Module 1 · Phase 1: Foundations from raw APIs · Weeks 1–2

LLM API Mastery

No frameworks. Raw HTTP/SDK calls only. Everything an agent does reduces to these mechanics: the message array, tool calling, structured outputs, streaming, tokens, and robust error handling.

After this module you can

▸Explain and implement the chat message format (system/user/assistant/tool roles) from memory
▸Implement tool calling end-to-end: schema → model emits call → you execute → return result → model continues
▸Produce validated structured outputs with JSON schema and recover from malformed JSON
▸Stream responses token-by-token and explain why streaming matters for agent UX
▸Count tokens, estimate cost per call, and reason about context-window budgets
▸Handle rate limits, timeouts, and refusals with exponential backoff and graceful degradation

Lessons

Messages Are the Only State

The single most important fact in agent engineering: the model is stateless. A 'conversation' is you resending an ever-growing array. Every agent pattern you'll ever build follows from this.

Sampling Parameters & Streaming

Temperature and top_p control the randomness of token selection — get them wrong and your agent is either erratic or uselessly rigid. Streaming turns dead air into perceived speed.

Tool Calling End-to-End

The mechanism that turns a text generator into something that can act. Crucial mental model: the model never executes anything — it emits structured JSON, and your code does the work.

Structured Outputs & JSON Schema

When you need data, not prose: forcing model output to conform to a schema, and what to do when it doesn't.

Errors, Rate Limits & Cost Control

An agent lives or dies on the unhappy path. Rate limits, timeouts, overloaded servers, refusals, context overflows — production behavior is defined by how you handle these.

12 questions · pass ≥ 80%

Lab: Tool-Calling CLI Assistant

Build a CLI assistant from scratch — raw SDK only, no frameworks — that answers questions using three tools: calculator, get_current_time, and read_file. This is the atom every later lab is built from. Starter code lives in labs/lab01-agent-loop/.

Best external resources

Curated reading, docs, and tools that pair with this module.

Anthropic — Tool use docs

The canonical reference for the message shapes used in Lab 01.

OpenAI — Function calling & structured outputs

Compare the two vendors' shapes — interviews ask about both.

Anthropic Cookbook

Runnable notebooks for every pattern in this module. Run the tool-use ones.

Prompt Engineering Guide

Reference for prompting techniques; skim the basics, bookmark the rest.

OpenAI Cookbook

The other vendor's runnable examples — structured outputs, function calling, streaming.

Chip Huyen — AI Engineering (book)

The best book-length treatment of this whole curriculum; chapters 1–2 pair with this module.