Public scan — anyone with this URL can view this analysis. Sign up to track your own repos privately, run scheduled re-scans, and get AI fix prompts via your dashboard.

vllm-project/vllm

https://github.com/vllm-project/vllm · scanned 2026-06-16 00:55 UTC (2 months, 1 week ago) · 10 languages

2224 raw signals (233 security + 1991 graph) 23rd percentile · Python · huge (>500K LoC)

File as issue Image

UNIFIED Repobility · multi-layer engine · AI coders

Complete repo analysis

Last scanned 2 months, 1 week ago · v2 · last Δ -21.4 (diff) · 1056 actionable findings from 2 signal sources. 202 repeated signals grouped for readability. Security checks, system graph analysis, and verified AI-agent feedback are merged into one review queue.

JSON

100.0% cov · 1011 findings

Score breakdown â 2026-05-17-v4 calibration-aware

Component	Sub-score	Weight	Contribution
`structure_score`	60.0	0.15	9.00
`security_score`	0.0	0.25	0.00
`testing_score`	100.0	0.20	20.00
`documentation_score`	86.0	0.15	12.90
`practices_score`	65.0	0.15	9.75
`code_quality`	70.0	0.10	7.00
Overall		1.00	58.6

Calibrated penalty buckets (security_score): agent: 3.2 Â· docker: 86.6 Â· threat: 75.1

Severity distribution — click a segment to filter

Active filters: excluding tests × Reset all

Severity: Critical 1 High 63 Medium 62 Low 427 9-Layer: Software Security Quality Integrity Frontend Hardware Data Network Cicd Source: Security checks 45 System graph 1011 Crowd 0 Layer: Cicd 16 Security 69 Software 29 Quality 888 Network 2 Hardware 1 Frontend 1 Api 50

Exclude dismissed / FP Exclude test files

Bug-class explainers. Each card groups findings of the same shape — these are the patterns most likely to ship to prod and reappear in future scans unless you systematically fix the cause, not just the instance.

Duplicates & near-duplicates 5 findings

What it is: Same function copy-pasted into multiple modules with minor variations.

Why it matters: Each copy drifts independently — bug fixes apply to one, miss the others.

How AI causes it: AI completes the same pattern in each file rather than refactoring to a shared helper.

Fix approach: Extract the duplicated logic into the most general module both call sites already import. Add tests at the helper level.

5 matching findings on this repo

low Near-duplicate function bodies in 2 places repo-level
low Near-duplicate function bodies in 3 places repo-level
low Near-duplicate function bodies in 6 places
low Near-duplicate function bodies in 4 places repo-level
low Near-duplicate function bodies in 5 places

View all duplicates & near-duplicates findings →

Commented-out code 491 findings

What it is: Lines of source that were intentionally disabled but never deleted.

Why it matters: Git already remembers history — commented code rots, becomes wrong, and adds noise to diffs.

How AI causes it: AI sometimes comments out broken code instead of fixing it. Reviewers approve out of inertia.

Fix approach: Delete. Trust `git log`. If you really need to remember, save it in a notes file under `docs/`.

12 matching findings on this repo

info Commented-code block (5 lines) in tests/conftest.py:1271
info Commented-code block (5 lines) in tests/utils.py:316
info Commented-code block (5 lines) in tests/compile/test_inductor_fallback_allow_li…
info Commented-code block (7 lines) in tests/compile/test_graph_partition.py:655
info Commented-code block (8 lines) in tests/compile/test_config.py:645
info Commented-code block (7 lines) in tests/compile/passes/test_fusion.py:319
info Commented-code block (5 lines) in tests/compile/passes/test_rope_kvcache_fusion…
info Commented-code block (8 lines) in tests/compile/passes/test_mla_rope_kvcache_ca…
info Commented-code block (5 lines) in tests/compile/passes/distributed/test_fusion_…
info Commented-code block (5 lines) in tests/compile/fusions_e2e/conftest.py:128
info Commented-code block (5 lines) in tests/compile/fusions_e2e/models.py:173
info Commented-code block (6 lines) in tests/compile/fullgraph/test_multimodal_compi…

View all commented-out code findings →

Config drift 28 findings

What it is: Settings duplicated across env files, Docker compose, K8s, and code defaults, all with slightly different values.

Why it matters: Production behaviour depends on whichever copy your loader reads first. Subtle bugs in staging that don't reproduce in dev.

How AI causes it: AI writes new config from memory rather than reading the existing source.

Fix approach: Pick one source of truth (env vars + a settings module). Have every other place import from there. Lint for duplicates in CI.

12 matching findings on this repo

low Very large file: tests/test_config.py (1559 lines)
low Very large file: vllm/config/model.py (2227 lines)
low Very large file: vllm/config/speculative.py (1141 lines)
low Very large file: vllm/config/compilation.py (1514 lines)
low Very large file: vllm/config/vllm.py (2314 lines)
medium Runtime service client appears to use placeholder configuration
low Old/deprecated-named symbol `set_splitting_ops_for_v1` in tests/compile/test_co…
info Commented-code block (8 lines) in tests/compile/test_config.py:645
low Old/deprecated-named symbol `deepseek_vl_v2` in vllm/transformers_utils/config.…
low Old/deprecated-named symbol `deepseek_v2` in vllm/transformers_utils/model_arch…
info Commented-code block (6 lines) in vllm/transformers_utils/configs/colpali.py:42
info Commented-code block (6 lines) in vllm/transformers_utils/configs/funaudiochat.…

View all config drift findings →

For AI agents: Voting guide (TP/FP) MCP manifest Stdio wrapper SARIF Integrate Findings queue Vote TP/FP on findings to calibrate the engine.

For AI agents + API integrations

Email me when this repo regresses

Free. We re-scan periodically; new criticals → your inbox. No signup required for the scan itself.

API access

This page is publicly accessible at: https://repobility.com/scan/cea585c0-944b-4614-9116-8abca2930bfe/

To check status programmatically (no auth required):

curl -s https://repobility.com/api/v1/public/scan/cea585c0-944b-4614-9116-8abca2930bfe/

Important — please don't re-submit the same URL repeatedly. The submission endpoint is idempotent: re-submitting the same git URL returns this same scan_token, not a new one. To re-scan this repo, sign up free and use the dashboard.