Runnable notebooks, examples, and guides for the JAZ framework — a Python framework for building and optimizing LLM agents through a code-first REPL loop.
This is the public home for JAZ's hands-on material. Install the package with
pip install jaz-lang (it imports as jaz).
notebooks/— Google Colab tutorials. Each runs in your browser, no local setup required, and each ships with its outputs so you can read it without running anything.
| Notebook | What it covers | Time |
|---|---|---|
| Inputs: prompts, tools, and data | One kind of argument: prompts, tools, data and your own objects all go in the same way | ~10 min |
| Recursive subagents and scope | Agents calling agents, why inputs stop at one call, and how scope reaches the whole tree |
~10 min |
| Hooks, events and effects | The extension model: typed returns, validation, budgets and limits are all one mechanism | ~12 min |
| Configuration | The LLM, the REPL and the protocol underneath — including what the agent is allowed to import | ~10 min |
| Putting it together | The homepage's program, using all of it: a DataFrame and a web search in, a typed, checked report out | ~12 min |
Run them in order — each one builds on the last.
See notebooks/README.md for how to run them.