.0 showcase / 1.2 --quick).",
+ )
+ args = parser.parse_args()
+
+ banner(quick=args.quick, loop=args.loop)
print("Bootstrapping demo environment...")
ensure_config()
@@ -108,14 +247,35 @@ def main() -> int:
# Set the demo flag — the pipeline checks this in _execute_run.
os.environ["APEX_DEMO_MODE"] = "1"
- # Per-run latency tunable — keep small so demo feels snappy.
- os.environ.setdefault("APEX_DEMO_RUN_SECONDS", "1.2")
+ # Per-run latency tunable. Default (auto-running showcase) is slower so
+ # each phase is visible long enough to film and narrate over.
+ if args.per_run_seconds is not None:
+ os.environ["APEX_DEMO_RUN_SECONDS"] = str(args.per_run_seconds)
+ elif args.quick:
+ os.environ.setdefault("APEX_DEMO_RUN_SECONDS", "1.2")
+ else:
+ # 2.5s synthetic delay; AI call latency dominates total runtime anyway
+ os.environ.setdefault("APEX_DEMO_RUN_SECONDS", "2.5")
+ # Skip per-run AI analysis during Phase 1 exploration so the demo
+ # finishes in ~3-4 min. Phase 2's autonomous AI loop (the headline
+ # feature) still streams reasoning live.
+ os.environ.setdefault("APEX_DEMO_SKIP_PHASE1_AI", "1")
- if "ANTHROPIC_API_KEY" not in os.environ:
- print(" [!] ANTHROPIC_API_KEY not set — AI reasoning will be skipped (synthetic metrics still flow).")
+ # AI reasoner pulls from env var OR config.yaml ai.anthropic_api_key
+ cfg_has_key = False
+ try:
+ cfg = yaml.safe_load(CONFIG.read_text()) if CONFIG.exists() else {}
+ cfg_has_key = bool(((cfg or {}).get("ai") or {}).get("anthropic_api_key", "").strip())
+ except Exception:
+ pass
+ if not (os.environ.get("ANTHROPIC_API_KEY", "").strip() or cfg_has_key):
+ print(" [!] No API key in env (ANTHROPIC_API_KEY) or config.yaml — AI reasoning will be skipped.")
+ else:
+ src = "config.yaml" if cfg_has_key else "env var"
+ print(f" [ok] API key found in {src} — AI reasoning enabled.")
print()
- print("Launching APEX server at http://localhost:5000 ...")
+ print(f"Launching APEX server at http://localhost:5000 (per-run: {os.environ['APEX_DEMO_RUN_SECONDS']}s) ...")
sys.path.insert(0, str(ROOT))
# Import after env vars are set so the pipeline picks them up.
import threading
@@ -126,11 +286,22 @@ def main() -> int:
import time as _t
_t.sleep(1.5)
try:
- webbrowser.open("http://localhost:5000")
+ # In auto-running mode jump straight to the dashboard so the user sees
+ # the live run unfold; in --quick mode land on the landing page.
+ url = "http://localhost:5000" if args.quick else "http://localhost:5000/dashboard"
+ webbrowser.open(url)
except Exception:
pass
threading.Thread(target=_open_browser, daemon=True).start()
+
+ if not args.quick:
+ threading.Thread(
+ target=_autostart_cinematic_run,
+ kwargs={"loop": args.loop},
+ daemon=True,
+ ).start()
+
socketio.run(
flask_app, host="0.0.0.0", port=5000,
debug=False, use_reloader=False, allow_unsafe_werkzeug=True,
diff --git a/optimizer/pipeline.py b/optimizer/pipeline.py
index f8d5d7e..f645a7b 100644
--- a/optimizer/pipeline.py
+++ b/optimizer/pipeline.py
@@ -898,11 +898,17 @@ class OptimizationPipeline:
delay = float(os.environ.get("APEX_DEMO_RUN_SECONDS", "1.5"))
time.sleep(max(0.05, delay))
- # Run AI reasoning if enabled — same flow as live mode
- try:
- self._reason_about_run(run_id, metrics, [], params)
- except Exception:
- pass
+ # Run AI reasoning if enabled — same flow as live mode.
+ # In showcase mode (APEX_DEMO_SKIP_PHASE1_AI=1) we skip Phase 1's
+ # per-run AI analysis to keep total runtime to ~3-4 min. The AI loop
+ # in Phase 2 — where reasoning actually drives parameter changes — is
+ # untouched, so the headline feature is still visible.
+ skip_p1_ai = os.environ.get("APEX_DEMO_SKIP_PHASE1_AI", "").strip() in ("1", "true", "yes")
+ if not (skip_p1_ai and phase.startswith("phase1")):
+ try:
+ self._reason_about_run(run_id, metrics, [], params)
+ except Exception:
+ pass
return ranker.make_result(run_id, params, phase, metrics)
diff --git a/ui/static/img/apex-logo-wide.png b/ui/static/img/apex-logo-wide.png
new file mode 100644
index 0000000..c9a452f
Binary files /dev/null and b/ui/static/img/apex-logo-wide.png differ
diff --git a/ui/static/img/apex-logo.png b/ui/static/img/apex-logo.png
new file mode 100644
index 0000000..9ef3749
Binary files /dev/null and b/ui/static/img/apex-logo.png differ
diff --git a/ui/static/img/apex-mark.png b/ui/static/img/apex-mark.png
new file mode 100644
index 0000000..84aefe8
Binary files /dev/null and b/ui/static/img/apex-mark.png differ
diff --git a/ui/static/img/favicon.png b/ui/static/img/favicon.png
new file mode 100644
index 0000000..2d6ff6c
Binary files /dev/null and b/ui/static/img/favicon.png differ
diff --git a/ui/templates/dashboard.html b/ui/templates/dashboard.html
index 277867c..af713a8 100644
--- a/ui/templates/dashboard.html
+++ b/ui/templates/dashboard.html
@@ -4,6 +4,7 @@
APEX — AI-Powered EA Optimizer
+
@@ -16,16 +17,19 @@
--card-hover: #131d35;
--border: rgba(255,255,255,0.07);
--border-hover: rgba(255,255,255,0.14);
- --accent: #4f46e5;
- --accent2: #7c6dfa;
+ --brand-cyan: #00d4ff;
+ --brand-blue: #4f6cff;
+ --brand-purple: #9333ea;
+ --accent: #4f6cff;
+ --accent2: #8b5cf6;
--teal: #00d4aa;
--text: #e2e8f0;
--muted: #64748b;
--green: #10b981;
--red: #ef4444;
--yellow: #f59e0b;
- --nav-active-bg: rgba(79,70,229,0.18);
- --nav-active-border: #4f46e5;
+ --nav-active-bg: rgba(79,108,255,0.18);
+ --nav-active-border: #4f6cff;
}
/* ── Reset ──────────────────────────────────────────────────── */
@@ -70,11 +74,17 @@
border-bottom: 1px solid var(--border);
margin-bottom: 0.75rem;
}
+ .logo-mark,
+ .cmd-mark {
+ display: block;
+ flex-shrink: 0;
+ filter: drop-shadow(0 0 10px rgba(79,108,255,0.3));
+ }
.logo-text {
font-size: 1.25rem;
font-weight: 800;
letter-spacing: 0.12em;
- background: linear-gradient(135deg, #7c6dfa, #00d4aa);
+ background: linear-gradient(180deg, #ffffff 0%, #d4dae6 100%);
-webkit-background-clip: text;
-webkit-text-fill-color: transparent;
background-clip: text;
@@ -82,11 +92,12 @@
.logo-tagline {
font-size: 0.58rem;
color: var(--muted);
- font-weight: 400;
- letter-spacing: 0.06em;
+ font-weight: 600;
+ letter-spacing: 0.08em;
+ text-transform: uppercase;
display: block;
line-height: 1;
- margin-top: -2px;
+ margin-top: 2px;
-webkit-text-fill-color: var(--muted);
}
@@ -217,7 +228,7 @@
content: '';
position: absolute;
top: 0; left: 0; right: 0; height: 1px;
- background: linear-gradient(90deg, rgba(0,229,255,0.45), rgba(79,70,229,0.55), rgba(124,58,237,0.45));
+ background: linear-gradient(90deg, rgba(0,212,255,0.55), rgba(79,108,255,0.65), rgba(147,51,234,0.55));
}
.cmd-logo {
display: flex;
@@ -234,19 +245,22 @@
font-size: 1.05rem;
font-weight: 800;
letter-spacing: 0.14em;
- background: linear-gradient(135deg, #00e5ff 0%, #4f46e5 50%, #7c3aed 100%);
+ background: linear-gradient(180deg, #ffffff 0%, #d4dae6 100%);
-webkit-background-clip: text;
-webkit-text-fill-color: transparent;
background-clip: text;
}
.cmd-tagline {
- font-size: 0.46rem;
- font-weight: 600;
- letter-spacing: 0.1em;
- color: var(--muted);
+ font-size: 0.5rem;
+ font-weight: 700;
+ letter-spacing: 0.18em;
text-transform: uppercase;
- -webkit-text-fill-color: var(--muted);
+ display: inline-flex;
+ gap: 0.35em;
}
+ .cmd-tagline .tag-cyan { color: #00d4ff; }
+ .cmd-tagline .tag-purple { color: #b06bff; }
+ .cmd-tagline .tag-white { color: #cfd5e0; }
.cmd-sep {
width: 1px;
height: 30px;
@@ -310,7 +324,7 @@
}
.cmd-prog-fill {
height: 100%;
- background: linear-gradient(90deg, #00e5ff, #4f46e5);
+ background: linear-gradient(90deg, #00d4ff, #4f6cff, #9333ea);
border-radius: 999px;
transition: width 0.5s ease;
}
@@ -347,8 +361,9 @@
white-space: nowrap;
}
.cmd-btn-start {
- background: linear-gradient(135deg, #4f46e5, #7c6dfa);
+ background: linear-gradient(135deg, #4f6cff 0%, #9333ea 100%);
color: white;
+ box-shadow: 0 0 0 1px rgba(147,51,234,0.4), 0 4px 14px rgba(79,108,255,0.25);
}
.cmd-btn-start:hover { opacity: 0.88; }
.cmd-btn-pause {
@@ -399,7 +414,7 @@
position: absolute;
top: 0; left: 0; right: 0;
height: 2px;
- background: linear-gradient(90deg, var(--accent), var(--accent2));
+ background: linear-gradient(90deg, #00d4ff, #4f6cff, #9333ea);
opacity: 0;
transition: opacity 0.2s;
}
@@ -1629,27 +1644,10 @@
-
+
APEX
- AI Optimizer
+ AI Powered Optimizer
@@ -1728,27 +1726,15 @@
-
+
APEX
- AI Powered EA Optimizer
+
+ AI
+ POWERED
+ EA
+ OPTIMIZER
+
diff --git a/ui/templates/index.html b/ui/templates/index.html
index 85f3426..ee7bd41 100644
--- a/ui/templates/index.html
+++ b/ui/templates/index.html
@@ -3,7 +3,8 @@
- MT5 EA Optimizer — Live Dashboard
+ APEX — Live Dashboard
+
diff --git a/ui/templates/landing.html b/ui/templates/landing.html
index d9a7255..85db958 100644
--- a/ui/templates/landing.html
+++ b/ui/templates/landing.html
@@ -4,12 +4,16 @@
APEX — AI-Powered EA Optimizer
+
