ชิ้นส่วนหน้าเว็บ (Components)

# one 27B model resident + embeddings — the golden rule of 48 GB
NAME                    ID              SIZE      PROCESSOR    UNTIL
turbo-fable:27b-nvfp4   a1b2c3d4e5f6    23 GB     100% GPU     23 hours from now
bge-m3:latest           0f9e8d7c6b5a    1.1 GB    100% GPU     3 hours from now
# 23 GB + 1.1 GB resident, ~16 GB headroom, swap 0 (docs/hardware.md)

NAME                  SERVICE       STATUS         PORTS
bomllm-litellm-1      litellm       Up 26 days     127.0.0.1:4000->4000/tcp
bomllm-open-webui-1   open-webui    Up 26 days     127.0.0.1:3000->3000/tcp
bomllm-postgres-1     postgres      Up 26 days     5432/tcp
bomllm-qdrant-1       qdrant        Up 26 days     6333/tcp
bomllm-valkey-1       valkey        Up 26 days     6379/tcp
bomllm-searxng-1      searxng       Up 26 days     127.0.0.1:8888->8080/tcp
# public path = Cloudflare Tunnel เท่านั้น (docs/security.md)

# 1. LiteLLM alive + DB connected
curl -sf https://llm.example.com/health/liveliness      # 200
curl -sf https://llm.example.com/health/readiness      # 200, db=connected
# 2. Unauth gate holds
curl -s -o /dev/null -w '%{http_code}' \
  -X POST https://llm.example.com/v1/chat/completions  # 401
# 3. Real chat through the whole chain (Thai)
curl -sf https://llm.example.com/v1/chat/completions \
  -H "Authorization: $SMOKE_KEY" \
  -d @smoke-payload.json
# 4. WebUI up   5. SearXNG local   6. Mac resident, no swap growth
curl -sf -o /dev/null -w '%{http_code}' https://chat.example.com/  # 200
curl -sf 'http://127.0.0.1:8888/search?q=test&format=json' -o /dev/null
ssh mac 'ollama ps && vm_stat | grep -i swap'

$ ssh mac 'vm_stat | grep -i swap'
swapins: 0.            swapouts: 0.
# swap 0 = memory pressure green — หนึ่งโมเดล resident ต่อช่วงเวลา
# อุบัติเหตุครั้งก่อน: 2 โมเดล resident → 47.7 GB, 3.8 GB swap,
#   TTFT 28-60 วิ → ไล่โมเดลที่สองออก (keep_alive: 0) ครบ 27 นาที สงบ