Skip to content

CHEATSHEET.md

Latest commit

 

History

History
37 lines (30 loc) · 2.74 KB

CHEATSHEET.md

File metadata and controls

37 lines (30 loc) · 2.74 KB
Raw

Cheatsheet - Stack It Yourself

Every step: cp env/stepN.env .env then docker compose up -d Fallen behind? Do exactly that for the step the room is on. You're caught up. What did a step change? git diff --no-index env/step2.env env/step3.env Why does each step matter? workshop-steps.md explains every step, with checkpoints and troubleshooting.

Step Command Open Look for
1 Baseline cp env/step1.env .env · docker compose up -d http://localhost:8000 class OllamaModel · API Chat Completions. The first reply takes 15–30 s while the model loads.
2 Responses API cp env/step2.env .env · docker compose up -d http://localhost:8000 class OpenAIResponsesModel · API Responses. Same server, one line changed.
3 Observability cp env/step3.env .env · docker compose up -d http://localhost:3000 → Dashboards → Stack It Yourself - Agent. Quicker: http://localhost:8000, ask a question, click trace ↗ under the reply The trace shows POST /chat → invoke_agent → chat → POST
4 Redis history cp env/step4.env .env · docker compose up -d http://localhost:8000 Chat, then docker compose rm -sf agent and docker compose up -d. Reload the page: the conversation is still there.
5 Tools + MCP cp env/step5.env .env · docker compose up -d http://localhost:8000 · Grafana Ask "What workshops are on Friday?", then open its trace. It spans agent and mcp-server.
6 Cluster (about 5 min to set up) ./scripts/cluster-up.sh http://localhost:8081 · http://localhost:3001 kubectl -n stack-it-yourself scale deployment/agent --replicas=3

Windows: cp works in PowerShell. In cmd, use copy. macOS/Linux: Permission denied running a script? chmod +x ./scripts/*.sh once.

Good questions to ask the agent

  • What workshops are on Friday?
  • Which workshop is about security?
  • What time is it in San Jose right now?
  • What's on in this room after this workshop?

Rescue commands

Problem Fix
Something is odd docker compose ps - everything should say healthy
See what the agent is doing docker compose logs -f agent
Ports already in use docker compose down, then find what else is on 8000, 3000 or 11434
Start clean docker compose --profile observability --profile state --profile tools down then redo the current step
Peek into Redis docker compose exec redis redis-cli KEYS 'conversation:*'
Hit the model directly curl http://localhost:11434/v1/models
Local model won't run Hosted fallback on Azure AI Foundry: cp env/hosted.env.example env/hosted.env, fill in the endpoint and key, then docker compose up -d