pixi-build-python fails with ModuleNotFoundError: No module named 'editables'

[weg-jr]

I am working through the pixi-build-python example and have a monorepo with 1 shared lib called common and 2 packages called pipeline_a and pipeline_b. pixi 0.76.1 is the newest version I have access to and I am using it inside of a conda environment. I get the following error running pixi install: DEBUG ModuleNotFoundError: No module named 'editables'.

I highly suspect I misunderstand something about my toml configuration. It will work if I add editables to my workspace dependencies but that doesnt seem like I should be the one specifying it. It seems there are 3 cases:

1. I misconfigured my project or package toml
2. This is intended behavior: then the docs should include that as a requirement in the example: https://prefix-dev.github.io/pixi-build-backends/backends/pixi-build-python/#example
3. This is unintended behavior for pixi-build

This is the build config. note that python is 3.13.15 but its using hatchling built for python 3.12 which seems like a problem to me

╭─ Running build for recipe: pipeline_b-0.1.0-pyh4616a5c_0
│
│ ╭─ Fetching source code
│ │ No sources to fetch
│ │
│ ╰─────────────────── (took 0 seconds)
│
│ ╭─ Externally resolved dependencies recipe: pipeline_b-0.1.0-pyh4616a5c_0
│ │
│ │ Resolved build dependencies(pipeline_b-0.1.0-pyh4616a5c_0):
│ │ ╭──────────────────┬─────────────────────┬───────────┬──────────────────────┬─────────────┬────────────╮
│ │ │ Package ┆ Spec ┆ Version ┆ Build ┆ Channel ┆ Size │
│ │ ╞══════════════════╪═════════════════════╪═══════════╪══════════════════════╪═════════════╪════════════╡
│ │ │ python ┆ python >=3.12,<3.14 ┆ 3.13.15 ┆ hb101c97_101_cp313 ┆ conda-forge ┆ 35.75 MiB │
│ │ │ _openmp_mutex ┆ ┆ 5.1 ┆ 52_gnu ┆ main ┆ 6.52 KiB │
│ │ │ bzip2 ┆ ┆ 1.0.8 ┆ hda65f42_10 ┆ conda-forge ┆ 251.77 KiB │
│ │ │ ca-certificates ┆ ┆ 2026.7.22 ┆ hbd8a1cb_0 ┆ conda-forge ┆ 128.69 KiB │
│ │ │ icu ┆ ┆ 78.3 ┆ h54a6638_2 ┆ conda-forge ┆ 13.79 MiB │
│ │ │ ld_impl_linux-64 ┆ ┆ 2.46.1 ┆ default_hbd61a6d_102 ┆ conda-forge ┆ 727.83 KiB │
│ │ │ libexpat ┆ ┆ 2.8.2 ┆ h7354ed3_1 ┆ main ┆ 126.34 KiB │
│ │ │ libffi ┆ ┆ 3.7.0 ┆ h3435931_0 ┆ conda-forge ┆ 65.99 KiB │
│ │ │ libgcc ┆ ┆ 16.1.0 ┆ ha9f2e26_1 ┆ conda-forge ┆ 1.01 MiB │
│ │ │ libgcc-ng ┆ ┆ 16.1.0 ┆ h69a702a_1 ┆ conda-forge ┆ 27.55 KiB │
│ │ │ liblzma ┆ ┆ 5.8.3 ┆ hb03c661_1 ┆ conda-forge ┆ 110.35 KiB │
│ │ │ libmpdec ┆ ┆ 4.0.1 ┆ h47b2149_0 ┆ main ┆ 88.50 KiB │
│ │ │ libsqlite ┆ ┆ 3.53.4 ┆ hf4e2dac_0 ┆ conda-forge ┆ 941.60 KiB │
│ │ │ libstdcxx ┆ ┆ 16.1.0 ┆ h934c35e_1 ┆ conda-forge ┆ 6.32 MiB │
│ │ │ libuuid ┆ ┆ 2.42.2 ┆ h5347b49_0 ┆ conda-forge ┆ 39.08 KiB │
│ │ │ libxcb ┆ ┆ 1.17.0 ┆ h9b100fa_0 ┆ main ┆ 430.36 KiB │
│ │ │ libzlib ┆ ┆ 1.3.2 ┆ h25fd6f3_3 ┆ conda-forge ┆ 62.22 KiB │
│ │ │ ncurses ┆ ┆ 6.6 ┆ hdb14827_1 ┆ conda-forge ┆ 889.84 KiB │
│ │ │ openssl ┆ ┆ 3.6.3 ┆ h35e630c_1 ┆ conda-forge ┆ 3.03 MiB │
│ │ │ pthread-stubs ┆ ┆ 0.4 ┆ hb03c661_1003 ┆ conda-forge ┆ 8.90 KiB │
│ │ │ python_abi ┆ ┆ 3.13 ┆ 8_cp313 ┆ conda-forge ┆ 6.84 KiB │
│ │ │ readline ┆ ┆ 8.3 ┆ h853b02a_0 ┆ conda-forge ┆ 336.99 KiB │
│ │ │ tk ┆ ┆ 8.6.15 ┆ h54e0aa7_0 ┆ main ┆ 3.44 MiB │
│ │ │ tzdata ┆ ┆ 2026c ┆ h151e31d_0 ┆ conda-forge ┆ 116.06 KiB │
│ │ │ xorg-libx11 ┆ ┆ 1.8.13 ┆ he1eb515_0 ┆ conda-forge ┆ 819.97 KiB │
│ │ │ xorg-libxau ┆ ┆ 1.0.12 ┆ hb03c661_2 ┆ conda-forge ┆ 16.03 KiB │
│ │ │ xorg-libxdmcp ┆ ┆ 1.1.5 ┆ hb03c661_2 ┆ conda-forge ┆ 20.62 KiB │
│ │ │ zlib ┆ ┆ 1.3.2 ┆ h25fd6f3_3 ┆ conda-forge ┆ 93.88 KiB │
│ │ │ zstd ┆ ┆ 1.5.7 ┆ hb78ec9c_6 ┆ conda-forge ┆ 587.28 KiB │
│ │ ╰──────────────────┴─────────────────────┴───────────┴──────────────────────┴─────────────┴────────────╯
│ │ Resolved host dependencies(pipeline_b-0.1.0-pyh4616a5c_0):
│ │ ╭───────────────────┬───────────┬─────────────┬──────────────────────┬─────────────┬────────────╮
│ │ │ Package ┆ Spec ┆ Version ┆ Build ┆ Channel ┆ Size │
│ │ ╞═══════════════════╪═══════════╪═════════════╪══════════════════════╪═════════════╪════════════╡
│ │ │ hatchling ┆ hatchling ┆ 1.31.0 ┆ py312h06a4308_0 ┆ main ┆ 160.62 KiB │
│ │ │ python ┆ python ┆ 3.12.13 ┆ hc5f7cf0_3 ┆ main ┆ 30.24 MiB │
│ │ │ uv ┆ uv ┆ 0.12.3 ┆ h2112641_0 ┆ conda-forge ┆ 16.50 MiB │
│ │ │ _openmp_mutex ┆ ┆ 5.1 ┆ 52_gnu ┆ main ┆ 6.52 KiB │
│ │ │ bzip2 ┆ ┆ 1.0.8 ┆ hda65f42_10 ┆ conda-forge ┆ 251.77 KiB │
│ │ │ ca-certificates ┆ ┆ 2026.7.22 ┆ hbd8a1cb_0 ┆ conda-forge ┆ 128.69 KiB │
│ │ │ icu ┆ ┆ 78.3 ┆ h54a6638_2 ┆ conda-forge ┆ 13.79 MiB │
│ │ │ ld_impl_linux-64 ┆ ┆ 2.46.1 ┆ default_hbd61a6d_102 ┆ conda-forge ┆ 727.83 KiB │
│ │ │ libexpat ┆ ┆ 2.8.2 ┆ h7354ed3_1 ┆ main ┆ 126.34 KiB │
│ │ │ libffi ┆ ┆ 3.4.8 ┆ h06d3fd0_3 ┆ main ┆ 136.92 KiB │
│ │ │ libgcc ┆ ┆ 16.1.0 ┆ ha9f2e26_1 ┆ conda-forge ┆ 1.01 MiB │
│ │ │ libgcc-ng ┆ ┆ 16.1.0 ┆ h69a702a_1 ┆ conda-forge ┆ 27.55 KiB │
│ │ │ liblzma ┆ ┆ 5.8.3 ┆ hb03c661_1 ┆ conda-forge ┆ 110.35 KiB │
│ │ │ liblzma-devel ┆ ┆ 5.8.3 ┆ hb03c661_1 ┆ conda-forge ┆ 480.44 KiB │
│ │ │ libsqlite ┆ ┆ 3.53.4 ┆ hf4e2dac_0 ┆ conda-forge ┆ 941.60 KiB │
│ │ │ libstdcxx ┆ ┆ 16.1.0 ┆ h934c35e_1 ┆ conda-forge ┆ 6.32 MiB │
│ │ │ libuuid ┆ ┆ 1.41.5 ┆ h5eee18b_0 ┆ main ┆ 27.45 KiB │
│ │ │ libxcb ┆ ┆ 1.17.0 ┆ h9b100fa_0 ┆ main ┆ 430.36 KiB │
│ │ │ libzlib ┆ ┆ 1.3.2 ┆ h25fd6f3_3 ┆ conda-forge ┆ 62.22 KiB │
│ │ │ ncurses ┆ ┆ 6.6 ┆ hdb14827_1 ┆ conda-forge ┆ 889.84 KiB │
│ │ │ openssl ┆ ┆ 3.6.3 ┆ h35e630c_1 ┆ conda-forge ┆ 3.03 MiB │
│ │ │ packaging ┆ ┆ 26.3 ┆ pyhc364b38_0 ┆ conda-forge ┆ 113.64 KiB │
│ │ │ pathspec ┆ ┆ 1.1.1 ┆ py312h06a4308_0 ┆ main ┆ 94.22 KiB │
│ │ │ pluggy ┆ ┆ 1.6.0 ┆ pyhf9edf01_1 ┆ conda-forge ┆ 25.27 KiB │
│ │ │ pthread-stubs ┆ ┆ 0.4 ┆ hb03c661_1003 ┆ conda-forge ┆ 8.90 KiB │
│ │ │ readline ┆ ┆ 8.3 ┆ h853b02a_0 ┆ conda-forge ┆ 336.99 KiB │
│ │ │ sqlite ┆ ┆ 3.53.4 ┆ h04a0ce9_0 ┆ conda-forge ┆ 201.85 KiB │
│ │ │ tk ┆ ┆ 8.6.15 ┆ h54e0aa7_0 ┆ main ┆ 3.44 MiB │
│ │ │ trove-classifiers ┆ ┆ 2026.6.1.19 ┆ py312h06a4308_0 ┆ main ┆ 29.04 KiB │
│ │ │ tzdata ┆ ┆ 2026c ┆ h151e31d_0 ┆ conda-forge ┆ 116.06 KiB │
│ │ │ xorg-libx11 ┆ ┆ 1.8.13 ┆ he1eb515_0 ┆ conda-forge ┆ 819.97 KiB │
│ │ │ xorg-libxau ┆ ┆ 1.0.12 ┆ hb03c661_2 ┆ conda-forge ┆ 16.03 KiB │
│ │ │ xorg-libxdmcp ┆ ┆ 1.1.5 ┆ hb03c661_2 ┆ conda-forge ┆ 20.62 KiB │
│ │ │ xz ┆ ┆ 5.8.3 ┆ ha02ee65_1 ┆ conda-forge ┆ 23.06 KiB │
│ │ │ xz-gpl-tools ┆ ┆ 5.8.3 ┆ ha02ee65_1 ┆ conda-forge ┆ 39.18 KiB │
│ │ │ xz-tools ┆ ┆ 5.8.3 ┆ hb03c661_1 ┆ conda-forge ┆ 94.20 KiB │
│ │ │ zlib ┆ ┆ 1.3.2 ┆ h25fd6f3_3 ┆ conda-forge ┆ 93.88 KiB │
│ │ │ zstd ┆ ┆ 1.5.7 ┆ hb78ec9c_6 ┆ conda-forge ┆ 587.28 KiB │
│ │ ╰───────────────────┴───────────┴─────────────┴──────────────────────┴─────────────┴────────────╯
│ │ Resolved run dependencies(pipeline_b-0.1.0-pyh4616a5c_0):
│ │ ╭──────────────────┬────────────╮
│ │ │ Name ┆ Spec │
│ │ ╞══════════════════╪════════════╡
│ │ │ Run dependencies ┆ │
│ │ │ common ┆ * │
│ │ │ pandas ┆ >=2.0,<3.0 │
│ │ │ python ┆ * │
│ │ ╰──────────────────┴────────────╯
│ │
│ ╰─────────────────── (took 0 seconds)

This is the error message I get when it uses uv to build pipeline-b:

DEBUG ModuleNotFoundError: No module named 'editables'
│ │ TRACE Released lock at <REDACTED>
│ │ × Failed to build pipeline-b @ │ │ │ file://<REDACTED>
...
│ │ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
│ │ File
│ │ "$PREFIX/lib/python3.12/site-packages/hatchling/builders/plugin/interface.py",
│ │ line 157, in build
│ │ artifact = version_api[version](directory, **build_data)
│ │ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
│ │ File
│ │ "$PREFIX/lib/python3.12/site-packages/hatchling/builders/wheel.py",
│ │ line 547, in build_editable
│ │ return self.build_editable_detection(directory, **build_data)
│ │ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
│ │ File
│ │ "$PREFIX/lib/python3.12/site-packages/hatchling/builders/wheel.py",
│ │ line 550, in build_editable_detection
│ │ from editables import EditableProject
│ │ ModuleNotFoundError: No module named 'editables'
│ │ hint: This error likely indicates that pipeline-b@0.1.0 depends on editables, but doesn't declare it as a build dependency. If pipeline-b is a first-
│ │ party package, consider adding editables to its build-system.requires. Otherwise, either add it to your pyproject.toml under:
│ │ [tool.uv.extra-build-dependencies]
│ │ pipeline-b = ["editables"]
│ │ or uv pip install editables into the environment and re-run with --no-build-isolation.

This is my project toml

[workspace]
name = "data-monorepo"
version = "0.1.0"
description = "Minimal example of a Python monorepo using pixi workspaces"
channels = ["<CONDA PROXIES>"]
platforms = ["linux-64"]
channel-priority = "disabled" # also disabled, dont use strict
conda-pypi-map = false
requires-pixi = "==0.76.1"
preview= ['pixi-build']
[workspace.dependencies]
python = ">=3.12,<3.14"
pandas = ">=2.0,<3.0"
pytest = ">=8.0"
pytest-cov = "*"
pixi-build-python = "*"
hatchling = "*"
common = { path = "libs/common" }
[dependencies]
python = { workspace = true }
pytest = { workspace = true }
common = { workspace = true }
pipeline_a = { path = "pipelines/pipeline_a" }
pipeline_b = { path = "pipelines/pipeline_b" }
[pypi-options]
index-url = "<pypi proxy>"

This is my package toml

[package]
name = "pipeline_b"
version = "0.1.0"
description = "Example data pipeline B"

[package.build.config]
ignore-pypi-mapping = true

[package.build]
backend = { name = "pixi-build-python", workspace = true }

[package.build-dependencies]
python = { workspace = true }

[package.host-dependencies]
hatchling = { workspace = true }

[package.run-dependencies]
common = { workspace = true }
pandas = { workspace = true }

This is my package pyproject.toml

[project]
name = "pipeline_b"
version = "0.1.0"
description = "Example data pipeline B"
# requires-python = ">=3.11"
dependencies = ["common"]

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

[tool.hatch.build.targets.wheel]
packages = ["src/pipeline_b"]

[SAGIRIxr]

The immediate cause is a missing host dependency for the editable build, not the Python 3.13 build environment using a Python 3.12 Hatchling package.

pixi install installs this workspace package as editable. pixi-build-python runs the installer with --no-deps --no-build-isolation, so the conda host environment must already contain every package needed by the backend. Hatchling 1.31 dynamically returns editables from get_requires_for_build_editable() and then imports it while producing the editable wheel. Your resolved host list contains hatchling and uv, but not editables, which matches the traceback exactly.

The smallest deterministic fix for this configuration is:

[package.host-dependencies]
hatchling = { workspace = true }
editables = "*"

Keeping editables in the workspace dependency pool and inheriting it here is equivalent:

[workspace.dependencies]
editables = "*"

[package.host-dependencies]
hatchling = { workspace = true }
editables = { workspace = true }

Then rerun pixi install. You do not need editables as a run dependency: it is needed only while Hatchling creates the editable wheel.

Adding it only to [build-system].requires will not solve this particular manifest while ignore-pypi-mapping = true, because that setting disables mapping build-system.requires into conda host dependencies. Explicitly declaring the host dependency is therefore the reliable workaround on the version/configuration shown.

The different Python versions are worth constraining separately if reproducibility requires them to match, but they are not the source of this ModuleNotFoundError; the failing interpreter is the host prefix and simply lacks the module.

References:

• pixi-build-python build process and no-build-isolation
• Editable behavior for pixi install
• Effect of ignore-pypi-mapping
• Hatchling 1.31's editable hook declares editables and imports it while building

[weg-jr]

It sounds like adding it as a dependency is the expected behavior. Since I have no control over whether pixi-build-conda installs my package as editable or not, do you think that the basic example should include that in the host dependencies?

[package.host-dependencies]
hatchling = "*"
editables="*"

My thinking is everyone will try to run and it will fail if they use pixi install but succeed for pixi build and it will be confusing since using [pypi dependencies] will sucessfully install in editable mode with just hatchling as a dependency and not editables