You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After the autoconf → autonerves rename (#135), PyAutoNerves still carries a legacy surface from the pre-rename era. A survey on 2026-07-23 separated it into three genuinely different problems, which this task ships as four independent steps.
The autoconf sightings are not tracked drift — git grep -ril autoconf over tracked files returns nothing. What remains is untracked working-copy cruft, one test that litters the repo root on every run, and four dormant one-shot scripts (two of which are already broken on their own imports).
Plan
Delete the untracked rename leftovers (autoconf/, test_autoconf/, autoconf.egg-info/, build/, bad/) from the working copy — no commit needed, but autoconf/ is not inert: it still resolves as a namespace package and shadows imports.
Fix the test that regenerates bad/ in the repo root on every run, rather than gitignoring the symptom.
Delete the dead EDEN tooling (eden.yaml, scripts/edenise.py) — its driver imports autofit.tools.edenise, which no longer exists in PyAutoFit.
Delete the one-shot config-migration tooling (scripts/, priors/, json_prior/generate.py) that bootstrapped the original configs years ago and is wired into no CI and no workspace.
1. Untracked rename leftovers.autoconf/, test_autoconf/, autoconf.egg-info/, build/ and bad/ are all present in the working copy and all untracked. autoconf/ and test_autoconf/ contain only __pycache__ — but they still import:
So they shadow imports rather than sitting inert. build/ and *.egg-info/ are already covered by .gitignore.
2. bad/ is regenerated on every test run.test_autonerves/test_config.py:44 sets BAD_PATH = "bad/path" and test_path_empty calls os.makedirs(BAD_PATH, exist_ok=True) into the CWD, never cleaning up. Deleting the directory without fixing the producer is a no-op.
3. Dead one-shot tooling in scripts/. Four drivers from the original config bootstrap, dormant for years, referenced from no CI and no workspace. Two are already broken:
Script
State
edenise.py
imports autofit.tools.edenise — removed from PyAutoFit, raises ImportError
convert_prior_configs.py
imports oyaml — not a declared dependency
convert_config.py
the .ini → YAML migration, long since run
generate_priors.py
CLI over autonerves/json_prior/generate.py, whose only other consumer is its own test
Steps
Step 0 — local cruft (no PR). rm -rf autoconf/ test_autoconf/ autoconf.egg-info/ build/ bad/. All untracked; nothing to commit.
Step 1 — stop bad/ regenerating. test_autonerves/test_config.py: move test_path_does_not_exist and test_path_empty onto tmp_path; delete the BAD_PATH constant and the remove_path() helper. Fixes the producer instead of gitignoring the symptom.
Step 2 — remove dead EDEN tooling.
Delete eden.yaml and scripts/edenise.py. This is the PyAutoNerves leg of PyAutoMind/draft/refactor/pyautofit/remove_eden_packaging_tooling.md.
PyAutoFit/eden.yaml is descoped — the branch survey found PyAutoFit claimed by the testmode-env-drift worktree. autofit.tools.edenise is already gone from PyAutoFit main, so only the orphan eden.yaml remains there: a one-file follow-up once that claim releases. The eden draft stays open until then.
Step 3 — remove the one-shot config-migration tooling.
Delete scripts/convert_config.py, scripts/convert_prior_configs.py, scripts/generate_priors.py, autonerves/json_prior/generate.py, test_autonerves/json_prior/test_generate.py, and priors/subconfig.json (a stray output artifact of generate_priors.py). Both scripts/ and priors/ disappear.
Step 4 — sort PR #136.
Force-push the branch with only the test-fixture re-anchor (test_workspace.py, test_setup_colab.py), reverting autonerves/__init__.py to 2026.7.9.1. Rewrite the PR body to describe a fixture cleanup — moving the colab mock off the yanked 2026.7.6.649 — not a version sync.
autonerves/json_prior/config.py is NOT in scope. It is the live half — imported by autonerves/conf.py and re-exported from autonerves/__init__.py (default_prior, make_config_for_class, path_for_class, JSONPriorConfig). Only the sibling generate.py goes.
generate.py is an import surface. Confirmed clean before deletion: no consumer in any PyAuto library, workspace, or doc outside its own script and test. Re-run grep -rn "json_prior import generate\|json_prior.generate" before deleting; if a live consumer surfaces, re-scope.
Run the full test_autonerves suite (151 tests at survey time) before each push.
Library-only; no downstream workspace impact expected.
Testing approach
Full pytest test_autonerves before each push. For step 1, additionally confirm bad/ is absent from the repo root after a run. For step 3, confirm import autonerves still succeeds and the four json_prior.config exports still resolve.
Related
Ships the PyAutoNerves leg of PyAutoMind/draft/refactor/pyautofit/remove_eden_packaging_tooling.md.
PyAutoNerves still has some references to autoconf (autoconf folder, test_autoconf). Are we able to clean up the cause of this? Just feels like legacy or drift from the old name. There is also an egg info file i guess we can remove, and the bad folder (prevent that from being made?) and the build folder. There is also python files which I think are now defunct so assess if we can remove them, for example we no longer edenise code so that can go (but check its depedency in autofit) and also converting configs to .ini, generate_priors.py are all things which we used to set up initial configs years ago. Maybe break this task down into logical steps as I've throw different things at you. remove eden.yaml from autonerves. also sort this: #136 and I think we can close this? #100, chekc it again autobrain
Overview
After the
autoconf→autonervesrename (#135), PyAutoNerves still carries a legacy surface from the pre-rename era. A survey on 2026-07-23 separated it into three genuinely different problems, which this task ships as four independent steps.The
autoconfsightings are not tracked drift —git grep -ril autoconfover tracked files returns nothing. What remains is untracked working-copy cruft, one test that litters the repo root on every run, and four dormant one-shot scripts (two of which are already broken on their own imports).Plan
autoconf/,test_autoconf/,autoconf.egg-info/,build/,bad/) from the working copy — no commit needed, butautoconf/is not inert: it still resolves as a namespace package and shadows imports.bad/in the repo root on every run, rather than gitignoring the symptom.eden.yaml,scripts/edenise.py) — its driver importsautofit.tools.edenise, which no longer exists in PyAutoFit.scripts/,priors/,json_prior/generate.py) that bootstrapped the original configs years ago and is wired into no CI and no workspace.__version__bump that contradicts the wheels+tags-only release model.Detailed implementation plan
Affected repositories
Branch survey
testmode-env-drift→ descopedSuggested branch:
feature/purge-autoconf-era-legacyWorktree root:
~/Code/PyAutoLabs-wt/purge-autoconf-era-legacy/Survey findings
1. Untracked rename leftovers.
autoconf/,test_autoconf/,autoconf.egg-info/,build/andbad/are all present in the working copy and all untracked.autoconf/andtest_autoconf/contain only__pycache__— but they still import:So they shadow imports rather than sitting inert.
build/and*.egg-info/are already covered by.gitignore.2.
bad/is regenerated on every test run.test_autonerves/test_config.py:44setsBAD_PATH = "bad/path"andtest_path_emptycallsos.makedirs(BAD_PATH, exist_ok=True)into the CWD, never cleaning up. Deleting the directory without fixing the producer is a no-op.3. Dead one-shot tooling in
scripts/. Four drivers from the original config bootstrap, dormant for years, referenced from no CI and no workspace. Two are already broken:edenise.pyautofit.tools.edenise— removed from PyAutoFit, raisesImportErrorconvert_prior_configs.pyoyaml— not a declared dependencyconvert_config.py.ini→ YAML migration, long since rungenerate_priors.pyautonerves/json_prior/generate.py, whose only other consumer is its own testSteps
Step 0 — local cruft (no PR).
rm -rf autoconf/ test_autoconf/ autoconf.egg-info/ build/ bad/. All untracked; nothing to commit.Step 1 — stop
bad/regenerating.test_autonerves/test_config.py: movetest_path_does_not_existandtest_path_emptyontotmp_path; delete theBAD_PATHconstant and theremove_path()helper. Fixes the producer instead of gitignoring the symptom.Step 2 — remove dead EDEN tooling.
Delete
eden.yamlandscripts/edenise.py. This is the PyAutoNerves leg ofPyAutoMind/draft/refactor/pyautofit/remove_eden_packaging_tooling.md.PyAutoFit/eden.yamlis descoped — the branch survey found PyAutoFit claimed by thetestmode-env-driftworktree.autofit.tools.edeniseis already gone from PyAutoFit main, so only the orphaneden.yamlremains there: a one-file follow-up once that claim releases. The eden draft stays open until then.Step 3 — remove the one-shot config-migration tooling.
Delete
scripts/convert_config.py,scripts/convert_prior_configs.py,scripts/generate_priors.py,autonerves/json_prior/generate.py,test_autonerves/json_prior/test_generate.py, andpriors/subconfig.json(a stray output artifact ofgenerate_priors.py). Bothscripts/andpriors/disappear.Step 4 — sort PR #136.
Force-push the branch with only the test-fixture re-anchor (
test_workspace.py,test_setup_colab.py), revertingautonerves/__init__.pyto2026.7.9.1. Rewrite the PR body to describe a fixture cleanup — moving the colab mock off the yanked2026.7.6.649— not a version sync.Key files
test_autonerves/test_config.py—tmp_pathmigration (step 1)eden.yaml,scripts/edenise.py— deleted (step 2)scripts/,priors/,autonerves/json_prior/generate.py,test_autonerves/json_prior/test_generate.py— deleted (step 3)autonerves/__init__.py— reverted on the test: re-anchor the version-handshake test fixtures #136 branch (step 4)Guardrails
autonerves/json_prior/config.pyis NOT in scope. It is the live half — imported byautonerves/conf.pyand re-exported fromautonerves/__init__.py(default_prior,make_config_for_class,path_for_class,JSONPriorConfig). Only the siblinggenerate.pygoes.__version__on any library main. The release model is wheels+tags-only (PyAutoBuild#118/feat: setup_colab registry for all 6 notebook repos, GPU/env fixes, tag-pinned clones #120); main's__version__is deliberately stale, and hand-bumping it rebuilds the cascade engine feat: setup_colab registry for all 6 notebook repos, GPU/env fixes, tag-pinned clones #120 deleted. This is exactly why test: re-anchor the version-handshake test fixtures #136 is being trimmed rather than merged as-is.generate.pyis an import surface. Confirmed clean before deletion: no consumer in any PyAuto library, workspace, or doc outside its own script and test. Re-rungrep -rn "json_prior import generate\|json_prior.generate"before deleting; if a live consumer surfaces, re-scope.test_autonervessuite (151 tests at survey time) before each push.Testing approach
Full
pytest test_autonervesbefore each push. For step 1, additionally confirmbad/is absent from the repo root after a run. For step 3, confirmimport autonervesstill succeeds and the fourjson_prior.configexports still resolve.Related
PyAutoMind/draft/refactor/pyautofit/remove_eden_packaging_tooling.md.MERGED; ledger atPyAutoMind/complete/2026/04/workspace-version-config-check.md), simply never closed on GitHub.Original Prompt
Click to expand starting prompt
PyAutoNerves still has some references to autoconf (autoconf folder, test_autoconf). Are we able to clean up the cause of this? Just feels like legacy or drift from the old name. There is also an egg info file i guess we can remove, and the bad folder (prevent that from being made?) and the build folder. There is also python files which I think are now defunct so assess if we can remove them, for example we no longer edenise code so that can go (but check its depedency in autofit) and also converting configs to .ini, generate_priors.py are all things which we used to set up initial configs years ago. Maybe break this task down into logical steps as I've throw different things at you. remove eden.yaml from autonerves. also sort this: #136 and I think we can close this? #100, chekc it again autobrain