docs: full Aurelio infrastructure audit + plan for 17/07

Complete audit of CT 208, host ASUS, host Gigabyte, CT 217. Includes: all services, ports, gbrain health/issues, bot status, resolved problems, and prioritized pending tasks for tomorrow.
2026-07-16 23:17:35 +01:00 · 2026-07-16 23:17:35 +01:00 · c068bf53dd
commit c068bf53dd
parent 6eb893b7a0
1 changed files with 200 additions and 0 deletions
--- a/.aurelio/plans/2026-07-16_aurelio-full-audit.md
+++ b/.aurelio/plans/2026-07-16_aurelio-full-audit.md
@ -0,0 +1,200 @@
+# Aurelio — Auditoria Completa (16/07/2026)
+
+> Plano para amanhã: estado actual, issues conhecidas, próximos passos.
+
+---
+
+## 1. Estrutura Geral
+
+### CT 208 — pf-forja-do-conhecimento (192.168.0.125)
+**Disco:** 49G, 37G used (78%) — cuidado, pouco espaço.
+**SO:** Debian 12 (LXC), PostgreSQL 16 no Docker (`172.17.0.1:5433`)
+
+#### Serviços Core
+
+| Serviço | Porta | Estado | Descrição |
+|---------|-------|--------|-----------|
+| `pf-gbrain.service` | `:18001` | ✅ active | GBrain v0.42.58 — PostgreSQL, 1362 páginas |
+| `gbrain-rest-proxy.service` | `:18002` | ✅ active | Bridge REST→MCP p/ bot-rs |
+| `aurelio-gabinete-rs.service` | `:18000` | ✅ active | Gabinete Rust (API central) |
+| `aurelio-bot-rs.service` | — | ✅ active | @Aurélio Omnisciente (Telegram) |
+| `aurelio-engineering-gateway.service` | `:8090` | ✅ active | MCP Gateway federado |
+| `aurelio-native-engine-gateway.service` | — | ✅ active | Dirac wrapper MCP gateway |
+| `aurelio-web.service` | — | ✅ active | Olhos-de-Orpheu MCP console |
+| `aurelio-webhook.service` | — | ✅ active | Forgejo→Telegram notifier |
+| `brain-sync.service` | `:8095` | ✅ active | Sync local (pushed=0 sempre) |
+| `lifestream-userbot.service` | — | ✅ active | Telethon Saved Messages monitor |
+| `pf-identity.service` | `:9091` | ✅ active | CA + Identity Management |
+| `pf-hermes-agent-208.service` | `:18300` | ✅ active | Hermes MCP server |
+
+#### MCP Servers (CT 208)
+
+| Serviço | Porta | Descrição |
+|---------|-------|-----------|
+| `mcp-sourcing-server.service` | `:8081` | Sourcing MCP |
+| `mcp-eda-server.service` | `:8082` | EDA MCP |
+| `mcp-knowledge-server.service` | `:8083` | Knowledge MCP |
+| `codebase-memory-mcp.service` | `:8084` | Codebase Memory MCP |
+| `mcp-civil-server.service` | — | Civil MCP |
+| `mcp-artist-server.service` | — | Artist MCP |
+| `mcp-mechanic-server.service` | — | Mechanic MCP |
+
+#### Kimi Swarm Agents
+
+| Serviço | Descrição |
+|---------|-----------|
+| `pf-kimi-swarm-orchestrator-208.service` | Orchestrator |
+| `pf-log-processor-agent-208.service` | Log processor |
+| `pf-ota-validator-agent-208.service` | OTA validator |
+| `pf-regression-detector-agent-208.service` | Regression detector |
+| `pf-report-synthesizer-agent-208.service` | Report synthesizer |
+| `pf-swarm-dashboard-208.service` | Swarm dashboard |
+
+#### Outros
+- `pf-forgejo-scribe-208.service` — Forgejo event scribe
+- `pf-github-scribe-208.service` — GitHub event scribe
+- `pf-grafica-208.service` — Gráfica print shop
+- `pf-olhos-208.service` — Olhos de Orpheu
+- `pf-savearth-workspace-208.service` — Savearth workspace
+- `savearth-mcp-208.service` — Savearth MCP
+
+### Host ASUS (192.168.0.38)
+
+| Serviço | Porta | Descrição |
+|---------|-------|-----------|
+| `aurelio-fabricd.service` | `:8081` | Fabric Control Plane (model gateway) |
+| `aurelio-orchestrator.service` | `:8200` | MicroVM Orchestrator |
+| `aurelio-asus-server-lab-node.service` | `:8002` | Hardware Lab Node |
+| `lab-gateway.service` | — | Lab gateway |
+| `cloudflared.service` | — | Cloudflare Tunnel (pf-tunnel-v2) |
+| `nginx.service` | `:80` | Reverse proxy |
+| `iptables-restore.service` | — | ✅ DNAT :18001 → CT 208 |
+| `pf-hardware-bridge.service` | — | Hardware API bridge |
+| `pf-watchdog.service` | — | Datacenter watchdog |
+
+### Host Gigabyte (192.168.0.104)
+- Docker media stack (Jellyfin, Sonarr, Radarr, etc.)
+- `signal-api` container (:18088)
+- Cloud emulators (MinIO :9000, LocalStack :4566, Azurite)
+
+### CT 217 — pf-hermes (192.168.0.27)
+- `hermes-gateway.service` — gateway 24/7 (Telegram/Signal/Matrix)
+- GBrain MCP configurado com token OAuth
+
+---
+
+## 2. GBrain — Estado e Issues
+
+### Métricas
+- **Páginas:** 1.362
+- **Embed coverage:** 12.7% (3.636/4.163 chunks não embedded)
+- **Brain Score:** 14/100 🚨
+- **Engine:** PostgreSQL (via binário)
+- **Versão:** 0.42.58.0
+
+### Issues Críticas
+
+| Issue | Severidade | Fix |
+|-------|-----------|-----|
+| Brain score 14/100 | ❌ fail | Embed + extrair links |
+| 1362/1362 páginas sem edges extraídos | ⚠️ warn | `gbrain extract --stale` |
+| 1312 páginas sem CR ladder | ⚠️ warn | `gbrain reindex --markdown` |
+| Schema pack `gbrain-base-v2` não encontrado | ⚠️ warn | `gbrain schema active` |
+| Subagent sem Anthropic key | ⚠️ warn | `gbrain config set agent.use_gateway_loop true` |
+| Embed coverage 12.7% | ⚠️ low | `gbrain embed --all` |
+
+### Fluxo de dados
+```
+Saved Messages (Telegram)
+    → Lifestream userbot (Telethon)
+    → SQLite DB (/opt/aurelio-lifestream/data/lifestream.db)
+    → Cron 03:00 → /opt/aurelio-lifestream/lifestream_to_brain.py
+    → GBrain (via REST→MCP proxy)
+```
+
+---
+
+## 3. Bot Aurélio — Estado
+
+| Comando | Estado | Notas |
+|---------|--------|-------|
+| `/brain <query>` | ✅ Fix | Proxy REST→MCP em :18002 |
+| `/think <q>` | ✅ | Usa fabricd como gateway |
+| `/status` | ✅ | Gateway, GBrain, SearXNG |
+| `/today` | ⏳ | Retorna "em desenvolvimento" |
+| `/onthisday` | ⏳ | Retorna "em desenvolvimento" |
+| `/sync` | ❓ | Não testado |
+| Mensagens livres | ✅ | Capturadas pelo Lifestream |
+
+### Config
+- **Gateway:** `http://192.168.0.38:8081` (fabricd)
+- **GBrain:** `http://127.0.0.1:18002` (proxy)
+- **Modelo:** `nvidia/nemotron-3-nano-30b-a3b`
+
+---
+
+## 4. O que ficou resolvido hoje
+
+| Problema | Antes | Depois |
+|----------|-------|--------|
+| PGLite lock timeout | ❌ Bun gbrain no host | ✅ Binário PostgreSQL CT 208 |
+| Duas instâncias gbrain | ❌ Host bun + CT 208 binary | ✅ Só CT 208 binary |
+| Bot não respondia | ❌ Host bot + CT 208 bot | ✅ Só CT 208 bot |
+| Gateway :9090 inexistente | ❌ Missing | ✅ Redireccionado p/ fabricd :8081 |
+| `/brain` dava 404 | ❌ gbrain v0.42 sem REST | ✅ Proxy REST→MCP em :18002 |
+| Hermes MCP sem auth | ❌ 401 no gbrain | ✅ Token OAuth configurado |
+
+---
+
+## 5. Pendências para Amanhã (Prioridades)
+
+### 🔴 Alta
+1. **Brain score 14/100** — Correr `gbrain extract --stale` + `gbrain embed --all` para construir edges e embeddings
+2. **Disco CT 208 a 78%** — Limpar sessions antigas em `.aurelio/brain/` (centenas de sessions opencode/kimi)
+3. **Schema pack** — `gbrain schema active` para resolver `gbrain-base-v2`
+
+### 🟡 Média
+4. **Lifestream sync** — Verificar se `lifestream_to_brain.py` funciona (cron 03:00). Testar sync manual.
+5. **Brain-sync** — Está a funcionar mas sempre `pushed=0`. Decidir se é necessário ou legacy.
+6. **REST API endpoints** — O gbrain v0.42 não tem REST API. O proxy cobre `/api/search`. Faltam outros endpoints?
+
+### 🟢 Baixa
+7. **`/today` e `/onthisday`** — Implementar chronicle no bot-rs
+8. **Forgejo push** — Token expirado; renovar para push automático
+9. **GBrain REST proxy** — Mover de `/opt/gbrain-rest-proxy.py` para `/opt/pf-services-208/aurelio-gbrain/scripts/`
+10. **Monitorização** — Adicionar watchdog para o bot-rs + gbrain
+
+---
+
+## 6. Diagrama de Rede
+
+```
+                    ┌─────────────────────┐
+                    │   Cloudflare Tunnel  │
+                    │  pf-tunnel-v2 (ASUS) │
+                    └──────────┬──────────┘
+                               │ :80/:443
+                    ┌──────────▼──────────┐
+                    │   ASUS (192.168.0.38)│
+                    │  ┌─────┬─────┬────┐ │
+                    │  │fabricd│orch.│nginx│ │
+                    │  │:8081 │:8200│:80  │ │
+                    │  └──┬──┴─────┴────┘ │
+                    │  DNAT :18001→CT 208 │
+                    └───────┬─────────────┘
+                            │
+        ┌───────────────────┼───────────────────┐
+        │                   │                   │
+┌───────▼──────┐   ┌───────▼──────┐   ┌───────▼──────┐
+│ CT 208       │   │ CT 217      │   │ Gigabyte     │
+│ pf-forja     │   │ pf-hermes   │   │ (media stack)│
+│              │   │              │   │              │
+│ gbrain :18001│   │ Hermes gw.  │   │ Docker:      │
+│ proxy :18002 │   │ Telegram/   │   │ Jellyfin     │
+│ gabinete     │   │ Signal/     │   │ Sonarr/Radarr│
+│  :18000      │   │ Matrix bots │   │ signal-api   │
+│ bot-rs       │   │              │   │ MinIO/Local  │
+│ lifestream   │   │              │   │   Stack      │
+│ MCP servers  │   │              │   │              │
+└──────────────┘   └──────────────┘   └──────────────┘
+```