Livebook: functional notebooks for AI and Web with Elixir
summary
Livebook demonstrates Elixir/Erlang VM features via runnable notebooks that treat each cell as a pure function for reproducible execution, spins up one million processes in ~3 seconds to show lightweight concurrency, and uses Kino visualizations and smart cells to trace messages and chart per-process memory/reductions. It runs a Bandit web server on port 6789 with a supervision tree, hosts Bumblebee models served by Nx serving for text and speech tasks with batching, and moves ML to a separate node without changing app code.
AI-written summary · about
chapters (12)
- 0:00Livebook and Elixir intro
- 3:11Livebook’s functional design
- 7:41Elixir’s lightweight processes
- 10:33Visualizing processes with Kino
- 13:23Monitoring with smart cells
- 20:47Building a web app
- 24:20Supervision trees in Elixir
- 27:31Adding request handling
- 28:56Integrating ML models
- 34:13Batching for ML efficiency
- 38:27Scaling AI apps
- 43:25Q&A
related talks
47mLivebook & Elixir: Where AI, Web & Concurrency Meet
53mElixir: a potion to mix complexity and concurrency in real‑time
41mIdioms for building fault-tolerant applications with Elixir
38mDemos and Lightning Talks
45mElixir: une potion pour mixer complexité et concurrence en temps réel
01h 13mKeynote: Phoenix LiveView - Interactive Apps without Javascript
35mKeynote: Fireside Chat on LiveView, (...)
37mKeynote: All Aboard The Stateful Train