ICE is a Python library and trace visualizer for language model programs.
Execution trace visualized in ICE
:warning: The ICE API may change at any point. The ICE interface is being actively developed and we may change the API at any point, including removing functionality, renaming methods, splitting ICE into multiple projects, and other similarly disruptive changes. Use at your own risk.
ICE requires Python 3.9, 3.10, or 3.11. If you don't have a supported version of Python installed, we recommend using pyenv to install a supported Python version and manage multiple Python versions.
If you use Windows, you'll need to run ICE inside of WSL.
As part of general good Python practice, consider first creating and activating a virtual environment to avoid installing ICE 'globally'. For example:
python -m venv venv
source venv/bin/activate
Install ICE:
pip install ought-ice
Run the Hello World recipe in the Primer to see the trace rendered.
Optionally, set secrets (like your OpenAI API key) in ~/.ought-ice/.env
. See .env.example
for the format. If these are not set, you'll be prompted for them when you run recipes that need them.
If you want to make changes to ICE itself, clone the repository, then install it in editable mode:
python -m venv venv
source venv/bin/activate
pip install --upgrade pip
pip install -e '.[dev]' --config-settings editable_mode=compat
pre-commit install
npm --prefix ui ci
npm --prefix ui run dev
If you're working on the backend, you might find it helpful to remove the cache of language model calls:
rm -r ~/.ought-ice/cache
pre-commit
complains if your code doesn't pass certain checks. It runs when you commit, and will possibly reject your commit and make you have to fix the problem(s) before you can commit again. (So you should probably use the same commit message you used the first time.)
Note that you don't technically need to run pre-commit install
, but not doing so may cause your commits to fail CI. (Which can be noisy, including by generating commits that will e.g. fix formatting.)
We use Storybook for UI tests. You can run them locally:
npm --prefix ui run storybook
Note that build-storybook
is only for CI and shouldn't be run locally.
Recipes are decompositions of a task into subtasks.
The meaning of a recipe is: If a human executed these steps and did a good job at each workspace in isolation, the overall answer would be good. This decomposition may be informed by what we think ML can do at this point, but the recipe itself (as an abstraction) doesn’t know about specific agents.
Agents perform atomic subtasks of predefined shapes, like completion, scoring, or classification.
Agents don't know which recipe is calling them. Agents don’t maintain state between subtasks. Agents generally try to complete all subtasks they're asked to complete (however badly), but some will not have implementations for certain task types.
The mode in which a recipe runs is a global setting that can affect every agent call. For instance, whether to use humans or agents. Recipes can also run with certain RecipeSettings
, which can map a task type to a specific agent_name
, which can modify which agent is used for that specific type of task.
Join the ICE Slack channel to collaborate with other people composing language model tasks. You can also use it to ask questions about using ICE.
Watch the recording of Ought's Lab Meeting to understand the high-level goals for ICE, how it interacts with Ought's other work, and how it contributes to alignment research.
Read the ICE announcement post for another introduction.
ICE is an open-source project by Ought. We're an applied ML lab building the AI research assistant Elicit.
We welcome community contributions:
For larger contributions, make an issue for discussion before submitting a PR.
And for even larger contributions, join us - we're hiring!
If you use ICE, please cite:
Iterated Decomposition: Improving Science Q&A by Supervising Reasoning Processes. Justin Reppert, Ben Rachbach, Charlie George, Luke Stebbing Jungwon Byun, Maggie Appleton, Andreas Stuhlmüller (2023). Ought Technical Report. arXiv:2301.01751 [cs.CL]
Bibtex:
@article{reppert2023iterated,
author = {Justin Reppert and Ben Rachbach and Charlie George and Luke Stebbing and Jungwon Byun and Maggie Appleton and Andreas Stuhlm\"{u}ller},
archivePrefix = {arXiv},
eprint = {2301.01751},
primaryClass = {cs.CL},
title = {Iterated Decomposition: Improving Science Q&A by Supervising Reasoning Processes},
year = 2023,
keywords = {language models, decomposition, workflow, debugging},
url = {https://arxiv.org/abs/2301.01751}
}