Week 2 — Resources

Links and references from the Week 2 sessions, grouped by class.

Getting set up

agentic-workflows — the Day 2 exercise repo. Clone it, run uv sync, then copy .env.example to .env and fill it in.
uv — the only thing you need to install. It fetches Python 3.12 for you.
Your AWS Bedrock key arrives by direct message and starts with ABSK. It goes in .env on one line, and it expires five days after it was issued — when calls suddenly start failing, that is usually why. Ask an instructor to re-issue it.
Course Langfuse — sign in with Google SSO using your course email, create a project, then Settings → API Keys → Create new API key and copy both halves into .env. The secret key is shown only once.
Not sure whether your setup works? Run uv run python exercises/01-hello-bedrock.py. It reports credential problems in plain English instead of a stack trace.

Frameworks: LangChain, Bedrock, and Langfuse

LangChain — overview — the uniform interface to the model.
Amazon Bedrock — the Converse API — what ChatBedrockConverse is calling underneath.
Langfuse docs — tracing concepts and the LangChain integration.

The course Langfuse server is v3, so the Python SDK is pinned below 4. Examples you find online are increasingly written for the v4 SDK and will not work here. If a snippet does not match from langfuse.langchain import CallbackHandler, it is the wrong version.

Structured Output, Tools, and LangGraph

LangGraph — overview — nodes, edges, state, and checkpointers.
Pydantic — the models behind structured output. Field descriptions are prompt text the model reads, not just documentation for humans.

LangGraph in this repo is 1.x. Most tutorials and blog posts still target the 0.x API, where the imports and the graph-building calls differ. When in doubt, read the installed package rather than a search result.

Capstone: Finding Relevant Research

NCICS Publications — the source the capstone reads. One static page, every year, no API.
Model Context Protocol — the standard behind the bonus exercise's local server.

Course Tools

Aca — AI Course Assistant — the AI assistant wired into this course.

Getting set up

agentic-workflows — the Day 2 exercise repo. Clone it, run uv sync, then copy .env.example to .env and fill it in.

uv — the only thing you need to install. It fetches Python 3.12 for you.

Your AWS Bedrock key arrives by direct message and starts with ABSK. It goes in .env on one line, and it expires five days after it was issued — when calls suddenly start failing, that is usually why. Ask an instructor to re-issue it.

Course Langfuse — sign in with Google SSO using your course email, create a project, then Settings → API Keys → Create new API key and copy both halves into .env. The secret key is shown only once.

Not sure whether your setup works? Run uv run python exercises/01-hello-bedrock.py. It reports credential problems in plain English instead of a stack trace.

Frameworks: LangChain, Bedrock, and Langfuse

LangChain — overview — the uniform interface to the model.

Amazon Bedrock — the Converse API — what ChatBedrockConverse is calling underneath.

Langfuse docs — tracing concepts and the LangChain integration.

Structured Output, Tools, and LangGraph

LangGraph — overview — nodes, edges, state, and checkpointers.

Pydantic — the models behind structured output. Field descriptions are prompt text the model reads, not just documentation for humans.