agentleFS
saintgo7saintgo7/claude-skills/gem-llm-overview/SKILL.md

gem-llm-overview

Skill0 starsChanged 6 months agoMITView on GitHub

GEM-LLM 프로젝트 전체 구조와 컴포넌트 위치를 빠르게 파악. 사용 시점 — "gem-llm이 뭐야", "전체 구성", "어디서 시작해", "프로젝트 구조", "포트 매핑", "어떤 모델 사용", "llm.pamout.com 어디로 가", "gem-llm overview". 시스템에 새로 합류하거나 어떤 컴포넌트가 어디 있는지 빠르게 확인할 때 사용. 깊은 디버깅은 다른 gem-llm-* skill로.

  • Reads credentials
  • Deletes or force-pushes

The file

Files
SKILL.md

gem-llm-overview

GEM-LLM = Gemma → Qwen Coder로 전환된 Qwen2.5-Coder-32B + Qwen3-Coder-30B-A3B 듀얼 서빙 코딩 어시스턴트 시스템.

한 그림

[curl https://llm.pamout.com] 
  → [Cloudflare Tunnel (master 노드 통과)]
  → [n1 cloudflared (port 8080 forward)]
  → [Gateway :8080 (FastAPI, API key 인증, quota, rate-limit)]
  → [vLLM :8001 Dense (Qwen2.5-Coder-32B, GPU 0-3, TP=4)]
  → [vLLM :8002 MoE (Qwen3-Coder-30B-A3B, GPU 4-7, TP=4)]

포트 매핑 (n1 = wku-vs-01-0)

포트서비스모델 / 역할
8001vLLM DenseQwen2.5-Coder-32B-Instruct (GPU 0-3, TP=4)
8002vLLM MoEQwen3-Coder-30B-A3B-Instruct (GPU 4-7, TP=4)
8080GatewayOpenAI 호환 + 인증 + quota
8090Admin UIFastAPI + Jinja2 + HTMX
외부https://llm.pamout.comCloudflare Tunnel → 8080

디렉토리

/home/jovyan/gem-llm/ (datavol-1 심볼릭)

  • src/ — vllm-serve, gateway, cli, admin-ui, common
  • docs/ — book-{ko,en} (~576p 각), manual-{ko,en} (~150p 각), paper-{ko,en} (KCI/IEEE), diagrams/ (40 SVG)
  • plan/ — 12 SPEC + 3 ADR + roadmap
  • tests/ — integration, load (locust + multi-user-bench), smoke
  • scripts/ — supervisor.sh, admin-cli.sh, build-docs.sh, build-diagrams.sh
  • _logs/ — 운영 로그
  • _data/ — gateway.db (SQLite, gitignore)
  • _trash/ — destructive 대신 격리 보존

주요 명령

bash scripts/supervisor.sh status        # 전체 스택 상태
bash scripts/supervisor.sh start         # 시작 (vLLM 로딩 5분)
bash scripts/admin-cli.sh add-user X X@y pro
bash scripts/admin-cli.sh issue-key <user_id>
make book-ko book-en                     # PDF 빌드
locust -f tests/load/locustfile.py       # 부하 테스트

외부 노출

  • llm.pamout.com — n1 cloudflared tunnel 10f3cb24-...로 라우팅
  • master tunnel (e9781f73-...) → paper.pamout.com (code-server)는 별개
  • DNS 라우팅 추가는 master에서 cloudflared tunnel route dns ... llm.pamout.com

알려진 함정

  • vLLM 0.19.1 사용 (0.20.0 DeepGEMM 빌드 실패). transformers 5.7.0, mistral_common 최신 필수.
  • SQLite WAL 파일 mv 금지 (case 13 — DB 손상). .gitignore 추가는 OK.
  • Gemma 4 미사용 (case 11 — layer_scalar weight 호환 깨짐). Qwen Coder 듀얼이 현재 운영.
  • n1↔master SSH는 양방향 (~/.ssh/config의 Host master/Host n3 cloudflared access ssh).
  • rm -rf 절대 금지 (feedback memory). 삭제 대신 _trash/<날짜>/ 로 mv.

깊은 디버깅 → 다른 skill

  • vLLM 모델 로딩/충돌 → gem-llm-deploy-vllm
  • 사용자/키 관리 → gem-llm-admin-cli
  • 부하 테스트 → gem-llm-load-test
  • 빌드 (PDF/SVG/논문) → gem-llm-build-docs
  • CLI 사용 → gem-llm-cli-client
  • 외부 라우팅 → gem-llm-cloudflare-tunnel
  • 에러 사례 → gem-llm-troubleshooting
  • 추론 검증 → gem-llm-test-inference
  • 프롬프트 리뷰 → gem-llm-review-prompt
  • MCP 디버깅 → gem-llm-debug-mcp

GitHub

  • Repo: saintgo7/gem-llm (Private)
  • 9 commits at last sync
  • credentials store: ~/.git-credentials (mode 600)

메모리 참조

  • project_gem_llm.md (정적 사실)
  • project_gem_llm_runtime.md (가동 상태, 버전, 함정)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.