| """DaisyChain dashboard (CLI: `daisychain-dashboard`). Tailwind-styled page with |
| a readiness banner, P2P connectivity scan, pooled resource + capacity plan, and |
| live training status. Config via DAISY_NODES_FILE. Serves on :8080.""" |
| import json |
| import os |
| from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer |
|
|
| try: |
| from daisychain.dashboard.scanner import scan_cluster |
| except Exception: |
| from scanner import scan_cluster |
|
|
| NODES_FILE = os.environ.get("DAISY_NODES_FILE", "config/nodes.example.json") |
| PORT = int(os.environ.get("DAISY_DASH_PORT", "8080")) |
| EXPECTED_WORLD = os.environ.get("DAISY_EXPECTED_WORLD") |
| BASE_BATCH = int(os.environ.get("DAISY_BASE_BATCH", "32")) |
|
|
|
|
| def load_nodes(): |
| with open(NODES_FILE) as f: |
| return json.load(f) |
|
|
|
|
| def _chip(ok, yes="OK", no="DOWN"): |
| cls = ("bg-emerald-500/20 text-emerald-600 dark:text-emerald-400" if ok |
| else "bg-rose-500/20 text-rose-600 dark:text-rose-400") |
| return f'<span class="px-2 py-0.5 rounded text-xs font-semibold {cls}">{yes if ok else no}</span>' |
|
|
|
|
| def render(d): |
| ready = d["ready"] |
| banner = ("bg-emerald-500", "✓ CLUSTER READY — all nodes connected") if ready \ |
| else ("bg-rose-500", f"✗ NOT READY — {d['reachable']}/{d['total']} nodes reachable") |
| rows = "" |
| for n in d["nodes"]: |
| lat = f'{n["latency_ms"]} ms' if n["latency_ms"] is not None else "—" |
| res = n.get("resources") or {} |
| dev = res.get("device", "—") |
| gpu = (res.get("gpu") or {}).get("name", "") |
| devlabel = f"{dev}" + (f" ({gpu})" if gpu else "") |
| rows += f"""<tr class="border-b border-slate-100 dark:border-slate-800"> |
| <td class="py-2 px-3 font-medium">{n['name']}</td> |
| <td class="py-2 px-3 text-slate-500">{n['host']}</td> |
| <td class="py-2 px-3">{_chip(n['reachable'],'reachable','unreachable')}</td> |
| <td class="py-2 px-3">{devlabel}</td> |
| <td class="py-2 px-3 tabular-nums">{lat}</td></tr>""" |
| plan = d["plan"] |
| pr = "" |
| for p in plan["per_node"]: |
| dv = p["device"] + (f" ({p['gpu']})" if p.get("gpu") else "") |
| ram = f'{p["ram_gb"]} GB' if p.get("ram_gb") is not None else "—" |
| pr += f"""<tr class="border-b border-slate-100 dark:border-slate-800"> |
| <td class="py-2 px-3 font-medium">{p['name']}</td> |
| <td class="py-2 px-3">{dv}</td> |
| <td class="py-2 px-3 tabular-nums">{ram}</td> |
| <td class="py-2 px-3 tabular-nums">{p['capacity']}</td> |
| <td class="py-2 px-3 tabular-nums">{p['weight']}</td> |
| <td class="py-2 px-3 tabular-nums">{p['batch']}</td></tr>""" |
| tot_ram = f'{plan["total_ram_gb"]} GB' if plan["total_ram_gb"] is not None else "—" |
| t = d["training"] |
| if t: |
| step, total = t.get("step", 0), t.get("total_steps", 1) |
| pct = int(100 * (step + 1) / max(1, total)); loss = t.get("cluster_avg_loss") |
| badge = _chip(True, "DONE") if t.get("done") else '<span class="text-xs text-sky-500 animate-pulse">training…</span>' |
| train = f"""<div class="flex items-center justify-between mb-2"> |
| <span class="text-sm text-slate-500">step {step} / {total}</span>{badge}</div> |
| <div class="w-full h-3 rounded-full bg-slate-200 dark:bg-slate-700 overflow-hidden"> |
| <div class="h-3 bg-sky-500" style="width:{pct}%"></div></div> |
| <p class="mt-3 text-2xl font-semibold tabular-nums">{loss:.5f} |
| <span class="text-sm font-normal text-slate-500">cluster-avg loss</span></p>""" |
| else: |
| train = '<p class="text-slate-400 text-sm">No active run detected (waiting for rank 0 status)…</p>' |
| return f"""<!doctype html><html><head><meta charset="utf-8"> |
| <meta name="viewport" content="width=device-width, initial-scale=1"> |
| <meta http-equiv="refresh" content="3"><title>DaisyChain — Cluster</title> |
| <script src="https://cdn.tailwindcss.com"></script></head> |
| <body class="bg-slate-50 dark:bg-slate-900 text-slate-800 dark:text-slate-100 min-h-screen"> |
| <div class="max-w-5xl mx-auto p-6 space-y-6"> |
| <header class="flex items-center justify-between"> |
| <div><h1 class="text-2xl font-bold">\U0001f33c DaisyChain</h1> |
| <p class="text-sm text-slate-500">Old Hardware Training Pipeline · scanned {d['scanned_at']}</p></div> |
| <span class="text-xs text-slate-400">auto-refresh 3s</span></header> |
| <div class="{banner[0]} text-white rounded-xl px-5 py-4 font-semibold text-lg shadow">{banner[1]}</div> |
| <div class="grid md:grid-cols-3 gap-4"> |
| <div class="rounded-xl border border-slate-200 dark:border-slate-700 p-4"><p class="text-xs uppercase tracking-wide text-slate-400">Nodes</p><p class="text-3xl font-bold tabular-nums">{d['reachable']}/{d['total']}</p></div> |
| <div class="rounded-xl border border-slate-200 dark:border-slate-700 p-4"><p class="text-xs uppercase tracking-wide text-slate-400">Total cores</p><p class="text-3xl font-bold tabular-nums">{plan['total_cores']}</p></div> |
| <div class="rounded-xl border border-slate-200 dark:border-slate-700 p-4"><p class="text-xs uppercase tracking-wide text-slate-400">Total RAM</p><p class="text-3xl font-bold tabular-nums">{tot_ram}</p></div></div> |
| <section class="rounded-xl border border-slate-200 dark:border-slate-700 overflow-hidden"> |
| <h2 class="px-4 py-3 font-semibold border-b border-slate-100 dark:border-slate-800">Connectivity scan</h2> |
| <table class="w-full text-sm"><thead class="text-left text-slate-400"><tr> |
| <th class="py-2 px-3">Node</th><th class="py-2 px-3">Host</th><th class="py-2 px-3">Agent</th> |
| <th class="py-2 px-3">Device</th><th class="py-2 px-3">Latency</th></tr></thead><tbody>{rows}</tbody></table></section> |
| <div class="grid md:grid-cols-2 gap-6"> |
| <section class="rounded-xl border border-slate-200 dark:border-slate-700 overflow-hidden"> |
| <h2 class="px-4 py-3 font-semibold border-b border-slate-100 dark:border-slate-800">Capacity plan · global batch {plan['global_batch']}</h2> |
| <table class="w-full text-sm"><thead class="text-left text-slate-400"><tr> |
| <th class="py-2 px-3">Node</th><th class="py-2 px-3">Device</th><th class="py-2 px-3">RAM</th> |
| <th class="py-2 px-3">Capacity</th><th class="py-2 px-3">Weight</th><th class="py-2 px-3">Batch</th></tr></thead><tbody>{pr}</tbody></table></section> |
| <section class="rounded-xl border border-slate-200 dark:border-slate-700 p-4"> |
| <h2 class="font-semibold mb-3">Live training</h2>{train}</section></div> |
| <footer class="text-center text-xs text-slate-400 pt-2">DaisyChainAI · pools compute, not memory · small models on spare hardware</footer> |
| </div></body></html>""" |
|
|
|
|
| class Handler(BaseHTTPRequestHandler): |
| def do_GET(self): |
| data = scan_cluster(load_nodes(), |
| int(EXPECTED_WORLD) if EXPECTED_WORLD else None, BASE_BATCH) |
| if self.path.startswith("/api"): |
| body = json.dumps(data).encode(); ctype = "application/json" |
| else: |
| body = render(data).encode(); ctype = "text/html; charset=utf-8" |
| self.send_response(200); self.send_header("Content-Type", ctype) |
| self.send_header("Content-Length", str(len(body))); self.end_headers() |
| self.wfile.write(body) |
|
|
| def log_message(self, *a): |
| pass |
|
|
|
|
| def main(): |
| print(f"[dashboard] :{PORT} nodes={NODES_FILE}", flush=True) |
| ThreadingHTTPServer(("0.0.0.0", PORT), Handler).serve_forever() |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
