feat(p46): Agent 編排矩陣新頁 — OpenClaw/Hermes/NemoTron/EA × Ollama × Gemini × MCP × RAG
All checks were successful
CD Pipeline / deploy (push) Successful in 2m30s

統帥要求:「好好把 OpenClaw/Hermes/NemoTron/ElephantAlpha + Ollama 多模型
+ 外部付費 Gemini + 內外 MCP + RAG 組合發揮出 AI 自動化新境界」

新頁面 /observability/agent_orchestration 一頁式呈現 4 Agent × 5 維度全景:

J-1: caller 自動分組
- OpenClaw: openclaw_qa/daily/meta/monthly/weekly/bot_main/bot_gemini/bot_nim
            + sales_copy + code_review_openclaw
- Hermes: hermes_analyst + hermes_intent + code_review_hermes
- NemoTron: nemotron_dispatch
- ElephantAlpha: ea_engine + code_review_elephant

J-2/3: 跨表 SQL JOIN(ai_calls × mcp_calls × rag_query_log)
每個 agent 顯示:
- 24h 呼叫 + Token + 成本
- 本地 Ollama 比例(細分 GCP-A/GCP-B/111)
- 付費 LLM 比例(細分 Gemini / 其他)
- MCP 編排率(透過 request_id 跨表 JOIN mcp_calls)
- RAG 命中率
- 錯誤率 + 平均耗時
- MCP server × caller 工作量明細

自動編排建議(5 條 rule-based):
1. 付費比例 > 50% 且 ollama < 20% → 改 Hermes-first 短路
2. 錯誤率 > 10% → 觸發 Code Review Pipeline
3. MCP 編排率 < 5% 但 calls > 50 → 擴大 MCP omnisearch/firecrawl
4. RAG 命中率 ≥ 40% → 推 Telegram 收 feedback 強化 promotion gate
5. 111 fallback 比例 > 20% → GCP 兩台異常,查 host_health AIOps

J-4: 入口
- sidebar AI 觀測 group 加「Agent 編排矩陣」(07b)
- /observability/overview 入口卡升級為 7 項,Agent 編排矩陣放第一

整體 KPI 卡片:
- 總呼叫 / 本地 Ollama 比例 / 付費 LLM 成本 / RAG 命中率
- 「組合發揮」一目瞭然

8 表跨 JOIN:ai_calls × mcp_calls × rag_query_log × ai_insights ×
learning_episodes × incidents × heal_logs × host_health_probes

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
OoO
2026-05-04 19:38:10 +08:00
parent 849e189b60
commit 347efb8ea1
4 changed files with 526 additions and 2 deletions

View File

@@ -0,0 +1,257 @@
{% extends "ewoooc_base.html" %}
{% block title %}Agent 編排矩陣{% endblock %}
{% block ewooo_content %}
<div class="container-fluid mt-3">
<h2 class="mb-3"><i class="fas fa-network-wired me-2"></i>Agent 編排矩陣
<small class="text-muted">{{ hours }}h 內 4 Agent × Ollama × Gemini × MCP × RAG 協作全景</small>
</h2>
{% if error %}
<div class="alert alert-warning"><strong><i class="fas fa-exclamation-triangle me-1"></i></strong> {{ error }}</div>
{% endif %}
<!-- 時間範圍 -->
<form method="get" class="row g-2 mb-3">
<div class="col-auto">
<select name="hours" class="form-select form-select-sm" onchange="this.form.submit()">
{% for h in [1, 6, 24, 72, 168] %}
<option value="{{ h }}" {% if hours == h %}selected{% endif %}>
{% if h < 24 %}過去 {{ h }} 小時{% else %}過去 {{ (h//24) }} {% endif %}
</option>
{% endfor %}
</select>
</div>
</form>
<!-- 整體 KPI -->
{% if overall %}
<div class="row g-3 mb-3">
<div class="col-lg-3 col-md-6">
<div class="card h-100">
<div class="card-body">
<small class="text-muted d-block"><i class="fas fa-network-wired me-1"></i>總呼叫</small>
<h3 class="mb-0">{{ "{:,}".format(overall.total_calls) }}</h3>
<small class="text-muted">Token{{ "{:,}".format(overall.total_tokens) }}</small>
</div>
</div>
</div>
<div class="col-lg-3 col-md-6">
<div class="card h-100" style="border-left: 4px solid #198754;">
<div class="card-body">
<small class="text-muted d-block"><i class="fas fa-server me-1"></i>本地 Ollama 比例</small>
<h3 class="mb-0 text-success">{{ "%.0f"|format(overall.local_pct) }}<small>%</small></h3>
<small class="text-muted">{{ "{:,}".format(overall.local_calls) }} 呼叫</small>
</div>
</div>
</div>
<div class="col-lg-3 col-md-6">
<div class="card h-100" style="border-left: 4px solid #ffc107;">
<div class="card-body">
<small class="text-muted d-block"><i class="fas fa-coins me-1"></i>付費 LLM 成本</small>
<h3 class="mb-0">${{ "%.2f"|format(overall.total_cost) }}</h3>
<small class="text-muted">{{ "{:,}".format(overall.paid_calls) }} 付費呼叫({{ "%.0f"|format(overall.paid_pct) }}%</small>
</div>
</div>
</div>
<div class="col-lg-3 col-md-6">
<div class="card h-100" style="border-left: 4px solid #6f42c1;">
<div class="card-body">
<small class="text-muted d-block"><i class="fas fa-magnifying-glass-chart me-1"></i>RAG 命中率</small>
<h3 class="mb-0" style="color: #6f42c1;">{{ "%.0f"|format(overall.rag_rate) }}<small>%</small></h3>
<small class="text-muted">{{ "{:,}".format(overall.rag_hits) }} hits</small>
</div>
</div>
</div>
</div>
{% endif %}
<!-- 4 Agent 矩陣 -->
<div class="card mb-3">
<div class="card-header">
<strong><i class="fas fa-th me-2"></i>4 Agent × LLM × MCP × RAG 矩陣</strong>
<small class="text-muted">資料來源ai_calls × mcp_calls × rag_query_logcaller 自動分組)</small>
</div>
<div class="card-body p-0">
<table class="table mb-0" style="font-size: 0.92em;">
<thead class="table-light">
<tr>
<th>Agent</th>
<th class="text-end">呼叫</th>
<th class="text-end">成本</th>
<th class="text-end">本地 Ollama</th>
<th class="text-end">付費 LLM</th>
<th class="text-end">MCP 編排</th>
<th class="text-end">RAG 命中</th>
<th class="text-end">錯誤率</th>
<th class="text-end">耗時</th>
</tr>
</thead>
<tbody>
{% for ag in agent_matrix %}
<tr>
<td>
<strong>{{ ag.label }}</strong>
<small class="d-block text-muted">{{ ag.desc }}</small>
</td>
<td class="text-end">
{% if ag.calls > 0 %}
<strong>{{ "{:,}".format(ag.calls) }}</strong>
<small class="d-block text-muted">{{ "{:,}".format(ag.tokens) }} tk</small>
{% else %}
<small class="text-muted"></small>
{% endif %}
</td>
<td class="text-end">
{% if ag.calls > 0 %}
${{ "%.2f"|format(ag.cost) }}
{% else %}
<small class="text-muted"></small>
{% endif %}
</td>
<td class="text-end">
{% if ag.calls > 0 %}
<strong class="text-success">{{ "%.0f"|format(ag.ollama_pct) }}%</strong>
<small class="d-block text-muted">
GCP-A {{ ag.ollama_gcp_a }} ·
GCP-B {{ ag.ollama_gcp_b }} ·
111 {{ ag.ollama_111 }}
</small>
{% else %}
<small class="text-muted"></small>
{% endif %}
</td>
<td class="text-end">
{% if ag.calls > 0 %}
<strong class="{% if ag.paid_pct > 50 %}text-danger{% elif ag.paid_pct > 20 %}text-warning{% endif %}">
{{ "%.0f"|format(ag.paid_pct) }}%
</strong>
<small class="d-block text-muted">
Gemini {{ ag.gemini }}{% if ag.other_paid %} · 其他 {{ ag.other_paid }}{% endif %}
</small>
{% else %}
<small class="text-muted"></small>
{% endif %}
</td>
<td class="text-end">
{% if ag.calls > 0 %}
<strong class="{% if ag.mcp_rate >= 30 %}text-info{% elif ag.mcp_rate >= 10 %}text-warning{% else %}text-muted{% endif %}">
{{ "%.1f"|format(ag.mcp_rate) }}%
</strong>
<small class="d-block text-muted">{{ ag.mcp_calls }} request_id</small>
{% else %}
<small class="text-muted"></small>
{% endif %}
</td>
<td class="text-end">
{% if ag.calls > 0 %}
<strong style="color: #6f42c1;">{{ "%.1f"|format(ag.rag_rate) }}%</strong>
<small class="d-block text-muted">{{ ag.rag_hits }} hits</small>
{% else %}
<small class="text-muted"></small>
{% endif %}
</td>
<td class="text-end">
{% if ag.calls > 0 %}
<strong class="{% if ag.error_rate >= 15 %}text-danger{% elif ag.error_rate >= 5 %}text-warning{% else %}text-success{% endif %}">
{{ "%.1f"|format(ag.error_rate) }}%
</strong>
<small class="d-block text-muted">{{ ag.errors }} 次</small>
{% else %}
<small class="text-muted"></small>
{% endif %}
</td>
<td class="text-end">
{% if ag.calls > 0 %}{{ ag.avg_ms }} ms{% else %}<small class="text-muted"></small>{% endif %}
</td>
</tr>
{% endfor %}
</tbody>
</table>
</div>
<div class="card-footer small text-muted">
<i class="fas fa-info-circle me-1"></i>
<strong>本地 Ollama</strong> = GCP-A + GCP-B + 111 三主機級聯(免費);
<strong>付費 LLM</strong> = Gemini / Claude / NIM / OpenRouter
<strong>MCP 編排率</strong> = caller 透過 ai_calls.request_id 串接到 mcp_calls 的比例;
<strong>RAG 命中</strong> = ai_calls.rag_hit=true 的 caller-level 比例。
</div>
</div>
<!-- 自動編排建議 -->
{% if recommendations %}
<div class="card mb-3" style="border-left: 4px solid #6f42c1;">
<div class="card-header bg-light">
<strong><i class="fas fa-lightbulb me-2"></i>編排策略自動建議</strong>
<small class="text-muted">— rule-based 規則引擎5 條判斷</small>
</div>
<div class="card-body p-2">
<ul class="list-unstyled mb-0">
{% for r in recommendations %}
<li class="mb-2 p-2" style="background: #fafafa; border-radius: 6px;
{% if r.severity == 'high' %}border-left: 3px solid #dc3545;
{% elif r.severity == 'med' %}border-left: 3px solid #ffc107;
{% else %}border-left: 3px solid #0dcaf0;{% endif %}">
<span class="badge {% if r.severity == 'high' %}bg-danger{% elif r.severity == 'med' %}bg-warning text-dark{% else %}bg-info text-dark{% endif %} me-1">{{ r.severity|upper }}</span>
<strong>{{ r.agent }}</strong>
<div class="small mt-1">
<i class="fas fa-search me-1"></i><strong>發現:</strong>{{ r.finding }}
</div>
<div class="small text-muted">
<i class="fas fa-arrow-right me-1"></i><strong>建議:</strong>{{ r.suggestion }}
</div>
</li>
{% endfor %}
</ul>
</div>
</div>
{% endif %}
<!-- MCP server × caller 細節 -->
{% if mcp_matrix %}
<div class="card mb-3">
<div class="card-header">
<strong><i class="fas fa-bolt me-2"></i>MCP server × caller 工作量明細</strong>
<small class="text-muted">資料來源mcp_calls過去 {{ hours }}h前 30 筆)</small>
</div>
<div class="card-body p-0">
<table class="table table-sm mb-0">
<thead class="table-light">
<tr>
<th>MCP Server</th>
<th>呼叫端 (caller)</th>
<th class="text-end">tool 呼叫</th>
<th class="text-end">cache 命中</th>
<th class="text-end">cache 率</th>
<th class="text-end">成本</th>
</tr>
</thead>
<tbody>
{% for m in mcp_matrix %}
<tr>
<td><code>{{ m.server }}</code></td>
<td><code>{{ m.caller }}</code></td>
<td class="text-end">{{ "{:,}".format(m.calls) }}</td>
<td class="text-end">{{ m.cache_hits }}</td>
<td class="text-end">
<span class="{% if m.cache_rate >= 50 %}text-success{% elif m.cache_rate >= 20 %}text-warning{% endif %}">
{{ "%.0f"|format(m.cache_rate) }}%
</span>
</td>
<td class="text-end">${{ "%.4f"|format(m.cost) }}</td>
</tr>
{% endfor %}
</tbody>
</table>
</div>
</div>
{% endif %}
<p class="text-muted mt-3"><small>
<i class="fas fa-robot me-1"></i>Operation Ollama-First v5.0 / Phase 46 — Agent 編排矩陣
8 表跨 JOINai_calls × mcp_calls × rag_query_log × ai_insights × learning_episodes
× incidents × heal_logs × host_health_probes
</small></p>
</div>
{% endblock %}