commit 1b08c9592985d3ab843a350c74983c76fc98c818
Author: hz4th_coder <hz4th_coder@tphai.com>
Date:   Wed Sep 2 13:14:16 2026 +0800

    v1.0.0 模型评测网站：模型运行速度排行(解码/预填充/首字) + 账号体系 + 接收llm-speed-tester一键发送 + 演示数据 + 图表(16016)

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..917a1f9
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,4 @@
+logs/
+data/
+__pycache__/
+*.pyc
diff --git a/API.md b/API.md
new file mode 100644
index 0000000..164afa9
--- /dev/null
+++ b/API.md
@@ -0,0 +1,89 @@
+# 模型评测网站 API 文档
+
+Base: `http://<IP>:16066`（端口 16066）
+
+## 1. 接收提交（llm-speed-tester 一键发送）
+
+`POST /api/submit`
+
+Header: `Content-Type: application/json`，`X-Token: meval16066`
+
+```bash
+curl -X POST http://127.0.0.1:16066/api/submit \
+  -H "Content-Type: application/json" \
+  -H "X-Token: meval16066" \
+  -d '{
+    "account": "性能评测组",            # 账号，不存在自动创建
+    "source_test_id": 25,             # llm-speed-tester 测试ID
+    "source_site": "llm-speed-tester",
+    "provider": "openai",
+    "model": "nvidia/Qwen3.6-35B-A3B-NVFP4",
+    "test_name": "我的测试",
+    "summary": { "samples_ok":2, "samples_total":2, "avg_ttft_ms":180,
+                 "avg_prefill_speed":320.5, "avg_decode_speed":88.2,
+                 "avg_output_tokens":128, "avg_total_ms":2500,
+                 "concurrency_levels":[1],
+                 "by_length": {"2048": {"avg_ttft_ms":150,"avg_prefill_speed":350,
+                                         "avg_decode_speed":92,"avg_prompt_tokens":1500,
+                                         "avg_output_tokens":128,"avg_total_ms":2000}},
+                 "by_concurrency": {} },
+    "gen": {"context_lengths":[2048,8192], "max_tokens":128, "samples":2, "concurrency_levels":[1]},
+    "runs": []
+  }'
+```
+
+响应：
+
+```json
+{ "ok": true, "id": 21, "account_id": 2, "account": "性能评测组",
+  "model": "nvidia/Qwen3.6-35B-A3B-NVFP4",
+  "url": "/model.html?provider=openai&model=nvidia/Qwen3.6-35B-A3B-NVFP4" }
+```
+
+## 2. 排行 / 统计
+
+```bash
+# 模型速度排行（按 provider+model 聚合）
+GET /api/leaderboard?sort=avg_decode_speed&order=desc&limit=200
+# sort 可选: avg_decode_speed / avg_prefill_speed / avg_ttft_ms / best_decode / cnt / accounts / last_tested
+
+GET /api/stats        # 统计卡片
+GET /api/health       # 健康检查 + stats
+```
+
+## 3. 模型详情
+
+```bash
+GET /api/model?provider=openai&model=nvidia/Qwen3.6-35B-A3B-NVFP4
+```
+
+返回：该模型全部提交（submissions，含账号）+ 按上下文长度聚合（by_length）+ 折线图数据（line_csv/line_payload）。
+
+## 4. 提交管理
+
+```bash
+GET    /api/submissions?page=1&page_size=20&q=关键词&account_id=2&model=xxx&provider=xxx
+GET    /api/submissions/<id>              # 单条详情
+DELETE /api/submissions/<id>              # 删除
+GET    /api/submissions/<id>/chart        # 上下文长度→速度折线图 PNG
+```
+
+## 5. 账号管理
+
+```bash
+GET    /api/accounts
+POST   /api/accounts    {"name":"xxx","remark":"yyy"}        # 新增（重名 400）
+PUT    /api/accounts/<id>  {"name":"xxx","remark":"yyy"}     # 改名/备注
+DELETE /api/accounts/<id>                                    # 删除（连带其提交）
+```
+
+## 6. 其他
+
+```bash
+POST /api/seed-demo          # 🎲 生成演示评测数据
+POST /api/chart              # 图表代理（转发 data-chart-tool 16016，返回 PNG）
+```
+
+## 错误
+
+所有接口失败返回 `{"ok": false, "error": "..."}` + 对应 HTTP 状态码。
diff --git a/README.md b/README.md
new file mode 100644
index 0000000..43d18b3
--- /dev/null
+++ b/README.md
@@ -0,0 +1,54 @@
+# ⚡ 模型评测网站（model-eval-site）
+
+聚合展示各种大模型的**运行速度评测**（解码 / 预填充 / 首字延迟），并支持从
+[llm-speed-tester](../llm-speed-tester/) 一键把测试结果发送到**对应账号**下展示。
+
+- 端口：**16066**
+- 技术栈：Flask + SQLite + data-chart-tool(16016) 图表
+
+## 页面
+
+| 页面 | 说明 |
+|------|------|
+| `/`（index.html） | 🏆 模型速度排行：按模型+提供商聚合所有提交，表头可排序 + 解码速度 TOP20 对比图 |
+| `/model.html` | 单模型详情：按上下文长度的速度折线图 + 各账号提交明细 |
+| `/submissions.html` | 📥 提交管理：分页/搜索/详情弹窗/删除 + 「🎲 生成演示数据」 |
+| `/accounts.html` | 👥 账号管理：账号列表（含提交数）/ 新增 / 编辑 / 删除 |
+
+## 与 llm-speed-tester 联动
+
+1. 在 llm-speed-tester 左侧「📤 评测站同步」填评测站地址（默认 `http://127.0.0.1:16066`）和账号名
+2. 测试完成后，在测试历史行点「📤 发送」，或在详情弹窗点「📤 发送到评测站」
+3. 结果以 **POST /api/submit** 提交，账号不存在自动创建，提交后排行榜实时更新
+
+## API
+
+| 方法 | 路径 | 说明 |
+|------|------|------|
+| POST | `/api/submit` | 接收 llm-speed-tester 提交（头 `X-Token: meval16066`） |
+| GET | `/api/health` | 健康检查 + 统计 |
+| GET | `/api/stats` | 统计卡片（模型数/提交数/账号数/采样数） |
+| GET | `/api/leaderboard` | 模型速度排行（可 `sort`/`order`/`limit`） |
+| GET | `/api/model` | 单模型详情（`provider` + `model` 参数） |
+| GET | `/api/submissions` | 提交列表（`page`/`page_size`/`q`/`account_id`/`model`/`provider`） |
+| GET | `/api/submissions/<id>` | 单条提交详情 |
+| DELETE | `/api/submissions/<id>` | 删除提交 |
+| GET | `/api/submissions/<id>/chart` | 单条提交：上下文长度→速度折线图 PNG |
+| GET/POST | `/api/accounts` | 账号列表 / 新增 |
+| PUT/DELETE | `/api/accounts/<id>` | 改账号 / 删账号（连带删其提交） |
+| POST | `/api/seed-demo` | 🎲 生成演示评测数据（便于看排行效果） |
+| POST | `/api/chart` | 图表代理（转发 data-chart-tool 16016） |
+
+## 数据
+
+- `data/model_eval.db`：SQLite（accounts / submissions）
+- 详细字段见 `database.py`
+
+## 部署
+
+```bash
+./start.sh          # 启动（默认 16066，PID 文件管理）
+./start.sh stop     # 停止
+# 依赖: openclaw conda 环境 flask/requests；图表走 data-chart-tool(16016)
+# Git: hz4th_coder/model-eval-site
+```
diff --git a/app.py b/app.py
new file mode 100644
index 0000000..7dba527
--- /dev/null
+++ b/app.py
@@ -0,0 +1,320 @@
+# -*- coding: utf-8 -*-
+"""模型评测网站 - Flask 主应用
+功能：
+  1. 聚合展示各种模型的运行速度（解码/预填充/首字延迟 排行 + 图表）
+  2. 账号体系：每个提交挂到对应账号下
+  3. POST /api/submit 接收 llm-speed-tester 一键发送的测试结果
+"""
+import io
+
+import requests
+from flask import Flask, jsonify, request, send_file, send_from_directory
+
+import config
+import database as db
+
+app = Flask(__name__, static_folder="static", static_url_path="")
+app.json.ensure_ascii = False
+
+db.init_db()
+
+
+@app.after_request
+def _cors(resp):
+    resp.headers["Access-Control-Allow-Origin"] = "*"
+    resp.headers["Access-Control-Allow-Headers"] = "Content-Type, X-Token"
+    resp.headers["Access-Control-Allow-Methods"] = "GET, POST, PUT, DELETE, OPTIONS"
+    return resp
+
+
+@app.route("/")
+def index():
+    return send_from_directory(app.static_folder, "index.html")
+
+
+@app.route("/health")
+@app.route("/api/health")
+def health():
+    return jsonify({"ok": True, "port": config.PORT, "service": "model-eval-site",
+                    "stats": db.get_stats()})
+
+
+# ───────────────────────── 接收 llm-speed-tester 提交 ─────────────────────────
+
+@app.route("/api/submit", methods=["POST"])
+def submit():
+    """接收速度测试结果，挂到对应账号下（账号不存在自动创建）"""
+    body = request.get_json(force=True) or {}
+    token = request.headers.get("X-Token") or body.get("token") or ""
+    if token != config.SUBMIT_TOKEN:
+        return jsonify({"ok": False, "error": "提交密钥错误"}), 403
+    model = (body.get("model") or "").strip()
+    if not model:
+        return jsonify({"ok": False, "error": "缺少模型名称"}), 400
+    summary = body.get("summary") or {}
+    if not (summary.get("samples_ok") or 0):
+        return jsonify({"ok": False, "error": "该测试没有成功采样数据，无法发布到评测站"}), 400
+
+    account_name = (body.get("account") or "").strip() or "默认账号"
+    aid = db.get_or_create_account(account_name, remark="来自 llm-speed-tester")
+    sid = db.add_submission(aid, body)
+    account = db.list_accounts()
+    acc = next((a for a in account if a["id"] == aid), None)
+    return jsonify({
+        "ok": True, "id": sid, "account_id": aid,
+        "account": acc["name"] if acc else account_name,
+        "model": model,
+        "url": "/model.html?provider=%s&model=%s" % (
+            requests.utils.quote(body.get("provider") or ""),
+            requests.utils.quote(model)),
+    })
+
+
+# ───────────────────────── 汇总 / 排行 ─────────────────────────
+
+@app.route("/api/stats")
+def stats():
+    return jsonify(db.get_stats())
+
+
+@app.route("/api/leaderboard")
+def leaderboard():
+    sort = request.args.get("sort", "avg_decode_speed")
+    order = request.args.get("order", "desc")
+    limit = min(int(request.args.get("limit", 200) or 200), 500)
+    rows = db.leaderboard(sort=sort, order=order, limit=limit)
+    # 按排行榜生成柱状图 CSV（前 20 名，解码速度）
+    top = rows[:20]
+    csv_lines = ["模型, 解码速度(tok/s), 预填充速度(tok/s), 首字延迟(ms)"]
+    for r in top:
+        label = "%s %s" % (r["provider"], r["model"])
+        csv_lines.append("%s, %s, %s, %s" % (
+            label.replace(",", " "), _fmt(r["avg_decode_speed"]),
+            _fmt(r["avg_prefill_speed"]), _fmt(r["avg_ttft_ms"])))
+    chart_csv = "\n".join(csv_lines)
+    bar_payload = {
+        "data": chart_csv, "chartType": "bar",
+        "title": "模型解码速度排行 TOP%s（tok/s）" % len(top),
+        "theme": "default", "showLegend": True, "showGrid": True, "showLabel": True,
+        "seriesTypes": ["bar", "bar", "line"],
+        "seriesAxis": [0, 0, 1],
+        "seriesStyles": ["solid", "hollow", "dashed"],
+        "width": 1100, "height": 560, "pixelRatio": 2,
+    }
+    return jsonify({"ok": True, "rows": rows, "chart_csv": chart_csv,
+                    "bar_payload": bar_payload})
+
+
+@app.route("/api/model")
+def model_detail():
+    provider = request.args.get("provider", "")
+    model = request.args.get("model", "")
+    if not model:
+        return jsonify({"ok": False, "error": "缺少模型名称"}), 400
+    d = db.get_model_detail(provider, model)
+    if not d["submissions"]:
+        return jsonify({"ok": False, "error": "该模型暂无评测数据"}), 404
+    # 按上下文长度聚合 → 折线图 CSV
+    csv_lines = ["上下文长度(tok), 解码速度(tok/s), 预填充速度(tok/s)"]
+    for r in d["by_length"]:
+        csv_lines.append("%d, %s, %s" % (r["length"], r["avg_decode_speed"], r["avg_prefill_speed"]))
+    line_csv = "\n".join(csv_lines)
+    line_payload = {
+        "data": line_csv, "chartType": "line",
+        "title": "%s %s · 解码速度随上下文长度变化" % (provider or "", model),
+        "theme": "default", "showLegend": True, "showGrid": True, "showLabel": False,
+        "smoothLine": True, "dualYAxis": True,
+        "leftAxisName": "预填充速度(tok/s)", "rightAxisName": "解码速度(tok/s)",
+        "seriesTypes": ["line", "line"], "seriesAxis": [0, 1],
+        "seriesStyles": ["dashed", "solid"],
+        "width": 1000, "height": 520, "pixelRatio": 2,
+    }
+    d["line_csv"] = line_csv
+    d["line_payload"] = line_payload
+    return jsonify({"ok": True, **d})
+
+
+# ───────────────────────── 提交管理 ─────────────────────────
+
+@app.route("/api/submissions")
+def submissions():
+    page = max(1, int(request.args.get("page", 1)))
+    page_size = min(max(1, int(request.args.get("page_size", 20))), 100)
+    q = request.args.get("q", "")
+    account_id = request.args.get("account_id") or None
+    model = request.args.get("model", "")
+    provider = request.args.get("provider", "")
+    return jsonify(db.list_submissions(page=page, page_size=page_size, q=q,
+                                       account_id=int(account_id) if account_id else None,
+                                       model=model, provider=provider))
+
+
+@app.route("/api/submissions/<int:sid>")
+def submission_detail(sid):
+    s = db.get_submission(sid)
+    if not s:
+        return jsonify({"ok": False, "error": "提交不存在"}), 404
+    return jsonify(s)
+
+
+@app.route("/api/submissions/<int:sid>", methods=["DELETE"])
+def submission_delete(sid):
+    db.delete_submission(sid)
+    return jsonify({"ok": True})
+
+
+@app.route("/api/submissions/<int:sid>/chart")
+def submission_chart(sid):
+    """单条提交：上下文长度 → 解码/预填充速度 折线图"""
+    s = db.get_submission(sid)
+    if not s:
+        return jsonify({"ok": False, "error": "提交不存在"}), 404
+    by = s["by_length"]
+    csv_lines = ["上下文长度(tok), 预填充速度(tok/s), 解码速度(tok/s)"]
+    for L in sorted(by, key=int):
+        bl = by[L]
+        pre = bl.get("avg_prefill_speed")
+        dec = bl.get("avg_decode_speed")
+        if pre is None or dec is None:
+            continue
+        csv_lines.append("%s, %.2f, %.2f" % (L, pre, dec))
+    if len(csv_lines) < 2:
+        return jsonify({"ok": False, "error": "该提交无可画图的长度分组数据"}), 400
+    payload = {
+        "data": "\n".join(csv_lines), "chartType": "line",
+        "title": "%s %s · 速度随上下文长度（提交#%d）" % (s["provider"], s["model"], sid),
+        "theme": "default", "showLegend": True, "showGrid": True, "showLabel": False,
+        "smoothLine": True, "dualYAxis": True,
+        "leftAxisName": "预填充速度(tok/s)", "rightAxisName": "解码速度(tok/s)",
+        "seriesTypes": ["line", "line"], "seriesAxis": [0, 1],
+        "seriesStyles": ["dashed", "solid"],
+        "width": 1000, "height": 480, "pixelRatio": 2,
+    }
+    return _chart_proxy(payload)
+
+
+# ───────────────────────── 账号管理 ─────────────────────────
+
+@app.route("/api/accounts")
+def accounts():
+    return jsonify(db.list_accounts())
+
+
+@app.route("/api/accounts", methods=["POST"])
+def account_add():
+    body = request.get_json(force=True) or {}
+    try:
+        aid = db.add_account(body.get("name", ""), body.get("remark", ""))
+        return jsonify({"ok": True, "id": aid})
+    except ValueError as e:
+        return jsonify({"ok": False, "error": str(e)}), 400
+
+
+@app.route("/api/accounts/<int:aid>", methods=["PUT"])
+def account_update(aid):
+    body = request.get_json(force=True) or {}
+    db.rename_account(aid, body.get("name", ""), body.get("remark", ""))
+    return jsonify({"ok": True})
+
+
+@app.route("/api/accounts/<int:aid>", methods=["DELETE"])
+def account_delete(aid):
+    db.delete_account(aid)
+    return jsonify({"ok": True})
+
+
+# ───────────────────────── 图表代理（data-chart-tool） ─────────────────────────
+
+def _chart_proxy(payload):
+    try:
+        resp = requests.post(config.CHART_API_BASE + "/api/chart", json=payload, timeout=60)
+    except requests.RequestException as e:
+        return jsonify({"ok": False, "error": "图表服务不可用: %s" % e}), 502
+    if resp.status_code != 200:
+        return jsonify({"ok": False, "error": "图表生成失败(%d): %s" % (resp.status_code, resp.text[:300])}), 502
+    return send_file(io.BytesIO(resp.content), mimetype="image/png")
+
+
+@app.route("/api/chart", methods=["POST"])
+def chart_proxy():
+    payload = request.get_json(force=True) or {}
+    return _chart_proxy(payload)
+
+
+# ───────────────────────── 演示数据 ─────────────────────────
+
+DEMO_MODELS = [
+    ("autodl", "qwen3.5-plus", [512, 2048, 4096, 8192, 16384]),
+    ("autodl", "glm-5.3-flash", [512, 2048, 4096, 8192]),
+    ("siliconflow", "deepseek-v4-flash", [512, 2048, 4096, 8192, 16384, 32768]),
+    ("siliconflow", "longcat", [512, 2048, 4096, 8192]),
+    ("local-qwen", "qwen3.5-plus", [512, 2048, 4096, 8192, 16384]),
+    ("openai", "gpt-4o-mini", [512, 2048, 4096, 8192, 16384, 32768]),
+    ("anthropic", "claude-3.5-haiku", [512, 2048, 4096, 8192, 16384]),
+    ("google", "gemini-2.0-flash", [512, 2048, 4096, 8192, 16384, 32768]),
+]
+
+
+@app.route("/api/seed-demo", methods=["POST"])
+def seed_demo():
+    """生成一批演示评测数据（便于查看排行榜效果），重复调用会追加"""
+    import random
+    random.seed()
+    accounts = ["测试小组A", "性能评测组", "模型研究所"]
+    cnt = 0
+    for ai, acc_name in enumerate(accounts):
+        aid = db.get_or_create_account(acc_name, remark="演示账号")
+        for mi, (provider, model, lens) in enumerate(DEMO_MODELS):
+            if (mi + ai) % 3 == 0 and ai != 0:
+                continue  # 让数据分布有点差异
+            # 基础解码速度：给每个模型一个基准（越靠后越快一些随机）
+            base_decode = 40 + mi * 12 + random.uniform(-5, 12)
+            base_prefill = base_decode * random.uniform(0.55, 0.9)
+            base_ttft = 200 + mi * 15 + random.uniform(-40, 80)
+            by_length = {}
+            for L in lens:
+                # 长度越长解码略降
+                k = 1 - (L / 65536) * 0.3
+                by_length[str(L)] = {
+                    "samples_ok": 2, "samples_total": 2,
+                    "avg_decode_speed": round(base_decode * k, 1),
+                    "avg_prefill_speed": round(base_prefill * k, 1),
+                    "avg_ttft_ms": round(base_ttft + L * 0.02, 0),
+                    "avg_prompt_tokens": int(L * 0.75), "avg_output_tokens": 128,
+                    "avg_total_ms": round((L * 0.75 / base_prefill + 128 / base_decode) * 1000, 0),
+                }
+            speeds = [v["avg_decode_speed"] for v in by_length.values()]
+            payload = {
+                "token": config.SUBMIT_TOKEN,
+                "account": acc_name,
+                "source_test_id": 9000 + cnt, "source_site": "llm-speed-tester(演示)",
+                "provider": provider, "model": model,
+                "test_name": "%s 基准评测" % model,
+                "summary": {
+                    "samples_ok": 2, "samples_total": 2,
+                    "avg_decode_speed": round(sum(speeds) / len(speeds), 1),
+                    "avg_prefill_speed": round(sum(v["avg_prefill_speed"] for v in by_length.values()) / len(by_length), 1),
+                    "avg_ttft_ms": round(sum(v["avg_ttft_ms"] for v in by_length.values()) / len(by_length), 0),
+                    "avg_output_tokens": 128, "avg_total_ms": 3200,
+                    "min_decode_speed": min(speeds), "max_decode_speed": max(speeds),
+                    "concurrency_levels": [1],
+                    "by_length": by_length, "by_concurrency": {},
+                },
+                "gen": {"context_lengths": lens, "max_tokens": 128,
+                        "samples": 2, "concurrency_levels": [1]},
+                "runs": [],
+            }
+            db.add_submission(aid, payload)
+            cnt += 1
+    return jsonify({"ok": True, "seeded": cnt, "stats": db.get_stats()})
+
+
+def _fmt(v):
+    try:
+        return "%.2f" % float(v)
+    except Exception:
+        return ""
+
+
+if __name__ == "__main__":
+    app.run(host=config.HOST, port=config.PORT, threaded=True, debug=False)
diff --git a/config.py b/config.py
new file mode 100644
index 0000000..1fbe464
--- /dev/null
+++ b/config.py
@@ -0,0 +1,17 @@
+# -*- coding: utf-8 -*-
+"""模型评测网站 - 全局配置"""
+import os
+
+PORT = 16066
+HOST = "0.0.0.0"
+
+BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+DATA_DIR = os.path.join(BASE_DIR, "data")
+LOG_DIR = os.path.join(BASE_DIR, "logs")
+DB_PATH = os.path.join(DATA_DIR, "model_eval.db")
+
+# 提交接口共享密钥（llm-speed-tester 发送时携带，防止被随意写入）
+SUBMIT_TOKEN = "meval16066"
+
+# data-chart-tool 图表服务地址（画速度对比图）
+CHART_API_BASE = "http://127.0.0.1:16016"
diff --git a/database.py b/database.py
new file mode 100644
index 0000000..41ad87e
--- /dev/null
+++ b/database.py
@@ -0,0 +1,350 @@
+# -*- coding: utf-8 -*-
+"""SQLite 存储：账号 / 评测提交 / 提交明细"""
+import os
+import json
+import sqlite3
+import threading
+
+import config
+
+_lock = threading.RLock()
+
+SCHEMA = """
+CREATE TABLE IF NOT EXISTS accounts(
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  name TEXT UNIQUE NOT NULL,
+  remark TEXT DEFAULT '',
+  created_at TEXT DEFAULT (datetime('now','localtime'))
+);
+
+CREATE TABLE IF NOT EXISTS submissions(
+  id INTEGER PRIMARY KEY AUTOINCREMENT,
+  account_id INTEGER NOT NULL,
+  source_test_id INTEGER DEFAULT 0,
+  source_site TEXT DEFAULT 'llm-speed-tester',
+  created_at TEXT DEFAULT (datetime('now','localtime')),
+  provider TEXT DEFAULT '',
+  model TEXT DEFAULT '',
+  test_name TEXT DEFAULT '',
+  samples_ok INTEGER DEFAULT 0,
+  samples_total INTEGER DEFAULT 0,
+  avg_ttft_ms REAL,
+  avg_prefill_speed REAL,
+  avg_decode_speed REAL,
+  avg_stream_decode REAL,
+  avg_prompt_tokens REAL,
+  avg_output_tokens REAL,
+  avg_total_ms REAL,
+  min_ttft_ms REAL, max_ttft_ms REAL,
+  min_prefill_speed REAL, max_prefill_speed REAL,
+  min_decode_speed REAL, max_decode_speed REAL,
+  min_total_ms REAL, max_total_ms REAL,
+  context_lengths TEXT DEFAULT '[]',
+  concurrency_levels TEXT DEFAULT '[]',
+  by_length TEXT DEFAULT '{}',
+  by_concurrency TEXT DEFAULT '{}',
+  runs TEXT DEFAULT '[]',
+  raw TEXT DEFAULT '{}'
+);
+
+CREATE INDEX IF NOT EXISTS idx_sub_model ON submissions(model, provider);
+CREATE INDEX IF NOT EXISTS idx_sub_account ON submissions(account_id);
+CREATE INDEX IF NOT EXISTS idx_sub_created ON submissions(created_at);
+"""
+
+
+def _connect():
+    os.makedirs(config.DATA_DIR, exist_ok=True)
+    conn = sqlite3.connect(config.DB_PATH, check_same_thread=False, timeout=30)
+    conn.row_factory = sqlite3.Row
+    conn.execute("PRAGMA journal_mode=WAL")
+    return conn
+
+
+def init_db():
+    with _lock:
+        conn = _connect()
+        try:
+            conn.executescript(SCHEMA)
+            conn.commit()
+        finally:
+            conn.close()
+
+
+# ───────────────────────── 账号 ─────────────────────────
+
+def get_or_create_account(name: str, remark: str = "") -> int:
+    name = (name or "").strip() or "默认账号"
+    with _lock:
+        conn = _connect()
+        try:
+            r = conn.execute("SELECT id FROM accounts WHERE name=?", (name,)).fetchone()
+            if r:
+                return r["id"]
+            cur = conn.execute("INSERT INTO accounts(name, remark) VALUES(?,?)",
+                               (name, (remark or "").strip()))
+            conn.commit()
+            return cur.lastrowid
+        finally:
+            conn.close()
+
+
+def add_account(name: str, remark: str = "") -> int:
+    name = (name or "").strip()
+    if not name:
+        raise ValueError("账号名称不能为空")
+    with _lock:
+        conn = _connect()
+        try:
+            r = conn.execute("SELECT id FROM accounts WHERE name=?", (name,)).fetchone()
+            if r:
+                raise ValueError("账号已存在")
+            cur = conn.execute("INSERT INTO accounts(name, remark) VALUES(?,?)", (name, remark))
+            conn.commit()
+            return cur.lastrowid
+        finally:
+            conn.close()
+
+
+def list_accounts():
+    with _lock:
+        conn = _connect()
+        try:
+            rows = conn.execute(
+                "SELECT a.id,a.name,a.remark,a.created_at,"
+                "(SELECT COUNT(*) FROM submissions s WHERE s.account_id=a.id) AS cnt "
+                "FROM accounts a ORDER BY cnt DESC, a.id").fetchall()
+            return [dict(r) for r in rows]
+        finally:
+            conn.close()
+
+
+def rename_account(aid: int, name: str, remark: str):
+    with _lock:
+        conn = _connect()
+        try:
+            conn.execute("UPDATE accounts SET name=?, remark=? WHERE id=?",
+                         (name.strip(), (remark or "").strip(), aid))
+            conn.commit()
+        finally:
+            conn.close()
+
+
+def delete_account(aid: int):
+    with _lock:
+        conn = _connect()
+        try:
+            conn.execute("DELETE FROM submissions WHERE account_id=?", (aid,))
+            conn.execute("DELETE FROM accounts WHERE id=?", (aid,))
+            conn.commit()
+        finally:
+            conn.close()
+
+
+# ───────────────────────── 提交 ─────────────────────────
+
+def add_submission(account_id: int, data: dict) -> int:
+    """data 是 llm-speed-tester 发来的整包（含 summary/gen/runs/config）"""
+    summary = data.get("summary") or {}
+    gen = data.get("gen") or {}
+    runs = data.get("runs") or []
+    with _lock:
+        conn = _connect()
+        try:
+            cur = conn.execute(
+                "INSERT INTO submissions(account_id,source_test_id,source_site,provider,model,test_name,"
+                "samples_ok,samples_total,avg_ttft_ms,avg_prefill_speed,avg_decode_speed,avg_stream_decode,"
+                "avg_prompt_tokens,avg_output_tokens,avg_total_ms,min_ttft_ms,max_ttft_ms,min_prefill_speed,"
+                "max_prefill_speed,min_decode_speed,max_decode_speed,min_total_ms,max_total_ms,"
+                "context_lengths,concurrency_levels,by_length,by_concurrency,runs,raw) VALUES("
+                "?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?)",
+                (account_id,
+                 int(data.get("source_test_id") or 0),
+                 data.get("source_site") or "llm-speed-tester",
+                 data.get("provider") or "",
+                 data.get("model") or "",
+                 data.get("test_name") or data.get("name") or "",
+                 int(summary.get("samples_ok") or 0),
+                 int(summary.get("samples_total") or 0),
+                 summary.get("avg_ttft_ms"), summary.get("avg_prefill_speed"),
+                 summary.get("avg_decode_speed"), summary.get("avg_stream_decode"),
+                 summary.get("avg_prompt_tokens"), summary.get("avg_output_tokens"),
+                 summary.get("avg_total_ms"),
+                 summary.get("min_ttft_ms"), summary.get("max_ttft_ms"),
+                 summary.get("min_prefill_speed"), summary.get("max_prefill_speed"),
+                 summary.get("min_decode_speed"), summary.get("max_decode_speed"),
+                 summary.get("min_total_ms"), summary.get("max_total_ms"),
+                 json.dumps(gen.get("context_lengths") or summary.get("context_lengths") or [],
+                            ensure_ascii=False),
+                 json.dumps(summary.get("concurrency_levels") or gen.get("concurrency_levels") or [1],
+                            ensure_ascii=False),
+                 json.dumps(summary.get("by_length") or {}, ensure_ascii=False),
+                 json.dumps(summary.get("by_concurrency") or {}, ensure_ascii=False),
+                 json.dumps(runs[:200], ensure_ascii=False),
+                 json.dumps(data, ensure_ascii=False)))
+            conn.commit()
+            return cur.lastrowid
+        finally:
+            conn.close()
+
+
+def get_submission(sid: int):
+    with _lock:
+        conn = _connect()
+        try:
+            r = conn.execute("SELECT * FROM submissions WHERE id=?", (sid,)).fetchone()
+            if not r:
+                return None
+            d = dict(r)
+            d["account"] = conn.execute("SELECT name FROM accounts WHERE id=?",
+                                        (d["account_id"],)).fetchone()["name"]
+            for k in ("context_lengths", "concurrency_levels", "by_length",
+                      "by_concurrency", "runs", "raw"):
+                d[k] = json.loads(d[k] or ("[]" if k in ("context_lengths", "concurrency_levels", "runs") else "{}"))
+            return d
+        finally:
+            conn.close()
+
+
+def list_submissions(page=1, page_size=20, q="", account_id=None, model="", provider=""):
+    with _lock:
+        conn = _connect()
+        try:
+            where = ["1=1"]
+            args = []
+            if q:
+                where.append("(s.model LIKE ? OR s.test_name LIKE ? OR a.name LIKE ? OR s.provider LIKE ?)")
+                like = "%" + q + "%"
+                args += [like, like, like, like]
+            if account_id:
+                where.append("s.account_id=?")
+                args.append(account_id)
+            if model:
+                where.append("s.model=?")
+                args.append(model)
+            if provider:
+                where.append("s.provider=?")
+                args.append(provider)
+            wsql = " AND ".join(where)
+            total = conn.execute(
+                "SELECT COUNT(*) c FROM submissions s LEFT JOIN accounts a ON a.id=s.account_id WHERE " + wsql,
+                args).fetchone()["c"]
+            rows = conn.execute(
+                "SELECT s.*, a.name AS account FROM submissions s "
+                "LEFT JOIN accounts a ON a.id=s.account_id WHERE " + wsql +
+                " ORDER BY s.id DESC LIMIT ? OFFSET ?",
+                args + [page_size, (page - 1) * page_size]).fetchall()
+            out = []
+            for r in rows:
+                d = dict(r)
+                for k in ("context_lengths", "concurrency_levels"):
+                    d[k] = json.loads(d[k] or "[]")
+                out.append(d)
+            return {"total": total, "items": out, "page": page, "page_size": page_size,
+                    "pages": max(1, -(-total // page_size))}
+        finally:
+            conn.close()
+
+
+def delete_submission(sid: int):
+    with _lock:
+        conn = _connect()
+        try:
+            conn.execute("DELETE FROM submissions WHERE id=?", (sid,))
+            conn.commit()
+        finally:
+            conn.close()
+
+
+# ───────────────────────── 汇总统计 ─────────────────────────
+
+def get_stats():
+    with _lock:
+        conn = _connect()
+        try:
+            model_cnt = conn.execute("SELECT COUNT(DISTINCT model) c FROM submissions WHERE model<>''").fetchone()["c"]
+            sub_cnt = conn.execute("SELECT COUNT(*) c FROM submissions").fetchone()["c"]
+            acc_cnt = conn.execute("SELECT COUNT(*) c FROM accounts").fetchone()["c"]
+            ok_cnt = conn.execute("SELECT SUM(samples_ok) c FROM submissions").fetchone()["c"] or 0
+            return {"models": model_cnt, "submissions": sub_cnt,
+                    "accounts": acc_cnt, "samples_ok": ok_cnt}
+        finally:
+            conn.close()
+
+
+def leaderboard(sort="avg_decode_speed", order="desc", limit=200):
+    """按 (provider, model) 聚合所有提交 → 模型速度排行"""
+    sort_whitelist = {
+        "avg_decode_speed": "avg_decode_speed", "avg_prefill_speed": "avg_prefill_speed",
+        "avg_ttft_ms": "avg_ttft_ms", "best_decode": "best_decode",
+        "cnt": "cnt", "last_tested": "last_tested", "avg_stream_decode": "avg_stream_decode",
+    }
+    srt = sort_whitelist.get(sort, "avg_decode_speed")
+    od = "DESC" if order == "asc" and srt not in ("avg_ttft_ms", "avg_total_ms") else (
+        "ASC" if order == "asc" else "DESC")
+    with _lock:
+        conn = _connect()
+        try:
+            rows = conn.execute(
+                "SELECT provider, model, "
+                "COUNT(*) AS cnt, "
+                "COUNT(DISTINCT account_id) AS accounts, "
+                "AVG(avg_decode_speed) AS avg_decode_speed, "
+                "AVG(avg_prefill_speed) AS avg_prefill_speed, "
+                "AVG(avg_stream_decode) AS avg_stream_decode, "
+                "AVG(avg_ttft_ms) AS avg_ttft_ms, "
+                "AVG(avg_output_tokens) AS avg_output_tokens, "
+                "AVG(avg_total_ms) AS avg_total_ms, "
+                "MAX(avg_decode_speed) AS best_decode, "
+                "MAX(created_at) AS last_tested "
+                "FROM submissions WHERE model<>'' "
+                "GROUP BY provider, model ORDER BY %s %s LIMIT ?" % (srt, od),
+                (limit,)).fetchall()
+            return [dict(r) for r in rows]
+        finally:
+            conn.close()
+
+
+def get_model_detail(provider, model):
+    """单个模型的所有提交 + 按上下文长度聚合"""
+    with _lock:
+        conn = _connect()
+        try:
+            rows = conn.execute(
+                "SELECT s.*, a.name AS account FROM submissions s "
+                "LEFT JOIN accounts a ON a.id=s.account_id "
+                "WHERE s.model=? AND s.provider=? ORDER BY s.id DESC",
+                (model, provider)).fetchall()
+            subs = []
+            agg_length = {}
+            for r in rows:
+                d = dict(r)
+                d["by_length"] = json.loads(d.pop("by_length") or "{}")
+                d["by_concurrency"] = json.loads(d.pop("by_concurrency") or "{}")
+                d["context_lengths"] = json.loads(d.pop("context_lengths") or "[]")
+                d["concurrency_levels"] = json.loads(d.pop("concurrency_levels") or "[]")
+                subs.append(d)
+                for L, bl in d["by_length"].items():
+                    a = agg_length.setdefault(str(L), {"count": 0, "sum_decode": 0.0,
+                                                       "sum_prefill": 0.0, "sum_ttft": 0.0})
+                    if bl.get("avg_decode_speed") is not None:
+                        a["count"] += 1
+                        a["sum_decode"] += float(bl["avg_decode_speed"] or 0)
+                        a["sum_prefill"] += float(bl.get("avg_prefill_speed") or 0)
+                        a["sum_ttft"] += float(bl.get("avg_ttft_ms") or 0)
+            length_rows = []
+            for L in sorted(agg_length, key=int):
+                a = agg_length[L]
+                if not a["count"]:
+                    continue
+                length_rows.append({
+                    "length": int(L), "count": a["count"],
+                    "avg_decode_speed": round(a["sum_decode"] / a["count"], 2),
+                    "avg_prefill_speed": round(a["sum_prefill"] / a["count"], 2),
+                    "avg_ttft_ms": round(a["sum_ttft"] / a["count"], 1),
+                })
+            return {"provider": provider, "model": model,
+                    "submissions": subs, "by_length": length_rows,
+                    "sub_cnt": len(subs)}
+        finally:
+            conn.close()
diff --git a/start.sh b/start.sh
new file mode 100755
index 0000000..1fb8286
--- /dev/null
+++ b/start.sh
@@ -0,0 +1,39 @@
+#!/bin/bash
+# 模型评测网站 启动/停止脚本
+cd "$(dirname "$0")"
+PY=/home/hz1/miniconda3/envs/openclaw/bin/python3
+PORT=16066
+PID_FILE=logs/app.pid
+mkdir -p logs data
+
+start() {
+  if [ -f "$PID_FILE" ] && kill -0 "$(cat "$PID_FILE")" 2>/dev/null; then
+    echo "已在运行 (PID $(cat "$PID_FILE"))"
+    return
+  fi
+  nohup "$PY" app.py >> logs/app.log 2>&1 &
+  echo $! > "$PID_FILE"
+  sleep 1
+  if kill -0 "$(cat "$PID_FILE")" 2>/dev/null; then
+    echo "启动成功 → http://<IP>:$PORT/"
+  else
+    echo "启动失败，请查看 logs/app.log"
+  fi
+}
+
+stop() {
+  if [ -f "$PID_FILE" ]; then
+    kill "$(cat "$PID_FILE")" 2>/dev/null
+    rm -f "$PID_FILE"
+    echo "已停止"
+  else
+    echo "未在运行"
+  fi
+}
+
+case "${1:-start}" in
+  start) start ;;
+  stop) stop ;;
+  restart) stop; sleep 1; start ;;
+  *) echo "用法: $0 [start|stop|restart]" ;;
+esac
diff --git a/static/accounts.html b/static/accounts.html
new file mode 100644
index 0000000..85dc8eb
--- /dev/null
+++ b/static/accounts.html
@@ -0,0 +1,39 @@
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+<meta charset="UTF-8">
+<meta name="viewport" content="width=device-width, initial-scale=1.0">
+<title>账号管理 · 模型评测站</title>
+<link rel="stylesheet" href="/css/style.css">
+</head>
+<body>
+<div class="topbar">
+  <div class="brand"><span class="logo">⚡</span> 模型评测站</div>
+  <nav class="nav">
+    <a href="/index.html">🏆 速度排行</a>
+    <a href="/submissions.html">📥 提交管理</a>
+    <a href="/accounts.html">👥 账号</a>
+  </nav>
+  <div class="spacer"></div>
+</div>
+
+<div class="container">
+  <div class="panel">
+    <h2>👥 测试账号 <span class="hint">（llm-speed-tester 发送时选择账号归属，不存在会自动创建）</span></h2>
+    <div class="form-row">
+      <div><label>账号名称</label><input type="text" id="acc-name" placeholder="如：性能评测组" style="width:220px"></div>
+      <div><label>备注</label><input type="text" id="acc-remark" placeholder="选填" style="width:220px"></div>
+      <button class="btn primary" id="btn-add-acc" style="margin-top:18px">➕ 新增账号</button>
+    </div>
+    <div class="tbl-wrap">
+      <table id="accs"><thead><tr>
+        <th>账号</th><th>备注</th><th>创建时间</th><th class="num">提交数</th><th>操作</th>
+      </tr></thead><tbody><tr><td colspan="5" class="empty">加载中…</td></tr></tbody></table>
+    </div>
+  </div>
+</div>
+
+<div id="toast"></div>
+<script src="/js/app.js"></script>
+</body>
+</html>
diff --git a/static/css/style.css b/static/css/style.css
new file mode 100644
index 0000000..5ee3784
--- /dev/null
+++ b/static/css/style.css
@@ -0,0 +1,172 @@
+/* 模型评测网站 - 深色仪表盘风格 */
+:root {
+  --bg: #0f1420;
+  --panel: #171e2e;
+  --panel2: #1c2436;
+  --border: #26314a;
+  --text: #e6ebf5;
+  --muted: #8b97b3;
+  --accent: #4f8cff;
+  --accent2: #22c55e;
+  --warn: #f59e0b;
+  --danger: #ef4444;
+  --radius: 12px;
+}
+* { box-sizing: border-box; margin: 0; padding: 0; }
+body {
+  background: var(--bg);
+  color: var(--text);
+  font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", "PingFang SC", "Microsoft YaHei", sans-serif;
+  font-size: 14px;
+  line-height: 1.5;
+}
+a { color: var(--accent); text-decoration: none; }
+a:hover { text-decoration: underline; }
+.hint { color: var(--muted); font-size: 12px; }
+
+/* 顶部导航 */
+.topbar {
+  display: flex; align-items: center; gap: 18px;
+  padding: 14px 24px; background: var(--panel);
+  border-bottom: 1px solid var(--border); position: sticky; top: 0; z-index: 20;
+}
+.brand { font-size: 16px; font-weight: 700; display: flex; align-items: center; gap: 8px; }
+.brand .logo { font-size: 20px; }
+.nav { display: flex; gap: 4px; }
+.nav a {
+  color: var(--muted); padding: 6px 12px; border-radius: 8px; font-size: 13px;
+}
+.nav a:hover { color: var(--text); text-decoration: none; background: var(--panel2); }
+.nav a.active { color: #fff; background: var(--accent); }
+.topbar .spacer { flex: 1; }
+
+.container { max-width: 1280px; margin: 0 auto; padding: 20px 24px 60px; }
+
+/* 统计卡片 */
+.stat-cards { display: grid; grid-template-columns: repeat(auto-fit, minmax(180px, 1fr)); gap: 14px; margin-bottom: 18px; }
+.stat-card {
+  background: var(--panel); border: 1px solid var(--border); border-radius: var(--radius);
+  padding: 16px 18px;
+}
+.stat-card .k { color: var(--muted); font-size: 12px; }
+.stat-card .v { font-size: 26px; font-weight: 700; margin-top: 4px; }
+.stat-card .v small { font-size: 13px; color: var(--muted); font-weight: 400; }
+
+.panel {
+  background: var(--panel); border: 1px solid var(--border);
+  border-radius: var(--radius); padding: 18px 20px; margin-bottom: 18px;
+}
+.panel h2 { font-size: 15px; margin-bottom: 12px; display: flex; align-items: center; gap: 8px; }
+.panel h3 { font-size: 14px; margin: 14px 0 10px; }
+
+/* 表格 */
+table { width: 100%; border-collapse: collapse; }
+th, td { padding: 9px 10px; text-align: left; border-bottom: 1px solid var(--border); font-size: 13px; }
+th { color: var(--muted); font-weight: 600; white-space: nowrap; cursor: pointer; user-select: none; }
+th.sortable:hover { color: var(--text); }
+th .arrow { font-size: 10px; color: var(--accent); }
+td.num, th.num { text-align: right; font-variant-numeric: tabular-nums; }
+tr:hover td { background: rgba(79, 140, 255, 0.05); }
+.tbl-wrap { overflow-x: auto; }
+
+/* 徽标 / pill */
+.pill {
+  display: inline-block; padding: 2px 9px; border-radius: 20px; font-size: 11px;
+  background: rgba(79, 140, 255, 0.15); color: var(--accent); white-space: nowrap;
+}
+.pill.green { background: rgba(34, 197, 94, 0.15); color: var(--accent2); }
+.pill.amber { background: rgba(245, 158, 11, 0.15); color: var(--warn); }
+.pill.red { background: rgba(239, 68, 68, 0.15); color: var(--danger); }
+.pill.gray { background: rgba(139, 151, 179, 0.15); color: var(--muted); }
+.badge-decode { font-weight: 700; color: var(--accent2); }
+
+/* 按钮 */
+.btn {
+  display: inline-block; padding: 6px 13px; border-radius: 8px; border: 1px solid var(--border);
+  background: var(--panel2); color: var(--text); font-size: 12.5px; cursor: pointer;
+  transition: all .15s; white-space: nowrap;
+}
+.btn:hover { border-color: var(--accent); color: #fff; }
+.btn.primary { background: var(--accent); border-color: var(--accent); color: #fff; }
+.btn.primary:hover { opacity: .88; }
+.btn.danger { color: var(--danger); border-color: rgba(239,68,68,.4); }
+.btn.danger:hover { background: rgba(239,68,68,.12); }
+.btn.small { padding: 3px 9px; font-size: 12px; }
+.btn.link { background: none; border: none; color: var(--accent); padding: 0; }
+.btn.link:hover { text-decoration: underline; }
+.btn:disabled { opacity: .5; cursor: not-allowed; }
+
+/* 表单 */
+input[type=text], input[type=search], input[type=number], select, textarea {
+  background: var(--panel2); border: 1px solid var(--border); color: var(--text);
+  border-radius: 8px; padding: 7px 10px; font-size: 13px; outline: none;
+}
+input:focus, select:focus, textarea:focus { border-color: var(--accent); }
+label { font-size: 12px; color: var(--muted); display: block; margin-bottom: 4px; }
+.form-row { display: flex; gap: 10px; align-items: center; flex-wrap: wrap; margin-bottom: 10px; }
+.form-row label { margin: 0; }
+.form-grid { display: grid; grid-template-columns: 1fr 1fr; gap: 12px; }
+@media (max-width: 760px) { .form-grid { grid-template-columns: 1fr; } }
+
+/* 弹窗 */
+.mask {
+  position: fixed; inset: 0; background: rgba(6, 9, 16, .72); z-index: 100;
+  display: flex; align-items: center; justify-content: center; padding: 24px;
+}
+.modal {
+  background: var(--panel); border: 1px solid var(--border); border-radius: 14px;
+  width: 100%; max-width: 860px; max-height: 88vh; display: flex; flex-direction: column;
+  box-shadow: 0 20px 60px rgba(0,0,0,.5);
+}
+.modal-head {
+  display: flex; align-items: center; justify-content: space-between;
+  padding: 14px 18px; border-bottom: 1px solid var(--border);
+}
+.modal-head h3 { margin: 0; font-size: 15px; }
+.modal-body { padding: 16px 18px; overflow-y: auto; }
+.modal-foot { padding: 12px 18px; border-top: 1px solid var(--border); display: flex; gap: 10px; justify-content: flex-end; }
+.mask[hidden] { display: none; }
+
+/* kv 信息块 */
+.kv { display: grid; grid-template-columns: repeat(auto-fill, minmax(170px, 1fr)); gap: 10px; }
+.kv-item { background: var(--panel2); border-radius: 8px; padding: 8px 12px; }
+.kv-k { font-size: 11px; color: var(--muted); }
+.kv-v { font-size: 14px; font-weight: 600; }
+
+/* 图表 */
+.chart-img-wrap { text-align: center; padding: 8px 0; }
+.chart-img-wrap img { max-width: 100%; border-radius: 8px; }
+.chart-csv {
+  width: 100%; min-height: 90px; font-family: ui-monospace, monospace; font-size: 12px;
+  margin-top: 6px;
+}
+
+/* 空态 / 提示 */
+.empty { color: var(--muted); text-align: center; padding: 30px 0; }
+.notice {
+  background: rgba(79,140,255,.1); border: 1px solid rgba(79,140,255,.3);
+  border-radius: 8px; padding: 10px 14px; font-size: 12.5px; color: #cfe0ff;
+  margin-bottom: 12px;
+}
+.notice.green { background: rgba(34,197,94,.1); border-color: rgba(34,197,94,.3); color: #d4f7e2; }
+.notice.red { background: rgba(239,68,68,.1); border-color: rgba(239,68,68,.3); color: #ffd7d7; }
+
+/* toast */
+#toast {
+  position: fixed; bottom: 26px; left: 50%; transform: translateX(-50%);
+  background: #1d2538; border: 1px solid var(--accent); color: #fff;
+  padding: 10px 18px; border-radius: 10px; z-index: 999; font-size: 13px;
+  box-shadow: 0 6px 20px rgba(0,0,0,.4); display: none; max-width: 80vw;
+}
+
+.pagination { display: flex; gap: 8px; align-items: center; margin-top: 12px; }
+.pagination button { padding: 4px 10px; }
+.pagination .info { color: var(--muted); font-size: 12px; }
+
+/* 详情弹窗内的小表 */
+.modal-body table.mini th { font-size: 12px; }
+.modal-body table.mini td { font-size: 12.5px; }
+
+/* 进度条 */
+.meter { background: var(--panel2); border-radius: 6px; height: 8px; overflow: hidden; margin-top: 6px; }
+.meter > div { height: 100%; background: linear-gradient(90deg, var(--accent), var(--accent2)); border-radius: 6px; }
diff --git a/static/index.html b/static/index.html
new file mode 100644
index 0000000..2490a40
--- /dev/null
+++ b/static/index.html
@@ -0,0 +1,65 @@
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+<meta charset="UTF-8">
+<meta name="viewport" content="width=device-width, initial-scale=1.0">
+<title>模型评测 · 运行速度排行</title>
+<link rel="stylesheet" href="/css/style.css">
+</head>
+<body>
+<div class="topbar">
+  <div class="brand"><span class="logo">⚡</span> 模型评测站</div>
+  <nav class="nav">
+    <a href="/index.html">🏆 速度排行</a>
+    <a href="/submissions.html">📥 提交管理</a>
+    <a href="/accounts.html">👥 账号</a>
+  </nav>
+  <div class="spacer"></div>
+  <span class="hint">llm-speed-tester 一键同步</span>
+</div>
+
+<div class="container">
+  <div class="stat-cards">
+    <div class="stat-card"><div class="k">评测模型数</div><div class="v" id="st-models">—</div></div>
+    <div class="stat-card"><div class="k">评测提交数</div><div class="v" id="st-subs">—</div></div>
+    <div class="stat-card"><div class="k">测试账号数</div><div class="v" id="st-accounts">—</div></div>
+    <div class="stat-card"><div class="k">成功采样</div><div class="v" id="st-samples">—</div></div>
+  </div>
+
+  <div class="panel">
+    <h2>🏆 模型运行速度排行 <span class="hint">（点击表头可排序；按模型+提供商聚合所有提交）</span></h2>
+    <div class="tbl-wrap">
+      <table id="lb">
+        <thead><tr>
+          <th class="num">#</th>
+          <th>提供商</th>
+          <th>模型</th>
+          <th class="num" data-sort="cnt">提交<span class="arrow"></span></th>
+          <th class="num" data-sort="accounts">账号<span class="arrow"></span></th>
+          <th class="num" data-sort="avg_decode_speed">平均解码 tok/s<span class="arrow"></span></th>
+          <th class="num" data-sort="avg_prefill_speed">平均预填充 tok/s<span class="arrow"></span></th>
+          <th class="num" data-sort="avg_ttft_ms">首字 ms<span class="arrow"></span></th>
+          <th class="num" data-sort="best_decode">最佳解码 tok/s<span class="arrow"></span></th>
+          <th class="num" data-sort="last_tested">最后测试<span class="arrow"></span></th>
+        </tr></thead>
+        <tbody><tr><td colspan="10" class="empty">加载中…</td></tr></tbody>
+      </table>
+    </div>
+  </div>
+
+  <div class="panel">
+    <h2>📊 解码速度 TOP20 对比图</h2>
+    <div class="chart-toolbar">
+      <button class="btn small primary" id="lb-chart-gen">🎨 生成/刷新图表</button>
+      <span class="hint" id="lb-chart-status"></span>
+    </div>
+    <div class="chart-img-wrap" id="lb-chart-img">
+      <div class="hint">点击上方按钮绘制：模型解码速度（实心柱）+ 预填充（空心柱）+ 首字延迟（虚线）</div>
+    </div>
+  </div>
+</div>
+
+<div id="toast"></div>
+<script src="/js/app.js"></script>
+</body>
+</html>
diff --git a/static/js/app.js b/static/js/app.js
new file mode 100644
index 0000000..2a73be6
--- /dev/null
+++ b/static/js/app.js
@@ -0,0 +1,482 @@
+/* 模型评测网站 前端逻辑 */
+"use strict";
+
+const $ = (s) => document.querySelector(s);
+const $$ = (s) => Array.from(document.querySelectorAll(s));
+
+const esc = (s) => String(s ?? "").replace(/[&<>"']/g,
+  (c) => ({ "&": "&amp;", "<": "&lt;", ">": "&gt;", '"': "&quot;", "'": "&#39;" }[c]));
+const fmt = (v, d = "—") => (v === null || v === undefined || isNaN(v) ? d : Number(v).toFixed(2).replace(/\.?0+$/, ""));
+const fmtInt = (v, d = "—") => (v === null || v === undefined || isNaN(v) ? d : Math.round(v));
+const toastEl = () => $("#toast");
+let toastTimer = null;
+function toast(msg) {
+  const el = toastEl();
+  el.textContent = msg;
+  el.style.display = "block";
+  clearTimeout(toastTimer);
+  toastTimer = setTimeout(() => { el.style.display = "none"; }, 3200);
+}
+
+async function api(path, method = "GET", body) {
+  const opt = { method, headers: {} };
+  if (body !== undefined) { opt.headers["Content-Type"] = "application/json"; opt.body = JSON.stringify(body); }
+  const resp = await fetch(path, opt);
+  const j = await resp.json().catch(() => ({}));
+  if (!resp.ok && !j.ok) throw new Error(j.error || `请求失败(${resp.status})`);
+  return j;
+}
+
+function providerLabel(p) {
+  const m = { autodl: "Autodl", siliconflow: "SiliconFlow", "local-qwen": "本地Qwen",
+              openai: "OpenAI", anthropic: "Anthropic", google: "Gemini", local: "本地" };
+  return m[p] || p || "—";
+}
+
+/* 顶部导航高亮 */
+function navActive() {
+  const path = location.pathname.split("/").pop() || "index.html";
+  $$(".nav a").forEach((a) => {
+    a.classList.toggle("active", a.getAttribute("href") === path);
+  });
+}
+
+/* ───────────────────────── 首页：排行 ───────────────────────── */
+let sortKey = "avg_decode_speed";
+let sortOrder = "desc";
+
+async function loadStats() {
+  try {
+    const s = await api("/api/stats");
+    $("#st-models").textContent = s.models;
+    $("#st-subs").textContent = s.submissions;
+    $("#st-accounts").textContent = s.accounts;
+    $("#st-samples").innerHTML = fmtInt(s.samples_ok) + " <small>次成功采样</small>";
+  } catch (e) { /* ignore */ }
+}
+
+function sortClick(key) {
+  if (sortKey === key) sortOrder = sortOrder === "desc" ? "asc" : "desc";
+  else { sortKey = key; sortOrder = "desc"; }
+  loadLeaderboard();
+}
+
+async function loadLeaderboard() {
+  const tb = $("#lb tbody");
+  if (!tb) return;
+  tb.innerHTML = '<tr><td colspan="9" class="empty">加载中…</td></tr>';
+  let d;
+  try {
+    d = await api(`/api/leaderboard?sort=${sortKey}&order=${sortOrder}`);
+  } catch (e) {
+    tb.innerHTML = `<tr><td colspan="9" class="empty">加载失败：${esc(e.message)}</td></tr>`;
+    return;
+  }
+  $$("#lb th[data-sort]").forEach((th) => {
+    th.querySelector(".arrow").textContent = (th.dataset.sort === sortKey) ? (sortOrder === "desc" ? " ▼" : " ▲") : "";
+  });
+  if (!d.rows.length) {
+    tb.innerHTML = '<tr><td colspan="9" class="empty">暂无评测数据。可在「提交管理」页面点击「🎲 生成演示数据」，或从 llm-speed-tester 一键发送。</td></tr>';
+    return;
+  }
+  const maxDecode = Math.max(...d.rows.map((r) => r.avg_decode_speed || 0));
+  tb.innerHTML = d.rows.map((r, i) => {
+    const meter = r.avg_decode_speed ? `<div class="meter"><div style="width:${(r.avg_decode_speed / maxDecode * 100).toFixed(1)}%"></div></div>` : "";
+    return `<tr>
+      <td class="num">${i + 1}</td>
+      <td><span class="pill">${esc(providerLabel(r.provider))}</span></td>
+      <td><a href="/model.html?provider=${encodeURIComponent(r.provider)}&model=${encodeURIComponent(r.model)}" title="查看详情">${esc(r.model)}</a></td>
+      <td class="num">${fmtInt(r.cnt)}</td>
+      <td class="num">${fmtInt(r.accounts)}</td>
+      <td class="num"><span class="badge-decode">${fmt(r.avg_decode_speed)}</span>${meter}</td>
+      <td class="num">${fmt(r.avg_prefill_speed)}</td>
+      <td class="num">${fmt(r.avg_ttft_ms)}</td>
+      <td class="num">${fmt(r.best_decode)}</td>
+      <td class="num">${esc((r.last_tested || "").slice(5, 16))}</td>
+    </tr>`;
+  }).join("");
+  window.__lbRows = d.rows;
+  window.__lbChart = d.bar_payload;
+}
+
+let lbChartUrl = "";
+async function genLbChart() {
+  if (!window.__lbChart) { toast("暂无数据可画图"); return; }
+  const status = $("#lb-chart-status");
+  const imgWrap = $("#lb-chart-img");
+  status.textContent = "⏳ 正在生成...";
+  try {
+    const resp = await fetch("/api/chart", {
+      method: "POST", headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(window.__lbChart),
+    });
+    if (!resp.ok) { const j = await resp.json().catch(() => ({})); status.textContent = "❌ " + (j.error || "生成失败"); return; }
+    const blob = await resp.blob();
+    if (lbChartUrl) URL.revokeObjectURL(lbChartUrl);
+    lbChartUrl = URL.createObjectURL(blob);
+    imgWrap.innerHTML = "";
+    const img = document.createElement("img");
+    img.src = lbChartUrl;
+    img.alt = "模型速度排行图";
+    img.onload = () => { status.textContent = "✅ 生成完成"; };
+    imgWrap.appendChild(img);
+  } catch (e) { status.textContent = "❌ " + e.message; }
+}
+
+/* ───────────────────────── 模型详情页 ───────────────────────── */
+async function loadModel() {
+  const params = new URLSearchParams(location.search);
+  const provider = params.get("provider") || "";
+  const model = params.get("model") || "";
+  if (!model) { $("#model-body").innerHTML = '<div class="empty">缺少模型参数</div>'; return; }
+  let d;
+  try {
+    d = await api(`/api/model?provider=${encodeURIComponent(provider)}&model=${encodeURIComponent(model)}`);
+  } catch (e) {
+    $("#model-body").innerHTML = `<div class="empty">${esc(e.message)}</div>`;
+    return;
+  }
+  document.title = `${d.model} · 模型评测`;
+  $("#m-title").textContent = `${d.model} 详情`;
+  $("#m-sub").textContent = `${providerLabel(d.provider)} · 共 ${d.sub_cnt} 条提交`;
+  $("#md-submissions").innerHTML = d.submissions.map((s) => {
+    const conc = (s.concurrency_levels || [1]).join("/");
+    return `<tr>
+      <td><a href="#" data-sid="${s.id}" class="s-link">#${s.id}</a></td>
+      <td>${esc(s.account)}</td>
+      <td>${esc(s.created_at)}</td>
+      <td class="num">${fmt(s.samples_ok)}/${fmt(s.samples_total)}</td>
+      <td class="num">${fmt(s.avg_ttft_ms)}</td>
+      <td class="num">${fmt(s.avg_prefill_speed)}</td>
+      <td class="num"><span class="badge-decode">${fmt(s.avg_decode_speed)}</span></td>
+      <td class="num">${fmt(s.avg_stream_decode)}</td>
+      <td class="num">${conc}</td>
+      <td class="num">${fmt(s.avg_output_tokens)}</td>
+      <td>${s.test_name ? `<span class="pill gray" title="${esc(s.test_name)}">${esc(s.test_name.slice(0, 16))}</span>` : "—"}</td>
+    </tr>`;
+  }).join("");
+
+  // 折线图：解码速度随上下文长度
+  window.__modelChart = d.line_payload;
+  window.__modelLineCsv = d.line_csv;
+  $("#m-chart-csv").value = d.line_csv;
+  $("#m-chart-img").innerHTML = '<div class="hint">点击「🎨 生成折线图」绘制</div>';
+  $("#m-chart-status").textContent = "";
+
+  // 数据统计（按长度聚合表）
+  $("#md-lengths").innerHTML = d.by_length.map((r) => `<tr>
+    <td class="num">${fmtInt(r.length)}</td>
+    <td class="num">${fmtInt(r.count)}</td>
+    <td class="num">${fmt(r.avg_prefill_speed)}</td>
+    <td class="num"><span class="badge-decode">${fmt(r.avg_decode_speed)}</span></td>
+    <td class="num">${fmt(r.avg_ttft_ms)}</td>
+  </tr>`).join("");
+}
+
+let modelChartUrl = "";
+async function genModelChart() {
+  if (!window.__modelChart) { toast("暂无数据可画图"); return; }
+  const status = $("#m-chart-status");
+  const imgWrap = $("#m-chart-img");
+  status.textContent = "⏳ 正在生成...";
+  try {
+    const resp = await fetch("/api/chart", {
+      method: "POST", headers: { "Content-Type": "application/json" },
+      body: JSON.stringify(window.__modelChart),
+    });
+    if (!resp.ok) { const j = await resp.json().catch(() => ({})); status.textContent = "❌ " + (j.error || "生成失败"); return; }
+    const blob = await resp.blob();
+    if (modelChartUrl) URL.revokeObjectURL(modelChartUrl);
+    modelChartUrl = URL.createObjectURL(blob);
+    imgWrap.innerHTML = "";
+    const img = document.createElement("img");
+    img.src = modelChartUrl;
+    img.alt = "速度随上下文长度";
+    img.onload = () => { status.textContent = "✅ 生成完成"; };
+    imgWrap.appendChild(img);
+  } catch (e) { status.textContent = "❌ " + e.message; }
+}
+
+/* ───────────────────────── 提交详情弹窗 ───────────────────────── */
+async function openSubmission(sid) {
+  const mask = $("#sub-mask");
+  if (!mask) return;
+  mask.hidden = false;
+  $("#sub-body").innerHTML = '<div class="empty">加载中…</div>';
+  let s;
+  try { s = await api(`/api/submissions/${sid}`); }
+  catch (e) { $("#sub-body").innerHTML = `<div class="empty">${esc(e.message)}</div>`; return; }
+  $("#sub-id").textContent = `#${sid}`;
+  const byLength = s.by_length || {};
+  const byConc = s.by_concurrency || {};
+  const lengthRows = Object.keys(byLength).sort((a, b) => a - b).map((L) => {
+    const bl = byLength[L];
+    return `<tr><td class="num">${L}</td><td class="num">${fmt(bl.avg_ttft_ms)}</td>
+      <td class="num">${fmt(bl.avg_prefill_speed)}</td><td class="num">${fmt(bl.avg_decode_speed)}</td>
+      <td class="num">${fmt(bl.avg_prompt_tokens)}</td><td class="num">${fmt(bl.avg_output_tokens)}</td></tr>`;
+  }).join("");
+  const concRows = Object.keys(byConc).sort((a, b) => a - b).map((C) => {
+    const bl = byConc[C];
+    return `<tr><td class="num">${C}</td><td class="num">${fmt(bl.avg_ttft_ms)}</td>
+      <td class="num">${fmt(bl.avg_prefill_speed)}</td><td class="num">${fmt(bl.avg_decode_speed)}</td>
+      <td class="num">${fmt(bl.avg_stream_decode)}</td><td class="num">${fmt(bl.avg_total_ms)}</td></tr>`;
+  }).join("");
+  const runs = s.runs || [];
+  const runsHtml = runs.length ? `<table class="mini"><thead><tr><th>#</th><th>上下文tok</th><th>提示词tok</th>
+    <th>首字ms</th><th>预填充tok/s</th><th>输出tok</th><th>解码tok/s</th><th>总耗时ms</th><th>备注</th></tr></thead><tbody>` +
+    runs.map((r, i) => {
+      const m = r.metrics || {};
+      return `<tr><td>${i + 1}</td><td class="num">${r.context_length || m.context_length || "—"}</td>
+        <td class="num">${fmt(m.prompt_tokens)}</td><td class="num">${fmt(m.ttft_ms)}</td>
+        <td class="num">${fmt(m.prefill_speed)}</td><td class="num">${fmt(m.output_tokens)}</td>
+        <td class="num">${fmt(m.decode_speed)}</td><td class="num">${fmt(m.total_ms)}</td>
+        <td>${r.error ? "❌" : "✅"}</td></tr>`;
+    }).join("") + "</tbody></table>"
+    : '<div class="hint">无采样明细（演示数据）</div>';
+
+  $("#sub-body").innerHTML = `
+    <div class="kv" style="margin-bottom:14px">
+      <div class="kv-item"><div class="kv-k">账号</div><div class="kv-v">${esc(s.account)}</div></div>
+      <div class="kv-item"><div class="kv-k">模型</div><div class="kv-v">${esc(s.model)}</div></div>
+      <div class="kv-item"><div class="kv-k">提供商</div><div class="kv-v">${esc(providerLabel(s.provider))}</div></div>
+      <div class="kv-item"><div class="kv-k">测试名称</div><div class="kv-v">${esc(s.test_name || "—")}</div></div>
+      <div class="kv-item"><div class="kv-k">提交时间</div><div class="kv-v">${esc(s.created_at)}</div></div>
+      <div class="kv-item"><div class="kv-k">来源</div><div class="kv-v">${esc(s.source_site)} #${s.source_test_id}</div></div>
+    </div>
+    <h3>📊 整体指标</h3>
+    <div class="kv" style="margin-bottom:14px">
+      <div class="kv-item"><div class="kv-k">采样(成功/总数)</div><div class="kv-v">${fmtInt(s.samples_ok)} / ${fmtInt(s.samples_total)}</div></div>
+      <div class="kv-item"><div class="kv-k">首字延迟</div><div class="kv-v">${fmt(s.avg_ttft_ms)} ms</div></div>
+      <div class="kv-item"><div class="kv-k">预填充速度</div><div class="kv-v">${fmt(s.avg_prefill_speed)} tok/s</div></div>
+      <div class="kv-item"><div class="kv-k">解码速度</div><div class="kv-v" style="color:var(--accent2)">${fmt(s.avg_decode_speed)} tok/s</div></div>
+      <div class="kv-item"><div class="kv-k">单流均解码</div><div class="kv-v">${fmt(s.avg_stream_decode)} tok/s</div></div>
+      <div class="kv-item"><div class="kv-k">平均输出</div><div class="kv-v">${fmtInt(s.avg_output_tokens)} tok</div></div>
+    </div>
+    <h3>📏 按上下文长度</h3>
+    <div class="tbl-wrap" style="margin-bottom:12px"><table class="mini"><thead><tr>
+      <th>长度tok</th><th>首字ms</th><th>预填充tok/s</th><th>解码tok/s</th><th>提示词tok</th><th>输出tok</th>
+    </tr></thead><tbody>${lengthRows || '<tr><td colspan="6" class="empty">无数据</td></tr>'}</tbody></table></div>
+    <h3>🔀 按并发数</h3>
+    <div class="tbl-wrap" style="margin-bottom:12px"><table class="mini"><thead><tr>
+      <th>并发</th><th>首字ms</th><th>预填充tok/s</th><th>解码tok/s</th><th>单流均解码tok/s</th><th>总耗时ms</th>
+    </tr></thead><tbody>${concRows || '<tr><td colspan="6" class="empty">仅单流</td></tr>'}</tbody></table></div>
+    <h3>📈 图表</h3>
+    <div class="chart-img-wrap" id="sub-chart-img"><div class="hint">点击「🎨 生成图表」</div></div>
+    <div style="margin-top:6px"><button class="btn small primary" id="sub-chart-gen">🎨 生成图表</button>
+      <span class="hint" id="sub-chart-status"></span></div>
+    <h3>📋 采样明细</h3>
+    ${runsHtml}
+  `;
+  window.__subId = sid;
+  let chartUrl = "";
+  $("#sub-chart-gen").addEventListener("click", async () => {
+    const status = $("#sub-chart-status");
+    status.textContent = "⏳ 正在生成...";
+    try {
+      const resp = await fetch(`/api/submissions/${sid}/chart`);
+      if (!resp.ok) { const j = await resp.json().catch(() => ({})); status.textContent = "❌ " + (j.error || "生成失败"); return; }
+      const blob = await resp.blob();
+      if (chartUrl) URL.revokeObjectURL(chartUrl);
+      chartUrl = URL.createObjectURL(blob);
+      const wrap = $("#sub-chart-img");
+      wrap.innerHTML = "";
+      const img = document.createElement("img");
+      img.src = chartUrl;
+      img.onload = () => { status.textContent = "✅ 生成完成"; };
+      wrap.appendChild(img);
+    } catch (e) { status.textContent = "❌ " + e.message; }
+  });
+}
+
+/* ───────────────────────── 提交管理页 ───────────────────────── */
+let subPage = 1;
+let subQuery = "";
+
+async function loadSubmissions(resetPage = false) {
+  const tb = $("#subs tbody");
+  if (!tb) return;
+  if (resetPage) subPage = 1;
+  tb.innerHTML = '<tr><td colspan="8" class="empty">加载中…</td></tr>';
+  let d;
+  try {
+    d = await api(`/api/submissions?page=${subPage}&page_size=20&q=${encodeURIComponent(subQuery)}`);
+  } catch (e) {
+    tb.innerHTML = `<tr><td colspan="8" class="empty">加载失败：${esc(e.message)}</td></tr>`;
+    return;
+  }
+  $("#sub-total").textContent = `共 ${d.total} 条`;
+  $("#sub-pg").textContent = `${d.page} / ${d.pages}`;
+  $("#sub-prev").disabled = d.page <= 1;
+  $("#sub-next").disabled = d.page >= d.pages;
+  if (!d.items.length) {
+    tb.innerHTML = '<tr><td colspan="8" class="empty">暂无提交记录。可点击右上角「🎲 生成演示数据」或从 llm-speed-tester 发送。</td></tr>';
+    return;
+  }
+  tb.innerHTML = d.items.map((s) => `<tr>
+    <td><a href="#" class="s-link" data-sid="${s.id}">#${s.id}</a></td>
+    <td>${esc(s.account)}</td>
+    <td>${esc(s.created_at)}</td>
+    <td><span class="pill">${esc(providerLabel(s.provider))}</span></td>
+    <td><a href="/model.html?provider=${encodeURIComponent(s.provider)}&model=${encodeURIComponent(s.model)}">${esc(s.model)}</a></td>
+    <td class="num">${fmt(s.samples_ok)}/${fmt(s.samples_total)}</td>
+    <td class="num">${fmt(s.avg_prefill_speed)}</td>
+    <td class="num"><span class="badge-decode">${fmt(s.avg_decode_speed)}</span></td>
+    <td class="num">${fmt(s.avg_ttft_ms)}</td>
+    <td>${s.test_name ? esc(s.test_name.slice(0, 14)) : "—"}</td>
+    <td>
+      <button class="btn small" data-view="${s.id}">查看</button>
+      <button class="btn small danger" data-del="${s.id}">删除</button>
+    </td>
+  </tr>`).join("");
+}
+
+/* ───────────────────────── 账号管理页 ───────────────────────── */
+async function loadAccounts() {
+  const tb = $("#accs tbody");
+  if (!tb) return;
+  let list;
+  try { list = await api("/api/accounts"); }
+  catch (e) { tb.innerHTML = `<tr><td colspan="5" class="empty">加载失败</td></tr>`; return; }
+  if (!list.length) {
+    tb.innerHTML = '<tr><td colspan="5" class="empty">暂无账号。当 llm-speed-tester 一键发送时账号会自动创建。</td></tr>';
+    return;
+  }
+  tb.innerHTML = list.map((a) => `<tr>
+    <td>${esc(a.name)}</td>
+    <td>${esc(a.remark || "—")}</td>
+    <td>${esc(a.created_at)}</td>
+    <td class="num"><span class="pill green">${fmtInt(a.cnt)} 条</span></td>
+    <td>
+      <button class="btn small" data-edit="${a.id}" data-name="${esc(a.name)}" data-remark="${esc(a.remark || "")}">编辑</button>
+      <button class="btn small danger" data-del="${a.id}">删除</button>
+    </td>
+  </tr>`).join("");
+}
+
+/* ───────────────────────── 事件绑定 ───────────────────────── */
+function bindIndex() {
+  $$("#lb th[data-sort]").forEach((th) => th.addEventListener("click", () => sortClick(th.dataset.sort)));
+  const gen = $("#lb-chart-gen");
+  if (gen) gen.addEventListener("click", genLbChart);
+}
+
+function bindModel() {
+  const gen = $("#m-chart-gen");
+  if (gen) gen.addEventListener("click", genModelChart);
+  const copy = $("#m-chart-copy");
+  if (copy) copy.addEventListener("click", () => {
+    const ta = $("#m-chart-csv");
+    ta.select();
+    if (navigator.clipboard && navigator.clipboard.writeText) navigator.clipboard.writeText(ta.value).then(() => toast("已复制"));
+    else { document.execCommand("copy"); toast("已复制"); }
+  });
+  const dl = $("#m-chart-dl");
+  if (dl) dl.addEventListener("click", () => {
+    if (!modelChartUrl) { toast("请先生成图表"); return; }
+    const a = document.createElement("a");
+    a.href = modelChartUrl;
+    a.download = `model_chart.png`;
+    document.body.appendChild(a); a.click(); a.remove();
+  });
+  const body = $("#md-submissions");
+  if (body) body.addEventListener("click", (e) => {
+    const link = e.target.closest(".s-link");
+    if (link) { e.preventDefault(); openSubmission(link.dataset.sid); }
+  });
+}
+
+function bindSubmissions() {
+  const seed = $("#btn-seed");
+  if (seed) seed.addEventListener("click", async () => {
+    seed.disabled = true;
+    seed.textContent = "生成中…";
+    try {
+      const r = await api("/api/seed-demo", "POST", {});
+      toast(`已生成 ${r.seeded} 条演示数据`);
+      loadSubmissions(true);
+    } catch (e) { toast("生成失败：" + e.message); }
+    seed.disabled = false;
+    seed.textContent = "🎲 生成演示数据";
+  });
+  const search = $("#sub-search");
+  if (search) {
+    search.addEventListener("keydown", (e) => { if (e.key === "Enter") { subQuery = search.value.trim(); loadSubmissions(true); } });
+    $("#btn-search").addEventListener("click", () => { subQuery = search.value.trim(); loadSubmissions(true); });
+  }
+  $("#sub-prev").addEventListener("click", () => { if (subPage > 1) { subPage--; loadSubmissions(); } });
+  $("#sub-next").addEventListener("click", () => { subPage++; loadSubmissions(); });
+  const tb = $("#subs tbody");
+  tb.addEventListener("click", (e) => {
+    const v = e.target.closest("[data-view]");
+    const d = e.target.closest("[data-del]");
+    const l = e.target.closest(".s-link");
+    if (v) { e.preventDefault(); openSubmission(v.dataset.view); }
+    if (l) { e.preventDefault(); openSubmission(l.dataset.sid); }
+    if (d) {
+      const id = d.dataset.del;
+      if (confirm(`确定删除提交 #${id}？`)) {
+        api(`/api/submissions/${id}`, "DELETE").then(() => { toast("已删除"); loadSubmissions(); });
+      }
+    }
+  });
+  const close = $("#sub-close");
+  if (close) {
+    close.addEventListener("click", () => { $("#sub-mask").hidden = true; });
+    $("#sub-mask").addEventListener("click", (e) => { if (e.target === $("#sub-mask")) $("#sub-mask").hidden = true; });
+  }
+}
+
+function bindAccounts() {
+  const add = $("#btn-add-acc");
+  if (add) add.addEventListener("click", async () => {
+    const name = $("#acc-name").value.trim();
+    const remark = $("#acc-remark").value.trim();
+    if (!name) { toast("请输入账号名称"); return; }
+    try {
+      await api("/api/accounts", "POST", { name, remark });
+      toast("账号已创建");
+      $("#acc-name").value = ""; $("#acc-remark").value = "";
+      loadAccounts();
+    } catch (e) { toast(e.message); }
+  });
+  const tb = $("#accs tbody");
+  tb.addEventListener("click", (e) => {
+    const ed = e.target.closest("[data-edit]");
+    const dl = e.target.closest("[data-del]");
+    if (ed) {
+      const name = prompt("账号名称：", ed.dataset.name);
+      if (name === null) return;
+      const remark = prompt("备注：", ed.dataset.remark) || "";
+      api(`/api/accounts/${ed.dataset.edit}`, "PUT", { name, remark }).then(() => { toast("已保存"); loadAccounts(); });
+    }
+    if (dl) {
+      const id = dl.dataset.del;
+      if (confirm("删除该账号会同时删除其下所有提交，确定？")) {
+        api(`/api/accounts/${id}`, "DELETE").then(() => { toast("已删除"); loadAccounts(); });
+      }
+    }
+  });
+}
+
+/* ───────────────────────── 初始化 ───────────────────────── */
+document.addEventListener("DOMContentLoaded", () => {
+  navActive();
+  const page = location.pathname.split("/").pop() || "index.html";
+  if (page === "index.html" || page === "") {
+    loadStats();
+    loadLeaderboard();
+    bindIndex();
+  } else if (page === "model.html") {
+    loadModel();
+    bindModel();
+  } else if (page === "submissions.html") {
+    loadSubmissions(true);
+    bindSubmissions();
+  } else if (page === "accounts.html") {
+    loadAccounts();
+    bindAccounts();
+  }
+  document.addEventListener("keydown", (e) => {
+    if (e.key === "Escape") { const m = $("#sub-mask"); if (m) m.hidden = true; }
+  });
+});
diff --git a/static/model.html b/static/model.html
new file mode 100644
index 0000000..78ed663
--- /dev/null
+++ b/static/model.html
@@ -0,0 +1,69 @@
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+<meta charset="UTF-8">
+<meta name="viewport" content="width=device-width, initial-scale=1.0">
+<title>模型详情 · 模型评测站</title>
+<link rel="stylesheet" href="/css/style.css">
+</head>
+<body>
+<div class="topbar">
+  <div class="brand"><span class="logo">⚡</span> 模型评测站</div>
+  <nav class="nav">
+    <a href="/index.html">🏆 速度排行</a>
+    <a href="/submissions.html">📥 提交管理</a>
+    <a href="/accounts.html">👥 账号</a>
+  </nav>
+  <div class="spacer"></div>
+  <a href="/index.html" class="hint">← 返回排行</a>
+</div>
+
+<div class="container">
+  <div class="panel">
+    <h2 id="m-title">模型详情</h2>
+    <div class="hint" id="m-sub"></div>
+  </div>
+
+  <div class="panel">
+    <h2>📈 解码速度随上下文长度变化（全部提交聚合）</h2>
+    <div class="chart-toolbar">
+      <button class="btn small primary" id="m-chart-gen">🎨 生成折线图</button>
+      <span class="hint" id="m-chart-status"></span>
+    </div>
+    <div class="chart-img-wrap" id="m-chart-img">
+      <div class="hint">左轴=预填充（虚线），右轴=解码（实线）</div>
+    </div>
+    <div class="chart-data-head" style="margin-top:8px;color:var(--muted);font-size:12px">📋 画图数据（CSV）</div>
+    <textarea class="chart-csv" id="m-chart-csv" readonly spellcheck="false"></textarea>
+    <div class="form-row" style="margin-top:8px">
+      <button class="btn small" id="m-chart-copy">📋 复制数据</button>
+      <button class="btn small" id="m-chart-dl">⬇ 下载 PNG</button>
+    </div>
+  </div>
+
+  <div class="panel">
+    <h2>📏 按上下文长度聚合</h2>
+    <div class="tbl-wrap">
+      <table><thead><tr>
+        <th class="num">长度 tok</th><th class="num">提交数</th>
+        <th class="num">平均预填充 tok/s</th><th class="num">平均解码 tok/s</th><th class="num">首字 ms</th>
+      </tr></thead><tbody id="md-lengths"><tr><td colspan="5" class="empty">—</td></tr></tbody></table>
+    </div>
+  </div>
+
+  <div class="panel">
+    <h2>📥 该模型的所有评测提交（按账号）</h2>
+    <div class="tbl-wrap">
+      <table><thead><tr>
+        <th>提交</th><th>账号</th><th>时间</th><th class="num">采样</th><th class="num">首字 ms</th>
+        <th class="num">预填充 tok/s</th><th class="num">解码 tok/s</th><th class="num">单流均解码</th>
+        <th class="num">并发</th><th class="num">输出 tok</th><th>测试名称</th>
+      </tr></thead><tbody id="md-submissions"><tr><td colspan="11" class="empty">加载中…</td></tr></tbody></table>
+    </div>
+  </div>
+</div>
+
+<div id="toast"></div>
+<script src="/js/app.js"></script>
+</body>
+</html>
diff --git a/static/submissions.html b/static/submissions.html
new file mode 100644
index 0000000..717c037
--- /dev/null
+++ b/static/submissions.html
@@ -0,0 +1,57 @@
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+<meta charset="UTF-8">
+<meta name="viewport" content="width=device-width, initial-scale=1.0">
+<title>提交管理 · 模型评测站</title>
+<link rel="stylesheet" href="/css/style.css">
+</head>
+<body>
+<div class="topbar">
+  <div class="brand"><span class="logo">⚡</span> 模型评测站</div>
+  <nav class="nav">
+    <a href="/index.html">🏆 速度排行</a>
+    <a href="/submissions.html">📥 提交管理</a>
+    <a href="/accounts.html">👥 账号</a>
+  </nav>
+  <div class="spacer"></div>
+  <button class="btn primary" id="btn-seed">🎲 生成演示数据</button>
+</div>
+
+<div class="container">
+  <div class="panel">
+    <h2>📥 评测提交 <span class="hint" id="sub-total"></span></h2>
+    <div class="notice green">💡 这些数据来自 llm-speed-tester 的「📤 发送到评测站」按钮（按账号归属）。</div>
+    <div class="form-row">
+      <input type="search" id="sub-search" placeholder="搜索：模型 / 账号 / 测试名称…" style="width:280px">
+      <button class="btn" id="btn-search">搜索</button>
+      <span class="hint">支持分页查看、详情、删除</span>
+    </div>
+    <div class="tbl-wrap">
+      <table id="subs"><thead><tr>
+        <th>提交</th><th>账号</th><th>时间</th><th>提供商</th><th>模型</th>
+        <th class="num">采样</th><th class="num">预填充 tok/s</th><th class="num">解码 tok/s</th>
+        <th class="num">首字 ms</th><th>测试名称</th><th>操作</th>
+      </tr></thead><tbody><tr><td colspan="11" class="empty">加载中…</td></tr></tbody></table>
+    </div>
+    <div class="pagination">
+      <button class="btn small" id="sub-prev">← 上一页</button>
+      <span class="info" id="sub-pg"></span>
+      <button class="btn small" id="sub-next">下一页 →</button>
+    </div>
+  </div>
+</div>
+
+<!-- 提交详情弹窗 -->
+<div class="mask" id="sub-mask" hidden>
+  <div class="modal">
+    <div class="modal-head"><h3>评测提交详情 <span id="sub-id"></span></h3>
+      <button class="btn small" id="sub-close">✕ 关闭</button></div>
+    <div class="modal-body" id="sub-body"></div>
+  </div>
+</div>
+
+<div id="toast"></div>
+<script src="/js/app.js"></script>
+</body>
+</html>