feat(bundles): add torch-free all-no-torch extra to lfx-bundles

[jordanrfrazier]

lfx[bundles] resolves to lfx-bundles[all], which pulls torch via the cuga and codeagents providers. Add a generated all-no-torch aggregate (all minus those two, 70 of 72 providers) so slim/CPU images can install the full provider set without torch. Generated by consolidate_bundles.py via the new TORCH_EXTRAS set; bump lfx-bundles 1.0.0 -> 1.1.0.

Add a guard test that re-resolves lfx-bundles[all-no-torch] and fails if any torch-family distribution appears, catching providers that pull torch transitively (e.g. via langchain/sentence-transformers) on future updates.

Summary by CodeRabbit

• New Features

  • Added a new all-no-torch install option for the bundle, making it easier to install the full set of extras without torch-related dependencies.
  • Updated the bundle metadata to clearly describe both all and all-no-torch options.

• Bug Fixes

  • Regenerates the torch-free aggregate consistently during bundle updates.
  • Added checks to ensure the new install option stays free of torch-family packages.

[coderabbitai]

Important

Review skipped

Auto incremental reviews are disabled on this repository.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: c42a5241-6b4a-47cc-a358-8493ceb4976a

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch lfx-bundles-all-no-torch-extra

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

✅ Test Coverage Advisor

No source changes detected without accompanying tests. Thanks for keeping coverage up! 🎉

Advisory check only — never blocks merge.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 59.62%. Comparing base (fa78cdd) to head (1dab9de).
⚠️ Report is 1 commits behind head on release-1.11.0.

Additional details and impacted files

@@                Coverage Diff                 @@
##           release-1.11.0   #13886      +/-   ##
==================================================
+ Coverage           59.58%   59.62%   +0.04%     
==================================================
  Files                2349     2349              
  Lines              224820   224820              
  Branches            33523    33471      -52     
==================================================
+ Hits               133954   134059     +105     
+ Misses              89347    89242     -105     
  Partials             1519     1519              

Flag	Coverage Δ
backend	67.26% <ø> (-0.10%)	⬇️
frontend	58.39% <ø> (+0.09%)	⬆️
lfx	56.47% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.
see 34 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/bundles/lfx-bundles/tests/test_all_no_torch_extra.py`:
- Around line 117-118: The guard in the test helper that runs uv pip compile is
treating every non-zero exit as an offline/registry issue, which hides real
resolver problems. Update the failure handling in the compile check so it only
calls pytest.skip for clearly transient environment failures (for example,
network/registry unavailability) and uses pytest.fail for unexpected resolver
errors such as invalid extras or dependency conflicts. Keep the logic localized
around the subprocess returncode handling in test_all_no_torch_extra.py so the
test still verifies the real behavior instead of masking failures.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

���️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 92e6e42c-608f-499e-8391-fa30133019f6

📥 Commits

Reviewing files that changed from the base of the PR and between fa78cdd and 3fc445a.

📒 Files selected for processing (3)

• scripts/migrate/consolidate_bundles.py
• src/bundles/lfx-bundles/pyproject.toml
• src/bundles/lfx-bundles/tests/test_all_no_torch_extra.py

[coderabbitai]

🎯 Functional Correctness | 🟠 Major | ⚡ Quick win

Don't skip every failed resolve.

Any non-zero uv pip compile exit currently turns this guard green, so an invalid extra or dependency conflict is indistinguishable from “offline/registry unavailable.” Only skip known transient environment failures; unexpected resolver errors should pytest.fail(...).

Suggested fix

     if proc.returncode != 0:
-        pytest.skip(f"could not resolve all-no-torch (offline / registry unavailable?):\n{proc.stderr[-800:]}")
+        stderr = proc.stderr.lower()
+        transient_markers = (
+            "offline",
+            "registry unavailable",
+            "temporary failure",
+            "timed out",
+            "connection",
+            "network",
+            "dns",
+            "tls",
+            "certificate",
+        )
+        if any(marker in stderr for marker in transient_markers):
+            pytest.skip(f"could not resolve all-no-torch (offline / registry unavailable?):\n{proc.stderr[-800:]}")
+        pytest.fail(f"uv failed to resolve all-no-torch:\n{proc.stderr[-800:]}")

As per coding guidelines, "verify the tests actually cover the new or changed behavior rather than acting as placeholders."

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	if proc.returncode != 0:
	pytest.skip(f"could not resolve all-no-torch (offline / registry unavailable?):\n{proc.stderr[-800:]}")
	if proc.returncode != 0:
	stderr = proc.stderr.lower()
	transient_markers = (
	"offline",
	"registry unavailable",
	"temporary failure",
	"timed out",
	"connection",
	"network",
	"dns",
	"tls",
	"certificate",
	)
	if any(marker in stderr for marker in transient_markers):
	pytest.skip(f"could not resolve all-no-torch (offline / registry unavailable?):\n{proc.stderr[-800:]}")
	pytest.fail(f"uv failed to resolve all-no-torch:\n{proc.stderr[-800:]}")

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/bundles/lfx-bundles/tests/test_all_no_torch_extra.py` around lines 117 -
118, The guard in the test helper that runs uv pip compile is treating every
non-zero exit as an offline/registry issue, which hides real resolver problems.
Update the failure handling in the compile check so it only calls pytest.skip
for clearly transient environment failures (for example, network/registry
unavailability) and uses pytest.fail for unexpected resolver errors such as
invalid extras or dependency conflicts. Keep the logic localized around the
subprocess returncode handling in test_all_no_torch_extra.py so the test still
verifies the real behavior instead of masking failures.

Source: Coding guidelines

[github-actions]

Frontend Unit Test Coverage Report

Coverage Summary

Lines	Statements	Branches	Functions
	44.51% (60131/135076)	69.29% (8176/11798)	42.78% (1356/3169)

Unit Test Results

Tests	Skipped	Failures	Errors	Time
5096	0 💤	0 ❌	0 🔥	14m 41s ⏱️

[erichare]

LGTM @jordanrfrazier , a couple small comments but not blockers. Maybe look at them and the code rabbit comments but this is much needed to have a torchless “full” install.

Findings
1. scripts/migrate/consolidate_bundles.py:182 — Comment references a non-existent all-torch aggregate. The new comment reads “Excluded from the generated all-torch aggregate so it resolves torch-free.” There is no all-torch extra; the providers in TORCH_EXTRAS are excluded from all-no-torch. A future maintainer reading this could be misled about which aggregate the exclusion applies to. Should read all-no-torch.
2. src/bundles/lfx-bundles/tests/test_all_no_torch_extra.py:108 — The “real guard” resolution test skips on any non-zero exit, so a genuine torch regression can pass silently. test_all_no_torch_resolves_without_torch calls pytest.skip(...) whenever proc.returncode != 0. The intent (per the docstring) is to tolerate offline/registry-unavailable CI, but a non-zero exit can also mean a real resolution conflict — including one introduced by the very provider change this test is meant to catch. The torch assertion only runs on a clean exit, so in a CI environment where this resolve doesn’t fully succeed, the guard is effectively a no-op and goes green. This is an intentional tradeoff, but worth confirming the guard actually runs (not just skips) in CI — otherwise the structural test is the only real protection. Consider logging/surfacing skips so a perpetually-skipped guard is visible.

[github-actions]

Build successful! ✅
Deploying docs draft.
Deploy successful! View draft