Agentic Training

Download

Course code package

Learners with full course access can download a zip of every Python and TypeScript sample from the lessons. This page is the map of that zip: layout, env scripts, and which API key unlocks which samples.

Download from Learning (full unlock) or from a lesson’s code panel. The tutor on the site uses Vercel AI Gateway separately; the zip is for running samples on your machine.

What’s in the zip

  • README.md — setup and how to run a sample
  • .env.example — one Anthropic, OpenAI, or AI Gateway key
  • scripts/env.py and scripts/env.mjs — generate and check
  • python/ — requirements.txt, course_env.py, then one folder per lesson id
  • typescript/ — package.json, course-env.ts, then the same lesson folders
  • YAML and other non-code snippets sit next to the lesson id at the zip root

Env: generate and check

Fill in one API key in .env. If more than one is set, also set LLM_PROVIDER to anthropic, openai, or gateway.

python3 scripts/env.py generate
python3 scripts/env.py check
# or: node scripts/env.mjs generate && node scripts/env.mjs check

generate creates .env from the example when needed, picks the provider from the filled key (or from LLM_PROVIDER), and writes the derived block: LLM_PROVIDER, MODEL, FALLBACK_MODEL, and gateway base URLs when the provider is gateway. It does not overwrite the key you typed.

check parses .env, confirms the selected key and MODEL, and prints which sample groups that key unlocks.

Which key runs what

KeySamples
ANTHROPIC_API_KEYAnthropic Messages lessons, LangChain / Pydantic AI, raw HTTP lab, Claude Agent SDK, Vercel AI SDK (Anthropic)
OPENAI_API_KEYOpenAI Agents SDK and OpenAI-shaped samples. Anthropic Messages samples ask you to switch key.
AI_GATEWAY_API_KEYSame Anthropic- and OpenAI-shaped samples through https://ai-gateway.vercel.sh. Model ids look like anthropic/claude-sonnet-4-5.

Shared client helpers

Samples do not hardcode provider clients. They import course_env (Python) or course-env.ts (TypeScript), which read LLM_PROVIDER and MODEL from the environment.

  • Python: run with PYTHONPATH=python so import course_env resolves
  • TypeScript: from a lesson folder, import { ... } from "../course-env.ts"
  • anthropic_client() / anthropicClient() for Messages API samples
  • messages_http() / messagesHttp() for the Week 1 raw HTTP lab
  • ensure_openai_env() / ensureOpenAIEnv() before OpenAI Agents SDK samples when using the gateway

Run a sample

After generate, check, and installing dependencies:

set -a && source .env && set +a
PYTHONPATH=python python python/w1-tokens/count_tokens_and_price_a_call.py

# from typescript/
node --env-file=../.env --import tsx w1-tokens/count_tokens_and_price_a_call.ts

From typescript/ you can also run npm run env:generate and npm run env:check.

How the zip is built

The site builds the zip at request time from the lesson markdown in agentic-ai-course/, plus the static helpers under course-package/ (scripts/, python/course_env.py, typescript/course-env.ts). Code fences with a filename become files under python/<lesson-id>/ or typescript/<lesson-id>/.