ice
ice copied to clipboard
Interactive Composition Explorer: a debugger for compositional language model programs
1. placebo recipes: 1. move most to `ice-ought-private` and refactor in this PR: https://github.com/oughtinc/ice-ought-private/pull/20 2. move recipes that are consumed by PlaceboTree and only by PlaceboTree 3. just delete some...
I've found that the presence & frequency penalties are rather important in tasks related to rephrasing - as the agent tends to fail via repeating input verbatim otherwise. These feel...
I would suggest using a safer version of eval by default in the [Interpreters](https://primer.ought.org/chapters/tool-use/interpreters) page (e.g. worry about generated code making web requests). [asteval](https://newville.github.io/asteval/) looks simple to use for mathematical...
When I run `./scripts/run-tests.sh`, all tests show PASSED except two: ``` tests/test_metrics.py::test_nubia FAILED tests/test_metrics.py::test_gold_paragraphs SKIPPED (Parses all PDFs - very slow) ``` 1. `test_nubia` fails with an obscure pydantic validation...
The README suggests in https://github.com/oughtinc/ice#running-tests: > Cheap integration tests: > > ``` > ./scripts/run-recipe.sh --mode test > ``` This had me confused for a while. 1. This prompts me select...
Currently an unhandled exception just shows the standard plain Python traceback. It can be helpful for debugging to show more information, and the dependencies are already there. One method is...
updates: - [github.com/pre-commit/pre-commit-hooks: v4.4.0 → v5.0.0](https://github.com/pre-commit/pre-commit-hooks/compare/v4.4.0...v5.0.0) - [github.com/asottile/reorder-python-imports: v3.10.0 → v3.14.0](https://github.com/asottile/reorder-python-imports/compare/v3.10.0...v3.14.0) - [github.com/psf/black: 23.3.0 → 24.10.0](https://github.com/psf/black/compare/23.3.0...24.10.0) - [github.com/PyCQA/flake8: 6.0.0 → 7.1.1](https://github.com/PyCQA/flake8/compare/6.0.0...7.1.1) - [github.com/pre-commit/mirrors-mypy: v1.4.1 → v1.14.1](https://github.com/pre-commit/mirrors-mypy/compare/v1.4.1...v1.14.1)
## Context I ran into some dependency issues when going through the Ought ICE [primer](https://primer.ought.org/intro/before-we-start). I thought I'd share the fix in case anyone else is in the same situation....
Updates to ice/settings and ice/recipe for fixing Pydantic import error: ``` pydantic.errors.PydanticImportError: `BaseSettings` has been moved to the `pydantic-settings` package. See https://docs.pydantic.dev/2.5/migration/#basesettings-has-moved-to-pydantic-settings for more details. ```