v10

Before You Begin

Summary — what this page covers The Day 2 pre-flight. The key difference from Day 1: this day runs on the Anthropic API (pay-as-you-go), which is separate from a Pro/Max subscription — an API key is required. It also covers the Docker dependency (Qdrant + TestContainers) and the free/local embedding alternative (Ollama) so the RAG section doesn't require a paid OpenAI key. Get everything ready before 9:00 AM.

API access (read carefully — this is different from Day 1)

A Claude.ai Pro or Max subscription powers Claude Code (Day 1) but does NOT grant API access. Day 2 needs an API key from the Anthropic Console.

Service Cost Notes
Anthropic API key (Console) Pay-as-you-go Required for every section. New Console accounts get a small free trial credit — enough for the workshop's calls.
OpenAI embedding key (Section 3) Pay-as-you-go Used by the RAG embedding step. Free alternative below.

Free-first note: to keep the RAG section on free/local infrastructure, swap the OpenAI embedding call for a local model via Ollama (nomic-embed-text, 768-dim) behind the same EmbeddingService interface, and set the Qdrant vector size to match (768). Install Ollama, then:

ollama pull nomic-embed-text     # download the embedding model (~270 MB)
ollama serve                     # run the local server on http://localhost:11434

No OpenAI key required. You'll point EmbeddingService at the local endpoint in Section 3.

Machine prerequisites

  • .NET 10 SDK + the BookTracker solution from Day 1 (cloned and building)

  • Anthropic API key set via user-secrets (don't commit it)

  • Docker Desktop running — needed for Qdrant (Section 3) and TestContainers (Section 5). Pull the image ahead of time so the lab doesn't wait on a download: docker pull qdrant/qdrant.

  • (Optional, free path) Ollama installed with nomic-embed-text pulled

dotnet user-secrets init --project BookTracker.Api
dotnet user-secrets set "Anthropic:ApiKey" "sk-ant-..." --project BookTracker.Api

A note on cost

Typical workshop usage runs from a few cents to a couple of dollars — comfortably inside the free trial credit on a new Console account. Prompt caching (Section 1) and batch processing reduce it further, and using Haiku for mechanical work keeps it low. See the Models & Cost Reference for current per-token pricing and the free/local alternatives.