菜单
Deploy the full LLM observability stack (Langfuse + ClickHouse + LibreChat + demo apps) end to end and verify it is demo-ready. Use when the user says "deploy this demo", "set this up", "get the stack running", or when a partial/failed deployment needs to be brought to a healthy state.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Diagnose and fix problems with the LLM observability stack — services not starting, traces not appearing, missing scores, agents without MCP tools, 401 errors, port conflicts. Use when anything in the demo stack is broken, unhealthy, or behaving unexpectedly.
Present, rehearse, or drive the LLM observability demo — pre-flight checks, fresh trace generation, and a guided walkthrough of what to show and say in Langfuse and LibreChat. Use when the user says "run the demo", "demo this", "walk me through the demo", "prep me for a customer demo", or asks what to show for a given audience or time slot.
Interact with Langfuse and access its documentation. Use when needing to (1) query or modify Langfuse data programmatically via the CLI — traces, prompts, datasets, scores, sessions, and any other API resource, (2) look up Langfuse documentation, concepts, integration guides, or SDK usage, or (3) understand how any Langfuse feature works. This skill covers CLI-based API access (via npx) and multiple documentation retrieval methods.
| name | deploy-demo |
| description | Deploy the full LLM observability stack (Langfuse + ClickHouse + LibreChat + demo apps) end to end and verify it is demo-ready. Use when the user says "deploy this demo", "set this up", "get the stack running", or when a partial/failed deployment needs to be brought to a healthy state. |
The canonical, non-interactive deploy runbook is AGENTS.md at the repo root. Read it and follow it exactly — this skill only adds the operating posture around it.
Check prerequisites (Docker running, docker compose, jq) using the commands
in AGENTS.md > Prerequisites. Fix what you can (brew install jq / apt-get install -y jq);
only stop if Docker itself is unavailable.
Resolve the one required secret. The only secret needed is an Anthropic API key.
Check .env first (grep '^ANTHROPIC_API_KEY=sk' .env), then the environment.
If absent, ask the user for it — never invent one, never proceed without it.
Deploy:
ANTHROPIC_API_KEY=sk-ant-... ./setup.sh --seed
--seed adds a few minutes of traces.docker compose up for initial deployment; setup.sh provisions
secrets and API-side config that compose alone does not.Verify with every check in AGENTS.md > Verify (health endpoints, LLM connection,
5 LibreChat agents, traces, evaluators) or ./setup.sh --status — all "Demo
Readiness" lines must be ✓. Do not declare success on a partial pass.
Report the URLs and credentials exactly as listed in AGENTS.md > "Report to the
user when done", and suggest the natural next steps: present it (run-demo skill,
docs/SA_FIELD_GUIDE.md) or explore it
(docs/USER_JOURNEY.md).
Use the AGENTS.md > Troubleshooting table first, then the troubleshoot skill for
deeper diagnosis. Re-run ./setup.sh after any fix. Never run ./scripts/reset.sh
(destructive) without explicit user confirmation.