[HARD_FACTS_BEGIN]
=== AI-LAB RUNTIME (HARD FACTS) ===

{
  "schema_version": "1.0",
  "runtime": {
    "mode": "plan",
    "version": "1.1.0",
    "timestamp": "2026-05-16T18:00:00Z",
    "uptime_hours": 2.2
  },
  "gpu_nodes": [
    {
      "name": "rx9070-node",
      "host": "192.168.1.50",
      "friendly_name": "RX9070",
      "online": true,
      "vram_gb": 16,
      "latency_ms": 2,
      "models": [
        {
          "id": "qwen2.5-coder-14b-instruct",
          "ctx": 32768,
          "skills": ["coding", "debugging", "refactor", "testing"]
        },
        {
          "id": "llama-3.1-8b-instruct",
          "ctx": 128000,
          "skills": ["fast", "general", "chat", "summarisation"]
        },
        {
          "id": "text-embedding-nomic-embed-text-v1.5",
          "ctx": 8192,
          "skills": ["embeddings", "semantic-search", "rag"]
        }
      ]
    },
    {
      "name": "rx7900xt-node",
      "host": "192.168.1.60",
      "friendly_name": "RX7900XT",
      "online": false,
      "vram_gb": 20,
      "latency_ms": null,
      "models": []
    }
  ],
  "system": {
    "hostname": "ubuntu-ialab",
    "ram": {"total_gib": 7.2, "used_gib": 2.5, "available_gib": 4.7},
    "disk": {"total_g": 97, "used_g": 70, "pct": 76},
    "docker_containers": 16,
    "load": [0.28, 0.39, 0.43],
    "uptime_hours": 2.2
  },
  "services": [
    {"name": "ailab-router", "active": true, "running": true},
    {"name": "ailab-docs", "active": true, "running": true},
    {"name": "ailab-gateway", "active": true, "running": true},
    {"name": "ailab-heartbeat", "active": true, "running": true},
    {"name": "ailab-live-api", "active": true, "running": true},
    {"name": "ailab-live-state", "active": true, "running": true},
    {"name": "ailab-runner", "active": true, "running": true},
    {"name": "ailab-traefik", "active": true, "running": false}
  ],
  "docker": {
    "total": 16,
    "main": [
      "traefik", "open-webui", "qdrant",  # LEGACY: ollama removed
      "cadvisor", "node-exporter", "promtail", "portainer"
    ],
    "nginx_sites": [
      "musquera-web", "agithome", "albertskills",
      "agitservices", "albertskills-amd-multi"
    ]
  },
  "routing": {
    "total_events": 107,
    "cognitive_snapshots": 106,
    "model_performance": {
      "qwen2.5-coder-14b-instruct": {
        "requests": 7,
        "success_rate": 1.0,
        "performance_index": 97,
        "failover_rate": 0.0
      }
    }
  },
  "health": {
    "watchdog": {"ok": 6, "total": 6, "status": "healthy"},
    "active_sessions": 1
  },
  "pending_implementations": [
    "routing_confidence",
    "latency_per_request",
    "puppet_ansible",
    "gateway_api_write",
    "rx7900xt_diagnosis",
    "ci_cd_automation",
    "auto_scaling"
  ],
  "sites": [
    {"url": "ai-lab.labrazahome.com", "access": "public", "tech": "Cloudflare Pages + Astro"},
    {"url": "blog-ai-lab.labrazahome.com", "access": "private", "tech": "Cloudflare Tunnel + Traefik"}
  ]
}

[HARD_FACTS_END]

=== DETALLE (texto) ===
GPU NODES:
  🟢 rx9070-node → 192.168.1.50 (ONLINE, 16GB VRAM, 2ms)
     Models:
       · qwen2.5-coder-14b-instruct (32768 ctx, coding/debugging)
       · llama-3.1-8b-instruct (128000 ctx, fast/general)
       · text-embedding-nomic-embed-text-v1.5 (8192 ctx, embeddings)
  🔴 rx7900xt-node → 192.168.1.60 (OFFLINE, 20GB VRAM)
     No models available (node offline)
SYSTEM RESOURCES:
  · RAM: 7,2Gi total / 2,5Gi used / 4,7Gi available
  · Disk: 97G total / 70G used (76%)
  · Uptime: 2 hours, 12 minutes
  · Load: 0.28, 0.39, 0.43
  · Docker: 16 containers running
...

Request → router → LM Studio → response
                          ↓
                   jsonl append (síncrono, inmediato)
                          ↓
                   Background embedder (cola asíncrona)
                          ↓
                   qdrant upsert (batch cada 10s o 20 eventos)

try:
    from runtime.memory.qdrant_store import store_embedding
    store_embedding(collection="routing_history", payload=data, embedding=emb)
except ImportError:
    pass

{
  "query": "¿Qué pasó la última vez que RX7900XT falló?",
  "collection": "incidents",
  "limit": 5,
  "threshold": 0.75
}

{
  "results": [
    {
      "score": 0.89,
      "payload": {
        "schema_version": "1.0",
        "event_type": "watchdog_failure",
        "timestamp": "2026-05-14T12:30:00Z",
        "severity": "high",
        "message": "RX7900XT timeout en health check",
        "node": "rx7900xt-node",
        "resolved": false
      }
    }
  ]
}

┌─────────────────────────────────────┐
│  🧠 Memoria Operacional Cognitiva   │
├─────────────────────────────────────┤
│ 🔍 [Buscar en memoria...        ]  │
├─────────────────────────────────────┤
│ ┌─ Episodios Recientes ───────────┐│
│ │ 🟢 12:30 - routing: qwen2.5... ││
│ │ 🟢 12:28 - cognitive snapshot  ││
│ │ 🔴 12:00 - watchdog check pass ││
│ └──────────────────────────────────┘│
│ ┌─ Incidentes Similares ──────────┐│
│ │ 📌 Score: 0.89                  ││
│ │ "RX7900XT timeout en health..." ││
│ │ 📌 Score: 0.76                  ││
│ │ "Failover a llama3.1 tras..."   ││
│ └──────────────────────────────────┘│
└─────────────────────────────────────┘