fix(security): floor litellm/wandb/lxml past known CVEs; relax stale click cap#1507
Open
nicgupta-nvidia wants to merge 6 commits into
Open
fix(security): floor litellm/wandb/lxml past known CVEs; relax stale click cap#1507nicgupta-nvidia wants to merge 6 commits into
nicgupta-nvidia wants to merge 6 commits into
Conversation
…click cap - litellm[caching] 1.83.14 -> 1.84.10: GHSA-4xpc-pv4p-pm3w (Critical) — 1.83.x leaks the API key to an arbitrary attacker-controlled Host header; fixed in 1.84.0. Minimal exact-pin jump; resolves with the existing httpx[http2]>=0.28.1 override (litellm 1.84.10 needs httpx>=0.28.0). - wandb -> >=0.27.1: the bundled wandb-core Go binary in older wheels ships golang.org/x/crypto 0.50.0 + Go 1.26.2 stdlib with 7 Critical / 13 High CVEs (incl. GHSA-x527-x647-q7gg et al.); 0.27.1 is the first release embedding patched x/crypto 0.52.0 (verified on both arches). - click < 8.2.0 cap removed + typer >= 0.16: the cap guarded against the typer/click-8.2 make_metavar break (ai-dynamo/dynamo#1039, closed 2025-06-26, fixed in typer >=0.16); wandb>=0.27.1 requires click>=8.2, and requires-python >=3.10 satisfies click 8.2's floor. - lxml -> >=6.1.0 (stem extra): GHSA-vfmq-68hx-4jfw (High). Validation: uv pip compile of core+pipeline (py3.10, with the pyproject overrides) resolves cleanly — litellm 1.84.10 / wandb 0.28.0 / click 8.4.2 / typer 0.26.8 / httpx 0.28.1; stem extra resolves with lxml 6.1.1. Runtime smoke on the resolved set: litellm/wandb import clean; typer --help rendering (the exact make_metavar crash path) passes under click 8.4. Signed-off-by: Nick Gupta <nicgupta@nvidia.com>
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Enterprise Run ID: 📒 Files selected for processing (1)
💤 Files with no reviewable changes (1)
📝 WalkthroughWalkthroughDependency constraints were updated across core, pipeline, stem, and test requirements, alongside a ChangesDependency security and compatibility updates
Estimated code review effort: 3 (Moderate) | ~30 minutes Suggested reviewers: 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Contributor
|
Note Unit test generation is a beta feature. Expect some limitations and changes as we gather feedback and continue to improve it. Generating unit tests... This may take up to 20 minutes. |
2 similar comments
Contributor
|
Note Unit test generation is a beta feature. Expect some limitations and changes as we gather feedback and continue to improve it. Generating unit tests... This may take up to 20 minutes. |
Contributor
|
Note Unit test generation is a beta feature. Expect some limitations and changes as we gather feedback and continue to improve it. Generating unit tests... This may take up to 20 minutes. |
Contributor
|
✅ Unit tests committed locally. Commit: |
Contributor
|
✅ Unit tests committed locally. Commit: |
Contributor
|
✅ Created PR with unit tests: #1508 |
CodeRabbit's test generation ran twice and committed two near-duplicate suites (test_dependency_pins.py + test_requirements_versions.py) covering the same pins. Keep the more robust one (operator-keyed specifier parsing instead of next(iter(specifier)), which is order-fragile on multi-spec requirements) and graft the three tests unique to the deleted file: pyproject stale-comment guards, pipeline-lines-parseable, and the wandb/typer click-comment consistency check. Also: - fix the copyright year (2026, not 2025) - add tomli (python_version < 3.11) to common-tests.txt so the pyproject override tests actually RUN on the CI's Python 3.10 instead of silently skipping (tomllib is stdlib only from 3.11) - add the missing trailing newline that failed the pre-commit end-of-file-fixer hook on the generated files 17 tests pass on Python 3.10 with the CI's -m 'not gpu' selection; pre-commit (pinned ruff) clean. Signed-off-by: Nick Gupta <nicgupta@nvidia.com>
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
🧹 Nitpick comments (1)
tests/test_requirements_versions.py (1)
95-96: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueExtract a shared helper for specifier-floor extraction.
The
{spec.operator: spec.version for spec in req.specifier}pattern is repeated across four test methods. A small helper reduces duplication and keeps the assertions consistent if the extraction logic ever needs to change (e.g., handling multiple specifiers of the same operator).♻️ Proposed helper
+def _get_specifier(req: Requirement, operator: str) -> Version: + """Return the parsed Version for the given specifier operator, asserting it's present.""" + specs = {spec.operator: spec.version for spec in req.specifier} + assert operator in specs, f"expected a '{operator}' specifier for {req.name}, got {req.specifier}" + return Version(specs[operator]) + + class TestCoreRequirements: ... def test_litellm_pin_fixes_ghsa_4xpc_pv4p_pm3w(self): req, comment = _find_requirement(CORE_REQUIREMENTS, "litellm") assert "caching" in req.extras, "litellm[caching] extra must be preserved" - - # Must be pinned to an exact version (== specifier) so the resolver is deterministic. - specs = {spec.operator: spec.version for spec in req.specifier} - assert "==" in specs, f"expected an exact pin for litellm, got specifier {req.specifier}" - - pinned_version = Version(specs["=="]) + pinned_version = _get_specifier(req, "==")Also applies to: 113-114, 156-157, 178-179
🤖 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 `@tests/test_requirements_versions.py` around lines 95 - 96, The specifier extraction logic is duplicated across multiple test methods in the requirements version tests, so add a shared helper for turning a requirement’s specifiers into a usable mapping or floor value. Update the affected assertions in the test class that currently build specs from req.specifier so they call this helper instead, keeping the exact-pin checks unchanged while centralizing the extraction behavior in one place.
🤖 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.
Nitpick comments:
In `@tests/test_requirements_versions.py`:
- Around line 95-96: The specifier extraction logic is duplicated across
multiple test methods in the requirements version tests, so add a shared helper
for turning a requirement’s specifiers into a usable mapping or floor value.
Update the affected assertions in the test class that currently build specs from
req.specifier so they call this helper instead, keeping the exact-pin checks
unchanged while centralizing the extraction behavior in one place.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 0ea160ee-449e-4976-b41b-78cdb3607f1a
📒 Files selected for processing (2)
requirements/common-tests.txttests/test_requirements_versions.py
…ed deps
test_requirements_versions.py only asserts the pins statically. Add functional
coverage that drives litellm 1.84.10, typer/click, and wandb through NeMo-Skills'
own code paths (CPU-only, hermetic — no sandbox, no live endpoint, no API keys)
so a resolve to a behavior-divergent version fails CI, not a production run:
* litellm: OpenAIModel.litellm_kwargs binds api_key to the configured
base_url/api_base only (GHSA-4xpc-pv4p-pm3w regression), generate_async
calls litellm.acompletion with those credentials and parses the 1.84
response, and the imported litellm exception/type surface still exists.
* typer/click: ns CLI --help + per-command help render Parameter.make_metavar
(the click 8.2 break typer>=0.16 fixes) and unknown commands are usage errors.
* wandb: log_random_samples matches the wandb 0.28 init/save/summary/finish
contract, and a real offline init/finish cycle runs with no account.
* lxml: importorskip-guarded real parse (optional stem extra).
Signed-off-by: Nick Gupta <nicgupta@nvidia.com>
Bring in #1509 (drop deleted vedas pin) so the sandbox image build — and therefore this PR's CI — is unblocked. Signed-off-by: Nick Gupta <nicgupta@nvidia.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Four dependency-only security fixes (no code changes):
litellm[caching]==1.83.14==1.84.10Hostheader; fixed in 1.84.0. Minimal exact-pin jump.wandb>=0.27.1wandb-coreGo binary built with golang.org/x/crypto 0.50.0 + Go 1.26.2 stdlib carrying 7 Critical / 13 High CVEs (x/crypto, Go stdlib, x/net, go-git, go-billy). 0.27.1 is the first release embedding patched x/crypto 0.52.0 (verified via Go buildinfo on both arches).click< 8.2.0make_metavarbreak (ai-dynamo/dynamo#1039) — closed 2025-06-26, fixed in typer ≥0.16.wandb>=0.27.1hard-requiresclick>=8.2, and this repo'srequires-python >=3.10satisfies click 8.2's floor.typer>= 0.13>= 0.16lxml(stem extra)>=6.1.0Also updates the
pyproject.tomloverride comment that referenced the oldlitellm==1.83.14httpx pin.Validation
uv pip compile core/requirements.txt requirements/pipeline.txt(Python 3.10, with the repo'shttpx[http2]>=0.28.1/urllib3>=2.6.3overrides) resolves cleanly: litellm 1.84.10, wandb 0.28.0, click 8.4.2, typer 0.26.8, httpx 0.28.1.uv pip compile requirements/stem.txtresolves with lxml 6.1.1.litellmandwandbimport clean; a typer CLI--helprender (the exactParameter.make_metavarcrash path from dynamo#1039) passes under click 8.4.Tests
tests/test_dependency_functional.pyadds functional coverage on top of the static pin guard intests/test_requirements_versions.py— it drives the bumped packages through NeMo-Skills' own code paths (CPU-only, hermetic: no sandbox, no live endpoint, no API keys), so a resolve to a behavior-divergent version fails CI rather than a production run. 13 tests, all green locally on the resolved set (litellm 1.84.10 / typer 0.26.8 / click 8.4.2 / wandb 0.28.0 / lxml 6.1.1):OpenAIModel.litellm_kwargsbinds the api_key to the configuredbase_url/api_baseonly (a regression guard for GHSA-4xpc-pv4p-pm3w, the key-to-arbitrary-Host leak this bump fixes);generate_asynccallslitellm.acompletionwith those credentials and parses the 1.84 response; the imported litellm exception/type surface still exists.nsCLI--help+ per-command help renderParameter.make_metavar(the click 8.2 breaktyper>=0.16fixes), and unknown commands are usage errors.log_random_samplesmatches the wandb 0.28init/save/summary/finishcontract, plus a real offlineinit/finishcycle with no account.importorskip-guarded real parse.Coverage caveat (lxml):
lxmlships only in the optionalstemextra, not in.[dev], sotest_lxml_parses_htmlis skipped in the existingunit-testsCI job (which installs-e .[dev]) — it does not give live CI signal there. Thelxml>=6.1.0floor remains covered by the static assertion intest_requirements_versions.py; to exercise lxml functionally in CI, add.[stem]to the unit-tests install. The other 12 tests run and give real signal in the current job.Notes
nltk(GHSA-p4gq-832x-fm9v) is intentionally NOT floored here: the advisory designates no patched version (first_patched: null), so a floor would only silence scanners without a verified fix.Summary by CodeRabbit
Bug Fixes
litellm[caching],wandb, andlxml.clickupper pin and raising thetyperminimum for compatibility.Tests
wandbbehavior, and HTML parsing.Documentation