Public report — supervision, published 2 Aug 2026.
Concrete security findings (CVE IDs, secret matches, dependency versions) are hidden in this version;
ask the repo owner for the full report.
198findings with an exact file:lineof 207 — the remainder are repo-wide signals (a dimension-level measurement, not a single line); open any file:line and verify
22/94dimensions across the health lenses41155 LoC — wide & deep
Executive summary
Read through the Preview lens: this repo is pre-1.0 / in development, so the colour bands are relaxed to what a preview needs — *green* means good enough for a preview, not yet production-stable. Code correctness and security stay near-strict even here; the score itself is absolute and comparable across repos.
roboflow/supervision is sound in substance but carries real gaps (61%). It is not in crisis, but the issues below raise the cost of changing it — friction its consumers ultimately inherit.
It is strongest in Architecture (100%) — the structure is clean and changes stay contained. Code Health (79%) is solid too.
The area that most needs attention is Security (55%) — exposure to security and compliance incidents is elevated. Maturity (59%) is the next concern — onboarding is slow — key decisions and the architecture aren't written down, so contributors have to reverse-engineer the intent.
Leadership focus, highest impact first: Record significant decisions one document per decision (Architecture documentation); 'Testing' section to the root README (Documentation (README)); 10 High finding(s) (Static Analysis (SAST)).
For scale: Medium (~41,155 production lines); rebuilding it from scratch would take roughly ~0.6 person-years (~1–2 engineers). Approximate, ±~30%.
It builds on a genuinely strong Architecture foundation (100%); the priorities above are the highest-leverage way to bring the rest up to that level.
How the score is built — each lens's share of the headlineWidth is the lens's weight in the worst-heaviest fold (the weakest area pulls hardest); colour is that lens's own band. A lens fixes the score in proportion to its width.
A full-fidelity diff against the previous run's complete recorded findings — line-move tolerant: a finding that only shifted line counts as unchanged, only genuinely new titles/files surface here.
0.8× (at 61% quality) — the last 20% of quality is most of the work
Size & shape
Medium · effort split not classified (source measured from disk; the effort-tier breakdown is a C#-only syntax walk)
This codebase represents roughly ~0.6 person-years of build effort (about ~€88,000 to rebuild). Its weakest lens is Security at 55% — the part of that asset most exposed by the findings below.
How we model this: boilerplate at a scaffolding rate + logic × domain Standard (×1.0) — standard service × a 0.8× quality factor, at €60–95/h; indicative, ±~30% · size measured directly from source · effort from total production LoC as straight-line logic (the tier split is a C#-only syntax walk), a conservative lower bound. Indicative only — most sensitive to the hourly rate and the domain tier (both tunable in config).
Top priorities
The highest-leverage moves; the full ranked list is in the Roadmap below.
1
Record significant decisions one document per decision — dated, stating the context, the decision and its consequences — and keep them together wherever your design docs already live (a conventional `docs/adr/` tree with `NNNN-title.md` names is the most discoverable form).
Value concentrated against a weak lens · Medium · Value at risk
This is a Medium asset (~0.6 person-years to rebuild), and its weakest lens is Security at 55%. The operational and business risk on an asset this size concentrates there — that's where remediation buys the most protection.
→ Direct remediation budget at Security first — highest risk-reduction per euro on an asset this size.
Highest-leverage move · Medium · Leverage
Of everything flagged, the best return on effort is: Record significant decisions one document per decision — dated, stating the context, the decision and its consequences — and keep them together wherever your design docs already live (a conventional `docs/adr/` tree with `NNNN-title.md` names is the most discoverable form). The rest can wait behind it.
Evidence: priority ranking: top of 5 ranked by impact/effort
→ Record significant decisions one document per decision — dated, stating the context, the decision and its consequences — and keep them together wherever your design docs already live (a conventional `docs/adr/` tree with `NNNN-title.md` names is the most discoverable form).
Architecture — module dependency matrix
149 modules, 8 dependencies — every dependency points down the layering, so there are no cycles. Rows and columns are the same modules, ordered so that a module only depends on ones above it. A cell means the row depends on the column, and its number is how many type pairs create that dependency. Read one thing: is anything above the diagonal? A mark there is a dependency cycle. (A cycle is all this shows — an unusual but cycle-free dependency sits below the diagonal like any other.)
Findings mapped to OWASP categories; the specific CVEs/secrets are in the Security dimension cards below and findings.md (redacted only on the public version of this report).
OWASP category
Findings
Severity
A03:2021 — Injection
11
High / Critical
A06:2021 — Vulnerable & Outdated Components
7
High / Critical
A02:2021 — Cryptographic Failures
2
High / Critical
Roadmap
Begin by establishing a central repository for architecture decisions, ensuring each is dated and clearly states the context and consequences. Next, update the root README to include a dedicated section on how to run the test suite. Address the ten high-priority static analysis findings, focusing first on the build and CI configuration files. Finally, integrate a security scanning step into the CI pipeline to prevent regressions, and verify that deployment protections and approval gates are correctly configured to prevent bad builds from reaching users.
Ranked by impact ÷ effort. "Helps" is the estimated gain on the 0–100 health score.
Do this
Helps
Effort
Dimension
Record significant decisions one document per decision — dated, stating the context, the decision and its consequences — and keep them together wherever your design docs already live (a conventional `docs/adr/` tree with `NNNN-title.md` names is the most discoverable form).
Watchdog is a deep, periodic assessment — run each sprint, monthly, or quarterly, taking the time to go wider and deeper than a quick check and surfacing in one coherent report what you'd otherwise piece together from a dozen separate tools. It scores deterministically: the same commit yields the same score, every run. 21 of 22 evaluated dimensions are computed purely by tools and static analysis (confidence 1.0); 1 documentation/naming judgement(s) are LLM-assisted and labelled advisory. Overall confidence is 0.5 — the weighted average across measured dimensions; it falls as more of the score leans on LLM-assisted judgement and rises when it's fully tool-backed.
Every figure here is one of three kinds, and we label which: ✓ Measured — a deterministic fact (LoC, complexity, coverage); ~ Modeled — an estimate from a stated model (cost, effort, value-at-risk), always a range with its assumptions, never a precise fact; ◐ Advisory — an LLM prose judgement. We never present a modelled estimate as if it were measured. Perfect or absent scores carry their provenance too (ADR-0011): ✓ Tool-verified means the property itself was measured across the surface; ○ Nothing flagged means the probes came back clean — a claim bounded by what a repository can show; ⊘ Not evidenced means a working control (a tested restore, an automated rollback) showed no positive evidence — absence of evidence is not evidence of a control, so it's excluded from the score rather than awarded a spurious 10; ◐ Sampled · advisory marks an LLM verdict over a bounded sample — advisory, never a deterministic measurement.
What we checked — 22 dimensions across the health lenses
Each chip is a dimension scored from real signals across architecture, testing, dependencies, security & compliance, documentation, git-history and code quality — in one coherent pass. A surface report typically covers a handful.
How to trust any code-health report — three questions
Can you open the finding? Real findings cite a repo-relative file and line you can open at the cited line — never an absolute scratch path. Here, 198 of 207 do; the remainder are repo-wide signals — a dimension-level measurement, not a single line. (Every path in this report is repo-relative by construction: paths are normalized at the producer and the report is rejected if any rooted path leaks through.)
Is there a tool behind the number? Every score below names the method that produced it — Roslyn, git, a scanner, or (for a handful of documentation/naming dimensions) an LLM labelled sampled · advisory — not a narrative.
Does re-running give the same result? Run it again on the same commit and the score — and this report, byte for byte — is identical. A report whose numbers move between runs is describing the run, not the code.
This report answers yes to all three. That's the bar to hold any assessment to.
Tools & methods
The actual versions used this run (captured at analysis time) — re-run on the same commit for the identical score.
Method
Backs
Version
Evaluator
Roslyn static analysis
Complexity, cohesion, coupling, dead code, API surface, layering
What ran differently this time — a tool absent, degraded, or that fell back to an estimate. Named openly, not folded silently into the scores. A degraded run also records its exact cause in diagnostics.md.
D19 Documentation Quality — LLM provider failed — The model provider returned an unusable result, so this LLM-assisted dimension fell back to a measurement gap (confidence 0) rather than a penalty. Re-run with a reachable provider to score it.
D30 Dependency Vulnerabilities — scanner not present in this environment — The backing tool was not installed where this scan ran, so this dimension was not scored. Install the tool (or run in the hosted environment, where it is always present) for a graded result.
Repo exclusion declarations: 1 pattern(s) declared (.gitattributes linguist-generated/vendored, .editorconfig generated_code) excluded 0 source file(s) from code-quality scoring. Declarations are the repo's own visible statement that a tree is machine-written or vendored — auditable in any diff, honored by GitHub the same way.
Limitations & what we did not check
Watchdog assesses the repository exactly as committed, and only the repository. By design it does not reach outside the source tree: the live cloud account, the running CI/CD pipeline, the host's branch-protection and approval rules, the production configuration, or a restore actually exercised against a backup are all out of scope. That boundary is a feature, not a gap — a repo-relative, deterministic scan re-runs identically on any commit and every finding opens at a real file and line, where a live audit can neither be reproduced nor traced. The visible consequence is that controls which leave no in-repo evidence are reported as "not evidenced" and excluded from the score rather than awarded a number a static scan cannot justify.
Per-dimension blind spots
For each dimension that was measured, what a static, repo-only scan structurally cannot see — the honest edge of the measurement, not a failure of it.
D1 Cyclomatic Complexity: Cyclomatic complexity counts branches statically — it cannot tell an essential decision tree from accidental tangle, nor see complexity that lives in data or configuration (large switch-case token tables, DSL lexers/parsers, data-as-code rule tables) rather than control flow: a tokenizer's many single-character cases read as high complexity though each branch is trivial.
D2 Cognitive Complexity: Cognitive-complexity heuristics approximate how hard code is to follow; genuine domain difficulty and well-named intent that eases reading are not captured.
D3 God Classes: "God class" is sized by members and responsibilities visible in the type — a deliberately broad facade over a coherent subsystem can read the same as an accidental grab-bag. For front-end JS the file-length check is cohesion-aware (a single-responsibility module — one class/IIFE — earns a 3× threshold), but cohesion is approximated from top-level declarations, not true dependency structure.
D4 Code Duplication: Duplication is token-similarity (jscpd) — it finds copy-paste, not semantic duplication expressed differently. Committed machine-written code (scaffolded migrations, designer/codegen output, protobuf/OpenAPI stubs, model snapshots) is EXCLUDED — its repetition is the tool's, not the team's — so the score reflects hand-written duplication only.
D13 Secret Scanning: Secret detection is signature- and entropy-based on the current tree — a secret that does not match a known pattern, or one already rotated, will not be flagged (a clean scan is "nothing matched", not "no secrets exist").
D15 Churn × Complexity Hotspots: Churn hotspots come from git history — a freshly imported or squashed repository has no churn signal, and recent rewrites can mask a historically risky file.
D16 Bus Factor: Bus-factor is a time-decayed model of commit attribution (who has recently, repeatedly worked a file), not comprehension — pairing, review and reading-without-committing spread knowledge it can't see; bot commits and shared accounts still distort it.
D21 Naming Consistency: Naming quality is an LLM judgement over a bounded sample — it assesses clarity/consistency of the names it sees, not domain-correctness, and is advisory.
D28 Secrets (history): Secret-history scanning sweeps the git log for known patterns — a secret that predates the available history, or never matched a signature, is not found (clean means "nothing matched in the history we can see").
D29 Static Analysis (SAST): SAST findings are pattern-based (semgrep) — it finds classes of bug it has rules for; logic flaws, auth/authorization gaps and issues needing runtime context are out of reach (and clean means "no rule matched").
D34 Knowledge Freshness: Freshness is decayed commit RECENCY, not comprehension — code read often but rarely committed reads as orphaned, and stable code that genuinely needs no changes is penalised the same as forgotten code; bot/squash commits distort it like the bus factor.
D35 Change Coupling: Change coupling is co-change in COMMITS — files split across separate commits, or coupled only through a shared config/build step, read as uncoupled, and a sweeping commit (rename/format) is excluded so it doesn't couple everything. It shows that files change together, not WHY: a high coupling can be a healthy cohesive pair as readily as a hidden leak.
M4 Documentation accuracy: Onboarding quality is an LLM read of the docs/setup present — it cannot run the onboarding or measure how long a real new joiner takes; the verdict is sampled and advisory.
P4 Deployment & Rollback: Approval/branch-protection rules live in repository settings the scan cannot see — only their in-repo evidence (config files, workflows) is checked, so a control enforced purely in the host's settings reads as "not evidenced".
P6 Release Hygiene: Rollback/observability controls are inferred from repo artefacts (pipelines, dashboards-as-code) — controls configured in external tooling, with no in-repo trace, cannot be credited.
The LLM boundary
LLM-set scores this run (2): D21, M4 (model: Local LLM). For these, a model reads a bounded sample and sets the numeric score (documentation, ADR quality, naming, comment value, onboarding) — D25 sets the ADR-conformance fraction over sampled code, D22 judges API accuracy over a sample. These are sampled and advisory by design: they vary at the margins between runs and are never a deterministic measurement. Every other score in this report is tool-computed at confidence 1.0.
What it measures: How tangled the control flow is — methods with many branches are hard to test and change.
Method: Cyclomatic complexity per method (1 + decision points), computed exhaustively across production source; test projects separated by convention. Deterministic.
+ 16 more group(s) — more in Appendix A; the complete list is findings.md.
What to do
Resolve the 3 internal.process_roboflow_result (cyclomatic 21) finding(s) in Cyclomatic Complexity — start with internal.py (3). — One of this dimension's main actionable groups (3 warning-level).
Resolve the 1 KeyPoints.select (cyclomatic 33) finding(s) in Cyclomatic Complexity — start with core.py. — One of this dimension's main actionable groups (1 warning-level).
Resolve the 1 _video._mux_audio (cyclomatic 31) finding(s) in Cyclomatic Complexity — start with _video.py. — One of this dimension's main actionable groups (1 warning-level).
Enforce Cyclomatic Complexity in CI to reach Verified (currently Documented). — Hardens enforcement from Documented toward Prevented — provenance only; does not change the score.
Detailed fixes: d1_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: How hard the code is for a person to follow, beyond raw branching.
Method: Cognitive complexity per method (Sonar-style nesting-penalized score), computed exhaustively over production code, excluding test projects. Deterministic.
+ 55 more group(s) — more in Appendix A; the complete list is findings.md.
What to do
Resolve the 3 internal.process_roboflow_result (cognitive 34) finding(s) in Cognitive Complexity — start with internal.py (3). — One of this dimension's main actionable groups (3 warning-level).
Resolve the 1 F1Score._compute (cognitive 60) finding(s) in Cognitive Complexity — start with f1_score.py. — One of this dimension's main actionable groups (1 warning-level).
Resolve the 1 Precision._compute (cognitive 60) finding(s) in Cognitive Complexity — start with precision.py. — One of this dimension's main actionable groups (1 warning-level).
Enforce Cognitive Complexity in CI to reach Verified (currently Documented). — Hardens enforcement from Documented toward Prevented — provenance only; does not change the score.
Detailed fixes: d2_recommendation.md · top locations in Appendix A, every location in findings.md.
Do you agree with this assessment?
D3 · God Classes8.7 / 10Strong✓ Tool-verified
What it measures: Over-large classes that try to do too much ("god classes").
Method: God-class detection by line and method-count thresholds per logical type (partial classes unified), filtered for generated code and registration/contract false positives. Deterministic.
Resolve the 11 FileTooLong finding(s) in God Classes — start with core.py (4), mean_average_precision.py, detection.py. — One of this dimension's main actionable groups (11 warning-level).
Enforce God Classes in CI to reach Verified (currently Documented). — Hardens enforcement from Documented toward Prevented — provenance only; does not change the score.
Detailed fixes: d3_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Copy-pasted code that should be shared instead.
Method: Code duplication via token-stream sliding windows with type-aware normalization (locals masked, type names preserved), density-scored per KLoC of production code. Deterministic.
What it measures: Whether any secrets (keys, tokens, passwords) have leaked into the code.
Method: In-process native secret scanner (entropy plus signature patterns) across all tracked files; no external tool. A clean result is a measured 10, not no-data zero. Deterministic.
What it measures: Files that change often and are also complex — the riskiest hotspots.
Method: Per production file churn times cyclomatic complexity over a rolling window, computed from git and Roslyn/JS/Razor analysis. Exhaustive, deterministic per commit date.
Detailed fixes: d15_recommendation.md · top locations in Appendix A, every location in findings.md.
Do you agree with this assessment?
D16 · Bus Factor9.4 / 10Exemplary✓ Tool-verified
What it measures: Whether knowledge is concentrated in too few people (the "bus factor").
Method: Living knowledge per author via time-decayed commit attribution (6-month half-life, focus weighting) across largest source files. Deterministic, avoids blame's mechanical-refactor false positives.
4 source file(s) have their living knowledge concentrated in one author (≥90% of recent, decayed contribution). The largest is src/supervision/detection/compact_mask.py.
Off-boarding risk: anonymized user #1
✓ On the Gold path — maintain.
Detailed fixes: d16_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Whether names — types, methods, variables — are clear and consistent.
Method: Judged by language model at low temperature (0.0-0.1) on a deterministic random symbol sample (fixed size, not exhaustive), with disclosed confidence band. Advisory, sampled.
What it measures: Whether any secrets were ever committed — scanned across the full git history, not just now.
Method: Git-history secret scan via gitleaks detect over full history in an isolated checkout; each match flagged High. Exhaustive; degrades cleanly when tool absent.
2 finding(s): 0 critical, 2 high, 0 medium, 0 low. Remediation for historically-committed secrets is credential rotation — they remain in history regardless of later deletion.
Secret: generic-api-key.github/workflows/publish-docs.yml:157detected by gitleaks finding
Rotate the exposed credentials — git history can't be un-committed
What to do
Resolve the 1 Secret finding(s) in Secrets (history) — start with publish-docs.yml. — One of this dimension's main actionable groups (1 issue-level).
Resolve the 1 Rotate the exposed credentials finding(s) in Secrets (history). — One of this dimension's main actionable groups (1 recommendation-level).
Detailed fixes: d28_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Real static-analysis (SAST) findings — likely security bugs in the code, any language.
Method: Polyglot static analysis via semgrep across the repo using the pinned, image-baked p/security-audit + p/owasp-top-ten rulesets (no scan-time registry fetch); severity rules (ERROR/WARNING/INFO) map to a full-band severity-weighted score. Exhaustive, deterministic; degrades on parse failure.
Coverage: semgrep pattern rules over all files — exhaustive for the rule set, blind to classes of bug without a rule (clean = no rule matched).
High: run-shell-injection · ×10.github/workflows/build-package.yml:32detected by semgrep finding
Medium: use-defused-xmlsrc/supervision/dataset/formats/pascal_voc.py:4detected by semgrep finding
What to do
Resolve the 10 High finding(s) in Static Analysis (SAST) — start with build-package.yml (3), ci-check-links.yml (2), publish-release.yml (2). — One of this dimension's main actionable groups (10 issue-level).
Resolve the 1 Medium finding(s) in Static Analysis (SAST) — start with pascal_voc.py. — One of this dimension's main actionable groups (1 warning-level).
Detailed fixes: d29_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Whether anyone still has living knowledge of each file, or it has been orphaned — last understood long ago by someone now gone quiet. The sibling of the bus factor: D16 asks who owns it, D34 asks whether anyone still knows it.
Method: File orphaning as total living-knowledge decay below one focused-commit's worth within a year, computed per-file from the D16 decay model. Exhaustive, deterministic over fixed history.
What it measures: Whether files that change together actually belong together — pairs that repeatedly co-change in git history despite having no explicit code dependency, surfacing the hidden/logical coupling (and boundaries in the wrong place) a static scan can't see.
Method: Pairwise co-occurrence over the per-commit file sets in git history (production source only — tests and generated dropped): Degree-of-Coupling = shared ÷ min individual revisions, reported above noise floors (each file ≥10 revisions, ≥5 shared commits, ≥50% strength); sweeping commits excluded. Deterministic over fixed history.
Coverage: Population: PRODUCTION source files only — test and generated files are dropped before pairing, so a class co-changing with its own test (trivially ~100%) can't drown the real production↔production coupling. Pairs ranked by Degree-of-Coupling; coupling through a build step, config, or non-source file isn't seen.
What it measures: Whether the build pipeline provides supply-chain integrity — generated provenance/attestation, signed artifacts (cosign/sigstore), an SBOM, and pinned build actions. Presence of the configuration, not a runtime guarantee.
Method: Supply-chain provenance/signing read deterministically from CI/build config (.github/workflows, .gitlab-ci.yml, azure-pipelines, Jenkinsfile, .circleci) + the release surface: four signals — generated provenance/attestation (SLSA/in-toto/actions-attest), artifact signing (cosign/sigstore/gitsign), an SBOM (syft/sbom-action/*.spdx.json/*.cdx.json), and SHA-pinned build actions — scored 10·present/denom. NotApplicable without a build pipeline. Detects configuration presence, not runtime enforcement.
Resolve the 1 Unpinned build actions finding(s) in Supply-chain Provenance & Signing. — One of this dimension's main actionable groups (1 warning-level).
Resolve the 1 No SBOM finding(s) in Supply-chain Provenance & Signing. — One of this dimension's main actionable groups (1 recommendation-level).
Detailed fixes: d36_recommendation.md · top locations in Appendix A, every location in findings.md.
What it measures: Whether dependencies have known published vulnerabilities (CVEs) per the OSV database — read natively from whatever lockfile the repository ships (Cargo, npm, Go, Python, Maven, RubyGems, …). D33 and D30 add ecosystem-specific scanners on top for npm and .NET.
Method: Multi-ecosystem dependency-CVE scan via osv-scanner --recursive (queries the osv.dev database + parses lockfiles natively across ecosystems: npm package-lock/yarn/pnpm/bun, Go go.mod, Rust Cargo.lock, Maven/Gradle pom.xml/gradle.lockfile, PyPI requirements.txt/poetry.lock/Pipfile.lock, Composer composer.lock, RubyGems Gemfile.lock, Hex mix.lock, pub pubspec.lock, Swift Package.resolved); severity tally (Critical/High/Medium/Low) to 0-10 tight normalizer (8.0). NotApplicable only when the repo declares no supported non-.NET dependency lockfile (a NuGet-only repo stays NotApplicable — .NET CVEs are D30's domain); coverage needs a resolved lockfile. Additive to D33 (trivy fs); exhaustive + deterministic, DB kept fresh.
High CVE: [GHSA redacted] · ×2uv.lockdetected by osv-scanner finding
High vulnerability: [GHSA redacted]uv.lockdetected by osv-scanner finding
Medium CVE: PYSEC-2026-2132 · ×4uv.lockdetected by osv-scanner finding
What to do
Resolve the 2 High CVE finding(s) in OSV Dependency Vulnerabilities — start with uv.lock (2). — One of this dimension's main actionable groups (2 issue-level).
Resolve the 1 High vulnerability finding(s) in OSV Dependency Vulnerabilities — start with uv.lock. — One of this dimension's main actionable groups (1 issue-level).
Resolve the 4 Medium CVE finding(s) in OSV Dependency Vulnerabilities — start with uv.lock (4). — One of this dimension's main actionable groups (4 warning-level).
Detailed fixes: d38_recommendation.md · top locations in Appendix A, every location in findings.md.
Maturity · Maturity — Whether the repo and its projects have a README, and whether it's substantive and current.
Method: Filesystem scan: README presence, word count, and headings for depth; git history for staleness. Exhaustive across root and project dirs, deterministic.
What to do
Add a 'Testing' section to the root README — how to run the test suite.
Add an 'Architecture' / 'How it works' section to the root README — the high-level shape.
Maturity · Maturity — Whether key decisions (ADRs) and the high-level shape (C4/diagrams) are written down.
Method: Filesystem scan: ADR folder/naming conventions or content, plus Mermaid/PlantUML/C4/architecture.md discovery. Exhaustive, deterministic.
No Architecture Decision Records found — no conventional ADR directory, no `NNNN-title.md` documents and nothing ADR-shaped by content. Design rationale recorded elsewhere (a design-notes tree, a mailing list, pull-request discussion) is not visible to this check and is not re-findable per decision, so a future maintainer cannot ask why one choice was made and get an answer.
No C4/PlantUML/Mermaid diagram or architecture.md — the high-level shape isn't documented.
What to do
Record significant decisions one document per decision — dated, stating the context, the decision and its consequences — and keep them together wherever your design docs already live (a conventional `docs/adr/` tree with `NNNN-title.md` names is the most discoverable form).
Add a C4 context/container diagram (Structurizr, PlantUML or Mermaid) or an architecture.md overview.
Maturity · Maturity — Whether the README actually describes the code that exists (LLM-judged, advisory).
Method: Judged by language model at low temperature: README accuracy versus actual projects, within a disclosed tolerance. Advisory, not a measured number.
Readiness · Readiness — Whether SAST, secret/dependency scanning and performance benchmarking are wired in (presence, not runtime).
Method: Filesystem scan: SAST configuration, dependency-update automation, secret scanning, and a benchmark harness or benchmark step — in this repository's own ecosystem. Exhaustive, deterministic.
No static application security testing detected. For this repository's stack, add bandit, `semgrep --config=p/python`, or CodeQL's python pack as a CI step.
What to do
Add a SAST step to CI running what this repository's stack ships: bandit, `semgrep --config=p/python`, or CodeQL's python pack — so a security regression fails the build instead of landing.
Add gitleaks/trufflehog in CI to block PRs that introduce committed secrets.
Readiness · Readiness — Whether releases are automated and safely reversible (probes, rolling updates, approval gates) — from manifests/pipeline files, not the live environment.
Method: Filesystem scan: deployment manifests/IaC (K8s YAML, Helm, Terraform) for rolling updates, probes, approval gates, migration hooks. Exhaustive, deterministic.
What to do
The release job declares an environment, but its protection rules are not visible from the repository — confirm required reviewers are attached, or publish as a draft release so a bad build can be stopped before users can download it.
Readiness · Readiness — Whether releases are traceable — a maintained changelog and explicit version stamping.
Method: Filesystem scan: changelog file presence and version tags in csproj or git tags. Exhaustive, deterministic.
Do you agree with this assessment?
Reference — by lens
The score is the rank-weighted fold of these lenses (worst-heaviest), each including its meta-dimensions; a lens with a Critical contributor is capped at Fair (its band reads "gated by …") and is never the strongest area however high its average.
Capped at Fair by a Critical contributor — resolve it before relying on this lens.
Not included — 72 check(s) not relevant to this codebase
These checks had nothing to measure here (no tests, no git history, the codebase is small, or the architecture style doesn't apply), so they're omitted above rather than scored low.
AC1 Text alternatives — No web markup found — accessibility is not applicable to this repository.
AC2 Forms & labels — No web markup found — accessibility is not applicable to this repository.
AC3 Page structure — No web markup found — accessibility is not applicable to this repository.
AC4 Keyboard semantics — No web markup found — accessibility is not applicable to this repository.
AC5 ARIA correctness — No web markup found — accessibility is not applicable to this repository.
AC6 Visual & motion safety — No web markup found — accessibility is not applicable to this repository.
AC7 A11y enforcement — No web markup found — accessibility is not applicable to this repository.
AX1 Captive dependencies — no DI registrations detected
AX10 Code composition — not assessed — code composition is computed by ROLE over a document set that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
AX2 Stateful singletons — no singleton implementations detected
AX3 Project dependency cycles — not assessed — project cycles and dependency direction are computed over a project-reference graph that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
AX4 Dependency direction — not assessed — project cycles and dependency direction are computed over a project-reference graph that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
AX5 Architecture & structure — not assessed — architecture style/structure is computed from a project graph (projects, types, module namespaces) that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
AX6 Interface segregation — not assessed — interface segregation is computed over a type surface that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
AX7 Slice cohesion — not applicable — not a vertical-slice architecture
AX8 Test isolation — not assessed — test isolation is computed from a project graph (which projects are test projects, and what they reference) that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
AX9 CQS / query purity — no CQRS query handlers detected — query purity is not applicable to this codebase
AXB2 Runtime readiness — Advisory — this card reports evidence and never carries a score, so there is nothing missing here.
C1 Data Protection — Not assessed: these personal data controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks personal data controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
C2 Access Controls — Not assessed: these authorization controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks authorization controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
C3 Audit Trail — Not assessed: these audit controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks audit controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
C4 Data Retention — Not assessed: these retention controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks retention controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
C5 Data-Subject Rights — Not assessed: these data-subject rights controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks data-subject rights controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
D10 Test Quality — ~38598 lines of test source are present (.py) but the test-quality collector reads C# only, so skipped/assertion-free tests couldn't be counted. Not scored — this is a gap in the analyzer, not a finding about this repository.
D11 Test Reliability — Test reliability not included
D12 Dependency Hygiene — Dependency hygiene not measured — dependency manifest found but not parsed for hygiene
D14 License Compliance — Not scored — this repository's package manifest is not parsed for licence data yet. A gap in the analyzer's language coverage, NOT a finding that the repository's licenses are compliant (a Python pyproject.toml/requirements.txt (pip/uv/Poetry)), which this pass does not parse yet — so this dimension asserts nothing about this repository's licensing in either direction.
D17 Explicit Debt — explicit-debt markers are read through a C# workspace today, so they were not read for this repository's language — this asserts nothing about how many markers the code carries. Not scored — this is a gap in the analyzer, not a finding about this repository
D18 Solution Shape — D18 scores the shape of a .NET solution; this repository has no .NET solution or project files, so the dimension does not apply.
D19 Documentation Quality — LLM evaluation failed
D20 ADR Quality — N/A — ADRs are expected on deployable products with a user-facing host, not consumed libraries; no ADR log is required here.
D22 Internal API Consistency — No exposed public API
D23 Boundary Type-Coupling — Production source is present (.py) but bounded contexts are resolved over the C#/VB project set, which exposed none, so context scope could not be assessed. Not scored — this is a gap in the analyzer, not a verdict about this repository. Declaring the codebase's bounded contexts (≥2) would let cross-boundary type coupling be assessed — see the recommendation on this dimension for where. Declare them in `.codehealth/config.yaml` at the repository root (create it if absent), mapping each context name to the module-path or namespace prefixes that belong to it — e.g. `architecture:` → `contexts:` → `Billing: ["src/billing", "Acme.Billing"]`, `Catalog: ["src/catalog", "Acme.Catalog"]`.
D24 Comment Value — No inline comments to assess — comment value is not applicable here.
D25 ADR Conformance — no ADRs to check
D26 Project Cohesion — Project cohesion is assessed over the .NET project set; this target exposed no projects, so project size and spread could not be assessed. Not scored — this is a gap in the analyzer's reach, not a verdict about this repository.
D27 Navigability — No calls could be sampled, so navigability was not assessed — tracing effort is measured over resolved call sites and this target exposed none. Not scored — this is a gap in the analyzer's reach, not a verdict about this repository.
D30 Dependency Vulnerabilities — Not scored — no dependency manifest in a supported ecosystem was read for this repository. A gap in the analyzer's language coverage, NOT a finding that the repository is free of vulnerable dependencies (a Python pyproject.toml/requirements.txt (pip/uv/Poetry) — not scanned yet) — where an OSV-supported manifest exists, dependency vulnerabilities for this repository are reported under D38 instead.
D31 IaC & Container Security — No Infrastructure-as-Code or container manifests found (Dockerfile, Terraform, Kubernetes/Helm, CloudFormation); nothing to scan.
D32 Data Compliance (PII/GDPR) — Data compliance (PII/GDPR) was not assessed in this scan — no ruleset is currently available for it. This says nothing about how this repository handles personal data, in either direction.
D33 JS/npm Dependency Vulnerabilities — No JS/npm manifest or lockfile found outside build output (package.json, package-lock.json, yarn.lock, pnpm-lock.yaml, bun.lockb); no JS dependencies to scan.
D37 Vulnerability-disclosure Policy — No vulnerability-disclosure policy file found (SECURITY.md/.markdown/.rst/.txt at root or under .github/.forgejo/.gitea/docs, .well-known/security.txt). A coordinated-disclosure policy may live off-repo, so this is not evidenced rather than failed.
D39 IL Efficiency — D39 measures the IL emitted by a .NET build; this repository has no .NET solution or project files, so the dimension does not apply.
D40 Network Egress Confinement — No Kubernetes/orchestration workloads found in the repository manifests; network egress policy is a cluster-native control that may live at the platform/firewall layer, so there is nothing to assess here.
D41 Kernel & Syscall Confinement — No Kubernetes/orchestration workloads found in the repository manifests; seccomp/AppArmor/SELinux confinement is a workload-level control, so there is nothing to assess here.
D42 Runtime Threat Enforcement — No Kubernetes/orchestration workloads found in the repository manifests; runtime threat-detection and admission-control policy are cluster-level controls, so there is nothing to assess here.
D5 Coupling — Inter-project coupling could not be assessed — no analyzable project graph was found for this repository. Not scored: a gap in the analyzer's reach, not a verdict about this repository. (Coupling here is Martin afferent/efferent/instability plus reference cycles across a project-reference graph, read today from .NET project files; other ecosystems' module graphs are not read yet.)
D6 Cohesion (LCOM4) — Cohesion (LCOM4) is measured over a C#/VB class graph, and this repository's production source is .py, which this pass does not read — so no class could be assessed. Not scored — this is a gap in the analyzer, not a finding about this repository.
D7 Architectural Integrity — no checkable ADRs and no dependency cycles — architectural integrity not assessed
D8 Code Coverage — Coverage not included — suite not readable by the collector
D9 Test Distribution — Test source is present (.py) but the test-pyramid classifier reads C# only, so its unit/integration/BDD/E2E split couldn't be counted. Not scored — this is a gap in the analyzer, not a finding about this repository.
DM1 Domain Modelling — not scored — this repository shows only 1 of the 3 signals this check looks for (20 value object(s))
ED1 Event-Driven — not scored — this repository shows none of the 3 signals this check looks for
ED5 Idempotency — no mutating command handlers or message consumers detected — idempotency check not applicable
ES1 Event Sourcing — not scored — this repository shows none of the 3 signals this check looks for
GD1 Unfinished & placeholder code — no source files
IC1 Incompleteness & stubs — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
P12 CI test-gate honesty — Reported, not scored — this card publishes what the CI gate does with the test inventory rather than grading it. The findings above are its output.
P2 Observability — Observability was not assessed: this check reads a source model that does not carry this repository's product — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of a logging idiom this check recognises is NOT evidence that this repo lacks structured logging (it may log through its own ecosystem's logger). This is a gap in the analyzer, not a finding about this repository.
P5 DR & Backup — not evidenced — repo shows no backup/RTO/RPO controls; absence of evidence is not evidence of a working control
P7 Outbound HTTP resilience — not measured — the application kind could not be determined for this repo
P8 Schema migrations — not assessed — schema-migration practice is read from a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
P9 Domain vs controller coverage — no coverage report found on disk — produce a coverage report in a standard format (`coverage run -m pytest` then `coverage xml`) into the repo working tree before the scan — a CI step is the usual place, since the artefact is commonly gitignored, or wire coverage collection into CI, to enable this cross-layer check
PF1 Benchmark discipline — Performance was not assessed: this lens reads a source model that was not loaded for this repository, because the repository is written in a language this lens does not yet model or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository — in particular it is NOT a statement that this repo is unpackaged or performance-careless.
PF2 Allocation hygiene — Performance was not assessed: this lens reads a source model that was not loaded for this repository, because the repository is written in a language this lens does not yet model or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository — in particular it is NOT a statement that this repo is unpackaged or performance-careless.
PF3 Async & latency hygiene — Performance was not assessed: this lens reads a source model that was not loaded for this repository, because the repository is written in a language this lens does not yet model or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository — in particular it is NOT a statement that this repo is unpackaged or performance-careless.
S1 Web-Security Posture — Not assessed: these web-security controls are read from a source model (declarative annotations, request middleware, entity/column names, guard methods) that was not loaded for this repository — because the repository is written in a language this check does not yet model, or because its projects failed to load. Absence of an idiom this check recognises is NOT evidence that this repository lacks web-security controls: it may implement them entirely in its own ecosystem. This is a gap in the analyzer's language coverage, not a finding about this repository.
X1 Async correctness — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
X2 Cancellation propagation — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
X3 Exception handling — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
X4 Structured logging — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
X5 Nullable reference types — not analysed — these correctness checks read a source model that was not loaded for this repository, because the repository is written in a language this check does not yet model, or because its projects failed to load. This is a gap in the analyzer, not a finding about this repository
Appendix A — Findings (grouped)
The findings behind the scores, grouped by severity, then by dimension and kind. The high-severity issues are enumerated in full below; items per group are capped at 25 with any overflow stated explicitly per group, never silently truncated. The complete machine-readable list of every finding (all severities) is the companion findings.md in this report's bundle.
High: run-shell-injection .github/workflows/build-package.yml:32— Using variable interpolation `${{...}}` with `github` context data in a `run:` step could allow an attacker to inject their own code into the runner. This would allow them to steal secrets and code. `github` context data can have arbitrary user input and should be treated as untrusted. Instead, use an intermediate environment variable with `env:` to store the data and use the environment variable in the `run:` script. Reference it as a shell VARIABLE rather than a `${{ }}` interpolation, using your shell's own syntax (`"$ENVVAR"` in bash, `$env:ENVVAR` in PowerShell), so the value is passed as data and never re-expanded as code.
High: github-actions-mutable-action-tag .github/workflows/build-package.yml:38— GitHub Actions step uses a mutable tag or branch reference. Tags and branch names can be silently repointed by the action owner, enabling supply-chain attacks — as seen in the trivy-action and kics-github-action compromises. Pin the reference to a full 40-character commit SHA instead, e.g. `uses: lycheeverse/lychee-action@<40-character SHA>`. This step references `lycheeverse/lychee-action@v2`; resolve the SHA it points at today with `gh api repos/lycheeverse/lychee-action/commits/v2 --jq .sha`.
High: github-actions-mutable-action-tag .github/workflows/build-package.yml:53— GitHub Actions step uses a mutable tag or branch reference. Tags and branch names can be silently repointed by the action owner, enabling supply-chain attacks — as seen in the trivy-action and kics-github-action compromises. Pin the reference to a full 40-character commit SHA instead, e.g. `uses: actions/upload-artifact@<40-character SHA>`. This step references `actions/upload-artifact@v7`; resolve the SHA it points at today with `gh api repos/actions/upload-artifact/commits/v7 --jq .sha`.
High: github-actions-mutable-action-tag .github/workflows/ci-check-links.yml:22— GitHub Actions step uses a mutable tag or branch reference. Tags and branch names can be silently repointed by the action owner, enabling supply-chain attacks — as seen in the trivy-action and kics-github-action compromises. Pin the reference to a full 40-character commit SHA instead, e.g. `uses: actions/checkout@<40-character SHA>`. This step references `actions/checkout@v7.0.1`; resolve the SHA it points at today with `gh api repos/actions/checkout/commits/v7.0.1 --jq .sha`. Note that `v7.0.1` is an exact release tag rather than a floating major: it is still mutable (a tag can be repointed), but by convention it moves only on a force-push, so pin the floating-major and branch references in this file first.
High: github-actions-mutable-action-tag .github/workflows/ci-check-links.yml:25— GitHub Actions step uses a mutable tag or branch reference. Tags and branch names can be silently repointed by the action owner, enabling supply-chain attacks — as seen in the trivy-action and kics-github-action compromises. Pin the reference to a full 40-character commit SHA instead, e.g. `uses: lycheeverse/lychee-action@<40-character SHA>`. This step references `lycheeverse/lychee-action@v2`; resolve the SHA it points at today with `gh api repos/lycheeverse/lychee-action/commits/v2 --jq .sha`.
High: github-actions-mutable-action-tag .github/workflows/pr-conflict-labeler.yml:23— GitHub Actions step uses a mutable tag or branch reference. Tags and branch names can be silently repointed by the action owner, enabling supply-chain attacks — as seen in the trivy-action and kics-github-action compromises. Pin the reference to a full 40-character commit SHA instead, e.g. `uses: eps1lon/actions-label-merge-conflict@<40-character SHA>`. This step references `eps1lon/actions-label-merge-conflict@v3`; resolve the SHA it points at today with `gh api repos/eps1lon/actions-label-merge-conflict/commits/v3 --jq .sha`.
High: github-actions-mutable-action-tag .github/workflows/publish-pre-release.yml:40— GitHub Actions step uses a mutable tag or branch reference. Tags and branch names can be silently repointed by the action owner, enabling supply-chain attacks — as seen in the trivy-action and kics-github-action compromises. Pin the reference to a full 40-character commit SHA instead, e.g. `uses: actions/download-artifact@<40-character SHA>`. This step references `actions/download-artifact@v8`; resolve the SHA it points at today with `gh api repos/actions/download-artifact/commits/v8 --jq .sha`.
High: github-actions-mutable-action-tag .github/workflows/publish-release.yml:34— GitHub Actions step uses a mutable tag or branch reference. Tags and branch names can be silently repointed by the action owner, enabling supply-chain attacks — as seen in the trivy-action and kics-github-action compromises. Pin the reference to a full 40-character commit SHA instead, e.g. `uses: actions/download-artifact@<40-character SHA>`. This step references `actions/download-artifact@v8`; resolve the SHA it points at today with `gh api repos/actions/download-artifact/commits/v8 --jq .sha`.
High: github-actions-mutable-action-tag .github/workflows/publish-release.yml:43— GitHub Actions step uses a mutable tag or branch reference. Tags and branch names can be silently repointed by the action owner, enabling supply-chain attacks — as seen in the trivy-action and kics-github-action compromises. Pin the reference to a full 40-character commit SHA instead, e.g. `uses: AButler/upload-release-assets@<40-character SHA>`. This step references `AButler/upload-release-assets@v4.0`; resolve the SHA it points at today with `gh api repos/AButler/upload-release-assets/commits/v4.0 --jq .sha`.
High: github-actions-mutable-action-tag .github/workflows/publish-testpypi.yml:32— GitHub Actions step uses a mutable tag or branch reference. Tags and branch names can be silently repointed by the action owner, enabling supply-chain attacks — as seen in the trivy-action and kics-github-action compromises. Pin the reference to a full 40-character commit SHA instead, e.g. `uses: actions/download-artifact@<40-character SHA>`. This step references `actions/download-artifact@v8`; resolve the SHA it points at today with `gh api repos/actions/download-artifact/commits/v8 --jq .sha`.
D38 · OSV Dependency Vulnerabilities· High CVE · ×2
High CVE: [GHSA redacted] uv.lock— cairosvg 2.8.2: [GHSA redacted] — cairosvg is not declared in this repo's manifests: it is pulled in transitively, so upgrade the dependency that requires it (or raise its floor in your own manifest and re-resolve (`uv lock --upgrade-package cairosvg`)).
High CVE: [GHSA redacted] uv.lock— jaraco-context 6.0.1: [GHSA redacted] — jaraco-context is not declared in this repo's manifests: it is pulled in transitively, so upgrade the dependency that requires it (or raise its floor in your own manifest and re-resolve (`uv lock --upgrade-package jaraco-context`)).
D38 · OSV Dependency Vulnerabilities· High vulnerability · ×1
High vulnerability: [GHSA redacted] uv.lock— gitpython 3.1.54: [GHSA redacted] — gitpython is not declared in this repo's manifests: it is pulled in transitively, so upgrade the dependency that requires it (or raise its floor in your own manifest and re-resolve (`uv lock --upgrade-package gitpython`)).
FileTooLong: annotators/core.py src/supervision/annotators/core.py:0— FileTooLong — 1713 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: detection/core.py src/supervision/detection/core.py:0— FileTooLong — 1101 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: metrics/mean_average_precision.py src/supervision/metrics/mean_average_precision.py:0— FileTooLong — 1002 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: metrics/detection.py src/supervision/metrics/detection.py:0— FileTooLong — 801 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: utils/iou_and_nms.py src/supervision/detection/utils/iou_and_nms.py:0— FileTooLong — 760 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: detection/compact_mask.py src/supervision/detection/compact_mask.py:0— FileTooLong — 662 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: dataset/core.py src/supervision/dataset/core.py:0— FileTooLong — 552 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: detection/vlm.py src/supervision/detection/vlm.py:0— FileTooLong — 550 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: key_points/core.py src/supervision/key_points/core.py:0— FileTooLong — 542 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: utils/image.py src/supervision/utils/image.py:0— FileTooLong — 525 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
FileTooLong: detection/line_zone.py src/supervision/detection/line_zone.py:0— FileTooLong — 510 significant lines (blank, comment-only and punctuation-only lines excluded). To reduce it, split the file along the responsibilities already in it: move each cohesive group of declarations into its own sibling file in the same module or package, so no one file has to be read whole to change one of them.
Hotspot: src/supervision/detection/core.py src/supervision/detection/core.py— src/supervision/detection/core.py changed 29 times in last 90 days, max complexity 22. 13 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Hotspot: src/supervision/key_points/core.py src/supervision/key_points/core.py— src/supervision/key_points/core.py changed 16 times in last 90 days, max complexity 33. 4 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Hotspot: src/supervision/metrics/detection.py src/supervision/metrics/detection.py— src/supervision/metrics/detection.py changed 15 times in last 90 days, max complexity 21. 6 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Hotspot: src/supervision/detection/utils/internal.py src/supervision/detection/utils/internal.py— src/supervision/detection/utils/internal.py changed 13 times in last 90 days, max complexity 21. 7 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Hotspot: src/supervision/dataset/formats/coco.py src/supervision/dataset/formats/coco.py— src/supervision/dataset/formats/coco.py changed 14 times in last 90 days, max complexity 17. 8 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Hotspot: src/supervision/metrics/f1_score.py src/supervision/metrics/f1_score.py— src/supervision/metrics/f1_score.py changed 9 times in last 90 days, max complexity 25. 6 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Hotspot: src/supervision/detection/utils/iou_and_nms.py src/supervision/detection/utils/iou_and_nms.py— src/supervision/detection/utils/iou_and_nms.py changed 12 times in last 90 days, max complexity 18. 5 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Hotspot: src/supervision/detection/vlm.py src/supervision/detection/vlm.py— src/supervision/detection/vlm.py changed 8 times in last 90 days, max complexity 25. 5 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Hotspot: src/supervision/metrics/precision.py src/supervision/metrics/precision.py— src/supervision/metrics/precision.py changed 7 times in last 90 days, max complexity 25. 5 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Hotspot: src/supervision/detection/utils/masks.py src/supervision/detection/utils/masks.py— src/supervision/detection/utils/masks.py changed 10 times in last 90 days, max complexity 15. 3 of those changes were fix/bug commits, so the churn is repair rather than feature work. Before the next change lands here, cover the area it touches with tests, then split that area out of the file so the following change is smaller than this one — a file this often edited pays the complexity back every time.
Duplicated block (10 lines × 2) examples/traffic_analysis/inference_example.py:105— examples/traffic_analysis/inference_example.py:105-114 | examples/traffic_analysis/ultralytics_example.py:102-111 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/traffic_analysis/inference_example.py:105` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 2) examples/traffic_analysis/inference_example.py:133— examples/traffic_analysis/inference_example.py:133-142 | examples/traffic_analysis/ultralytics_example.py:130-139 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/traffic_analysis/inference_example.py:133` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 2) src/supervision/annotators/core.py:1452— src/supervision/annotators/core.py:1452-1461 | src/supervision/annotators/core.py:1802-1811 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/annotators/core.py:1452` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 2) src/supervision/dataset/formats/coco.py:528— src/supervision/dataset/formats/coco.py:528-537 | src/supervision/dataset/formats/createml.py:31-40 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/dataset/formats/coco.py:528` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 2) src/supervision/detection/utils/iou_and_nms.py:934— src/supervision/detection/utils/iou_and_nms.py:934-943 | src/supervision/detection/utils/iou_and_nms.py:1026-1035 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
Duplicated block (10 lines × 2) src/supervision/metrics/mean_average_precision.py:1508— src/supervision/metrics/mean_average_precision.py:1508-1518 | src/supervision/metrics/mean_average_precision.py:1587-1596 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (8 lines × 2) examples/compact_mask/bench_inference_api.py:332— examples/compact_mask/bench_inference_api.py:332-339 | examples/compact_mask/benchmark.py:927-938 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (8 lines × 2) examples/time_in_zone/rfdetr_stream_example.py:97— examples/time_in_zone/rfdetr_stream_example.py:97-104 | examples/time_in_zone/ultralytics_stream_example.py:34-43 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/time_in_zone/rfdetr_stream_example.py:97` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 2) examples/tracking/inference_example.py:37— examples/tracking/inference_example.py:37-44 | examples/tracking/ultralytics_example.py:26-33 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (8 lines × 2) examples/traffic_analysis/inference_example.py:92— examples/traffic_analysis/inference_example.py:92-99 | examples/traffic_analysis/ultralytics_example.py:89-96 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/traffic_analysis/inference_example.py:92` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 2) src/supervision/detection/tools/inference_slicer.py:510— src/supervision/detection/tools/inference_slicer.py:510-517 | src/supervision/detection/tools/inference_slicer.py:600-607 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/detection/tools/inference_slicer.py:510` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (8 lines × 2) src/supervision/key_points/annotators.py:103— src/supervision/key_points/annotators.py:103-110 | src/supervision/key_points/annotators.py:351-358 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
D38 · OSV Dependency Vulnerabilities· Medium CVE · ×4
Medium CVE: PYSEC-2026-2132 uv.lock— click 8.1.8: PYSEC-2026-2132 — click is not declared in this repo's manifests: it is pulled in transitively, so upgrade the dependency that requires it (or raise its floor in your own manifest and re-resolve (`uv lock --upgrade-package click`)).
Medium CVE: PYSEC-2026-2987 uv.lock— pygments 2.19.2: PYSEC-2026-2987 — pygments is not declared in this repo's manifests: it is pulled in transitively, so upgrade the dependency that requires it (or raise its floor in your own manifest and re-resolve (`uv lock --upgrade-package pygments`)).
Medium CVE: [GHSA redacted] uv.lock— pymdown-extensions 10.21.3: [GHSA redacted] — pymdown-extensions is not declared in this repo's manifests: it is pulled in transitively, so upgrade the dependency that requires it (or raise its floor in your own manifest and re-resolve (`uv lock --upgrade-package pymdown-extensions`)).
Medium CVE: [GHSA redacted] uv.lock— setuptools 80.9.0: [GHSA redacted] — setuptools is not declared in this repo's manifests: it is pulled in transitively, so upgrade the dependency that requires it (or raise its floor in your own manifest and re-resolve (`uv lock --upgrade-package setuptools`)).
Duplicated block (11 lines × 2) src/supervision/dataset/formats/coco.py:538— src/supervision/dataset/formats/coco.py:538-549 | src/supervision/dataset/formats/createml.py:41-51 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/dataset/formats/coco.py:538` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11 lines × 2) src/supervision/dataset/formats/pascal_voc.py:223— src/supervision/dataset/formats/pascal_voc.py:223-233 | src/supervision/dataset/formats/yolo.py:255-265 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (11 lines × 2) src/supervision/key_points/annotators.py:468— src/supervision/key_points/annotators.py:468-478 | src/supervision/key_points/annotators.py:568-578 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/key_points/annotators.py:468` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (11 lines × 2) src/supervision/metrics/mean_average_precision.py:1533— src/supervision/metrics/mean_average_precision.py:1533-1543 | src/supervision/metrics/mean_average_precision.py:1614-1624 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
internal.process_roboflow_result (cyclomatic 21) src/supervision/detection/utils/internal.py:259— internal.process_roboflow_result has cyclomatic complexity 21 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
internal.process_roboflow_result (cyclomatic 21) src/supervision/detection/utils/internal.py:274— internal.process_roboflow_result has cyclomatic complexity 21 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
internal.process_roboflow_result (cyclomatic 21) src/supervision/detection/utils/internal.py:288— internal.process_roboflow_result has cyclomatic complexity 21 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
internal.process_roboflow_result (cognitive 34) src/supervision/detection/utils/internal.py:259— internal.process_roboflow_result has cognitive complexity 34 (threshold 15). Drivers by points: if/else 27, ternaries 4, boolean chains 2, loops 1 (nesting depth added 14). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
internal.process_roboflow_result (cognitive 34) src/supervision/detection/utils/internal.py:274— internal.process_roboflow_result has cognitive complexity 34 (threshold 15). Drivers by points: if/else 27, ternaries 4, boolean chains 2, loops 1 (nesting depth added 14). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
internal.process_roboflow_result (cognitive 34) src/supervision/detection/utils/internal.py:288— internal.process_roboflow_result has cognitive complexity 34 (threshold 15). Drivers by points: if/else 27, ternaries 4, boolean chains 2, loops 1 (nesting depth added 14). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
Duplicated block (25 lines × 2) src/supervision/detection/utils/boxes.py:249— src/supervision/detection/utils/boxes.py:249-273 | src/supervision/detection/utils/boxes.py:282-312 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (25 lines × 2) src/supervision/draw/color.py:161— src/supervision/draw/color.py:161-185 | src/supervision/draw/color.py:189-213 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (25 lines × 2) src/supervision/draw/color.py:217— src/supervision/draw/color.py:217-241 | src/supervision/draw/color.py:247-271 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/draw/color.py:217` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (13 lines × 3) examples/speed_estimation/inference_example.py:73— examples/speed_estimation/inference_example.py:73-85 | examples/speed_estimation/ultralytics_example.py:59-71 | examples/speed_estimation/yolo_nas_example.py:60-72 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/speed_estimation/inference_example.py:73` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (13 lines × 3) examples/speed_estimation/inference_example.py:128— examples/speed_estimation/inference_example.py:128-140 | examples/speed_estimation/ultralytics_example.py:112-124 | examples/speed_estimation/yolo_nas_example.py:115-127 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (13 lines × 3) examples/time_in_zone/inference_file_example.py:39— examples/time_in_zone/inference_file_example.py:39-51 | examples/time_in_zone/rfdetr_file_example.py:118-130 | examples/time_in_zone/ultralytics_file_example.py:39-51 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (13 lines × 2) examples/count_people_in_zone/inference_example.py:184— examples/count_people_in_zone/inference_example.py:184-196 | examples/count_people_in_zone/ultralytics_example.py:172-184 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (13 lines × 2) src/supervision/dataset/core.py:188— src/supervision/dataset/core.py:188-200 | src/supervision/dataset/core.py:1119-1131 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (13 lines × 2) src/supervision/metrics/mean_average_precision.py:256— src/supervision/metrics/mean_average_precision.py:256-268 | src/supervision/metrics/mean_average_recall.py:232-244 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (12 lines × 2) examples/count_people_in_zone/inference_example.py:48— examples/count_people_in_zone/inference_example.py:48-59 | examples/count_people_in_zone/ultralytics_example.py:46-57 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/count_people_in_zone/inference_example.py:48` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (12 lines × 2) examples/time_in_zone/inference_stream_example.py:52— examples/time_in_zone/inference_stream_example.py:52-63 | examples/time_in_zone/ultralytics_stream_example.py:52-63 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/time_in_zone/inference_stream_example.py:52` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (12 lines × 2) examples/time_in_zone/ultralytics_file_example.py:53— examples/time_in_zone/ultralytics_file_example.py:53-64 | examples/time_in_zone/ultralytics_naive_stream_example.py:56-67 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (10 lines × 3) examples/time_in_zone/inference_stream_example.py:65— examples/time_in_zone/inference_stream_example.py:65-74 | examples/time_in_zone/rfdetr_stream_example.py:125-134 | examples/time_in_zone/ultralytics_stream_example.py:65-74 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/time_in_zone/inference_stream_example.py:65` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (10 lines × 3) src/supervision/detection/utils/iou_and_nms.py:520— src/supervision/detection/utils/iou_and_nms.py:520-529 | src/supervision/detection/utils/iou_and_nms.py:1632-1641 | src/supervision/detection/utils/iou_and_nms.py:1757-1766 — all 3 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited.
Duplicated block (10 lines × 3) src/supervision/metrics/mean_average_recall.py:487— src/supervision/metrics/mean_average_recall.py:487-496 | src/supervision/metrics/precision.py:324-333 | src/supervision/metrics/recall.py:291-300 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/metrics/mean_average_recall.py:487` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9 lines × 3) examples/time_in_zone/inference_stream_example.py:21— examples/time_in_zone/inference_stream_example.py:21-29 | examples/time_in_zone/rfdetr_stream_example.py:85-93 | examples/time_in_zone/ultralytics_stream_example.py:22-30 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/time_in_zone/inference_stream_example.py:21` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9 lines × 3) src/supervision/metrics/f1_score.py:327— src/supervision/metrics/f1_score.py:327-335 | src/supervision/metrics/precision.py:323-331 | src/supervision/metrics/recall.py:290-298 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/metrics/f1_score.py:327` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (9 lines × 3) src/supervision/validators/__init__.py:86— src/supervision/validators/__init__.py:86-94 | src/supervision/validators/__init__.py:109-117 | src/supervision/validators/__init__.py:171-179 — all 3 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited.
Duplicated block (9 lines × 2) examples/time_in_zone/rfdetr_file_example.py:55— examples/time_in_zone/rfdetr_file_example.py:55-63 | examples/time_in_zone/rfdetr_naive_stream_example.py:55-63 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/time_in_zone/rfdetr_file_example.py:55` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (9 lines × 2) examples/traffic_analysis/inference_example.py:39— examples/traffic_analysis/inference_example.py:39-47 | examples/traffic_analysis/ultralytics_example.py:37-45 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (9 lines × 2) examples/traffic_analysis/inference_example.py:164— examples/traffic_analysis/inference_example.py:164-172 | examples/traffic_analysis/ultralytics_example.py:161-169 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (6 lines × 2) src/supervision/annotators/utils.py:324— src/supervision/annotators/utils.py:324-329 | src/supervision/annotators/utils.py:332-337 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
Duplicated block (6 lines × 2) src/supervision/detection/core.py:2488— src/supervision/detection/core.py:2488-2493 | src/supervision/key_points/core.py:1275-1280 — the copies span different directories, so extracting a shared function means choosing where it lives: put it somewhere both call sites can already reach — a location they all depend on today, or a new shared one if there is none — and call it from each site; until then, every change has to be made twice.
Duplicated block (6 lines × 2) src/supervision/detection/utils/masks.py:341— src/supervision/detection/utils/masks.py:341-346 | src/supervision/detection/utils/masks.py:356-361 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/detection/utils/masks.py:341` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (5 lines × 2) src/supervision/_cv2/_image.py:235— src/supervision/_cv2/_image.py:235-239 | src/supervision/_cv2/_image.py:260-264 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (5 lines × 2) src/supervision/detection/tools/inference_slicer.py:527— src/supervision/detection/tools/inference_slicer.py:527-532 | src/supervision/detection/tools/inference_slicer.py:614-618 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
Duplicated block (5 lines × 2) src/supervision/detection/utils/iou_and_nms.py:644— src/supervision/detection/utils/iou_and_nms.py:644-648 | src/supervision/detection/utils/iou_and_nms.py:650-655 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (15 lines × 2) examples/compact_mask/benchmark.py:813— examples/compact_mask/benchmark.py:813-827 | src/supervision/utils/image.py:871-885 — the copies span different directories, so extracting a shared function means choosing where it lives: put it somewhere both call sites can already reach — a location they all depend on today, or a new shared one if there is none — and call it from each site; until then, every change has to be made twice.
Duplicated block (15 lines × 2) src/supervision/tracker/byte_tracker/core.py:215— src/supervision/tracker/byte_tracker/core.py:215-229 | src/supervision/tracker/byte_tracker/core.py:264-278 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
Duplicated block (7 lines × 2) src/supervision/detection/utils/masks.py:87— src/supervision/detection/utils/masks.py:87-93 | src/supervision/detection/utils/masks.py:98-104 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited.
Duplicated block (7 lines × 2) src/supervision/detection/vlm.py:257— src/supervision/detection/vlm.py:257-263 | src/supervision/detection/vlm.py:724-730 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
KeyPoints.select (cyclomatic 33) src/supervision/key_points/core.py:948— KeyPoints.select has cyclomatic complexity 33 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
_video._mux_audio (cyclomatic 31) src/supervision/_cv2/_video.py:282— _video._mux_audio has cyclomatic complexity 31 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
vlm.from_google_gemini_2_5 (cyclomatic 25) src/supervision/detection/vlm.py:733— vlm.from_google_gemini_2_5 has cyclomatic complexity 25 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
F1Score._compute (cyclomatic 25) src/supervision/metrics/f1_score.py:147— F1Score._compute has cyclomatic complexity 25 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
Precision._compute (cyclomatic 25) src/supervision/metrics/precision.py:149— Precision._compute has cyclomatic complexity 25 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
Detections.from_vlm (cyclomatic 22) src/supervision/detection/core.py:1611— Detections.from_vlm has cyclomatic complexity 22 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
InferenceSlicer.__call__ (cyclomatic 21) src/supervision/detection/tools/inference_slicer.py:309— InferenceSlicer.__call__ has cyclomatic complexity 21 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
ConfusionMatrix.evaluate_detection_batch (cyclomatic 21) src/supervision/metrics/detection.py:861— ConfusionMatrix.evaluate_detection_batch has cyclomatic complexity 21 (threshold 15). To reduce it, split the body: these branches sit side by side rather than nested inside one another, so extracting each one on its own would leave a function per branch. Group the statements between the checks into named steps and move each step into its own function, so the body reads as a short sequence of named stages.
Recall._compute (cyclomatic 20) src/supervision/metrics/recall.py:150— Recall._compute has cyclomatic complexity 20 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
ByteTrack.update_with_tensors (cyclomatic 20) src/supervision/tracker/byte_tracker/core.py:185— ByteTrack.update_with_tensors has cyclomatic complexity 20 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
CompactMask.from_coco_rle (cyclomatic 19) src/supervision/detection/compact_mask.py:736— CompactMask.from_coco_rle has cyclomatic complexity 19 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
compact_mask._rle_split_cols (cyclomatic 19) src/supervision/detection/compact_mask.py:59— compact_mask._rle_split_cols has cyclomatic complexity 19 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
iou_and_nms.oriented_box_iou_batch (cyclomatic 18) src/supervision/detection/utils/iou_and_nms.py:449— iou_and_nms.oriented_box_iou_batch has cyclomatic complexity 18 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
COCOEvaluator._evaluate_image (cyclomatic 18) src/supervision/metrics/mean_average_precision.py:822— COCOEvaluator._evaluate_image has cyclomatic complexity 18 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
COCOEvaluator._accumulate (cyclomatic 18) src/supervision/metrics/mean_average_precision.py:947— COCOEvaluator._accumulate has cyclomatic complexity 18 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
coco.detections_to_coco_annotations (cyclomatic 17) src/supervision/dataset/formats/coco.py:238— coco.detections_to_coco_annotations has cyclomatic complexity 17 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
Detections.from_sam3 (cyclomatic 17) src/supervision/detection/core.py:856— Detections.from_sam3 has cyclomatic complexity 17 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
InferenceSlicer._run_callback_batch (cyclomatic 17) src/supervision/detection/tools/inference_slicer.py:554— InferenceSlicer._run_callback_batch has cyclomatic complexity 17 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
MeanAveragePrecision._prepare_targets (cyclomatic 16) src/supervision/metrics/mean_average_precision.py:1500— MeanAveragePrecision._prepare_targets has cyclomatic complexity 16 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
video.process_video (cyclomatic 16) src/supervision/utils/video.py:377— video.process_video has cyclomatic complexity 16 (threshold 15). To reduce it, separate the cases: extract each independent branch into its own named function, and where the body has guards that only reject input, fold those into early returns at the top.
LLM evaluation failed — JSON parse error: Expected end of string, but instead reached end of data. Path: $.findings[0].issue | LineNumber: 0 | BytePositionInLine: 958.
F1Score._compute (cognitive 60) src/supervision/metrics/f1_score.py:147— F1Score._compute has cognitive complexity 60 (threshold 15). Drivers by points: if/else 51, boolean chains 4, ternaries 4, loops 1 (nesting depth added 37). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
Precision._compute (cognitive 60) src/supervision/metrics/precision.py:149— Precision._compute has cognitive complexity 60 (threshold 15). Drivers by points: if/else 51, boolean chains 4, ternaries 4, loops 1 (nesting depth added 37). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
vlm.from_google_gemini_2_5 (cognitive 54) src/supervision/detection/vlm.py:733— vlm.from_google_gemini_2_5 has cognitive complexity 54 (threshold 15). Drivers by points: if/else 41, error handling 6, boolean chains 3, loops 2, ternaries 2 (nesting depth added 25). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
COCOEvaluator._accumulate (cognitive 54) src/supervision/metrics/mean_average_precision.py:947— COCOEvaluator._accumulate has cognitive complexity 54 (threshold 15). Drivers by points: if/else 28, loops 20, ternaries 5, boolean chains 1 (nesting depth added 37). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
Recall._compute (cognitive 52) src/supervision/metrics/recall.py:150— Recall._compute has cognitive complexity 52 (threshold 15). Drivers by points: if/else 45, ternaries 4, boolean chains 2, loops 1 (nesting depth added 33). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
_video._mux_audio (cognitive 48) src/supervision/_cv2/_video.py:282— _video._mux_audio has cognitive complexity 48 (threshold 15). Drivers by points: if/else 24, ternaries 14, boolean chains 7, loops 2, error handling 1 (nesting depth added 18). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
compact_mask._rle_split_cols (cognitive 42) src/supervision/detection/compact_mask.py:59— compact_mask._rle_split_cols has cognitive complexity 42 (threshold 15). Drivers by points: if/else 32, loops 7, ternaries 2, boolean chains 1 (nesting depth added 24). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
KeyPoints.select (cognitive 41) src/supervision/key_points/core.py:948— KeyPoints.select has cognitive complexity 41 (threshold 15). Drivers by points: if/else 30, boolean chains 10, ternaries 1 (nesting depth added 11). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
InferenceSlicer.__call__ (cognitive 40) src/supervision/detection/tools/inference_slicer.py:309— InferenceSlicer.__call__ has cognitive complexity 40 (threshold 15). Drivers by points: if/else 23, loops 12, boolean chains 5 (nesting depth added 18). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
COCOEvaluator._evaluate_image (cognitive 40) src/supervision/metrics/mean_average_precision.py:822— COCOEvaluator._evaluate_image has cognitive complexity 40 (threshold 15). Drivers by points: if/else 24, loops 10, boolean chains 5, ternaries 1 (nesting depth added 22). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
coco.detections_to_coco_annotations (cognitive 36) src/supervision/dataset/formats/coco.py:238— coco.detections_to_coco_annotations has cognitive complexity 36 (threshold 15). Drivers by points: if/else 31, boolean chains 4, loops 1 (nesting depth added 16). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
Detections.from_sam3 (cognitive 35) src/supervision/detection/core.py:856— Detections.from_sam3 has cognitive complexity 35 (threshold 15). Drivers by points: if/else 25, loops 6, boolean chains 4 (nesting depth added 16). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
MeanAveragePrecision._prepare_targets (cognitive 34) src/supervision/metrics/mean_average_precision.py:1500— MeanAveragePrecision._prepare_targets has cognitive complexity 34 (threshold 15). Drivers by points: if/else 28, boolean chains 3, loops 3 (nesting depth added 18). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
Detections.from_vlm (cognitive 33) src/supervision/detection/core.py:1611— Detections.from_vlm has cognitive complexity 33 (threshold 15). Drivers by points: if/else 33 (nesting depth added 12). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
CompactMask.from_coco_rle (cognitive 31) src/supervision/detection/compact_mask.py:736— CompactMask.from_coco_rle has cognitive complexity 31 (threshold 15). Drivers by points: if/else 24, boolean chains 4, error handling 2, loops 1 (nesting depth added 11). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
EdgeAnnotator.annotate (cognitive 31) src/supervision/key_points/annotators.py:153— EdgeAnnotator.annotate has cognitive complexity 31 (threshold 15). Drivers by points: if/else 23, loops 3, ternaries 3, boolean chains 2 (nesting depth added 17). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
_geometry._simplify_slices (cognitive 30) src/supervision/_cv2/_geometry.py:32— _geometry._simplify_slices has cognitive complexity 30 (threshold 15). Drivers by points: if/else 20, loops 9, boolean chains 1 (nesting depth added 14). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
core._merge_detection_group (cognitive 30) src/supervision/detection/core.py:3278— core._merge_detection_group has cognitive complexity 30 (threshold 15). Drivers by points: if/else 20, loops 5, ternaries 4, boolean chains 1 (nesting depth added 11). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
ConfusionMatrix.plot (cognitive 30) src/supervision/metrics/detection.py:1151— ConfusionMatrix.plot has cognitive complexity 30 (threshold 15). Drivers by points: if/else 12, ternaries 12, loops 5, boolean chains 1 (nesting depth added 14). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
ByteTrack.update_with_tensors (cognitive 30) src/supervision/tracker/byte_tracker/core.py:185— ByteTrack.update_with_tensors has cognitive complexity 30 (threshold 15). Drivers by points: if/else 22, loops 8 (nesting depth added 6). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
InferenceSlicer._run_callback_batch (cognitive 29) src/supervision/detection/tools/inference_slicer.py:554— InferenceSlicer._run_callback_batch has cognitive complexity 29 (threshold 15). Drivers by points: if/else 19, loops 6, boolean chains 4 (nesting depth added 12). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
MeanAveragePrecision._prepare_predictions (cognitive 28) src/supervision/metrics/mean_average_precision.py:1582— MeanAveragePrecision._prepare_predictions has cognitive complexity 28 (threshold 15). Drivers by points: if/else 24, loops 3, boolean chains 1 (nesting depth added 16). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
ConfusionMatrix.evaluate_detection_batch (cognitive 27) src/supervision/metrics/detection.py:861— ConfusionMatrix.evaluate_detection_batch has cognitive complexity 27 (threshold 15). Drivers by points: if/else 14, loops 7, boolean chains 3, ternaries 3 (nesting depth added 5). To reduce it, split the body: most of this score is breadth rather than depth — checks laid out side by side rather than stacked — so group the statements between the checks into named steps and move each step into its own function. Some of it IS depth: where a check sits inside another whose only job is to reach it, merge the two into one condition, and where an else follows a branch that already returns, drop the trailing else and let the rest of the body continue at one level.
video.process_video (cognitive 27) src/supervision/utils/video.py:377— video.process_video has cognitive complexity 27 (threshold 15). Drivers by points: if/else 15, error handling 6, loops 3, boolean chains 2, ternaries 1 (nesting depth added 11). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
labelme.labelme_shapes_to_detections (cognitive 26) src/supervision/dataset/formats/labelme.py:61— labelme.labelme_shapes_to_detections has cognitive complexity 26 (threshold 15). Drivers by points: if/else 20, ternaries 3, boolean chains 2, loops 1 (nesting depth added 11). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
MeanAverageRecall._compute (cognitive 26) src/supervision/metrics/mean_average_recall.py:386— MeanAverageRecall._compute has cognitive complexity 26 (threshold 15). Drivers by points: if/else 24, boolean chains 1, loops 1 (nesting depth added 14). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
_geometry._intersect_convex_convex (cognitive 24) src/supervision/_cv2/_geometry.py:186— _geometry._intersect_convex_convex has cognitive complexity 24 (threshold 15). Drivers by points: if/else 15, ternaries 4, loops 3, boolean chains 2 (nesting depth added 10). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
_contours._follow_border (cognitive 22) src/supervision/_cv2/_contours.py:24— _contours._follow_border has cognitive complexity 22 (threshold 15). Drivers by points: if/else 17, loops 4, boolean chains 1 (nesting depth added 10). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
masks.filter_segments_by_distance (cognitive 22) src/supervision/detection/utils/masks.py:370— masks.filter_segments_by_distance has cognitive complexity 22 (threshold 15). Drivers by points: if/else 19, loops 2, boolean chains 1 (nesting depth added 8). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
benchmark.run_scenario (cognitive 21) examples/compact_mask/benchmark.py:626— benchmark.run_scenario has cognitive complexity 21 (threshold 15). Drivers by points: if/else 9, ternaries 9, boolean chains 2, loops 1 (nesting depth added 5). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
iou_and_nms.oriented_box_iou_batch (cognitive 21) src/supervision/detection/utils/iou_and_nms.py:449— iou_and_nms.oriented_box_iou_batch has cognitive complexity 21 (threshold 15). Drivers by points: if/else 13, boolean chains 3, ternaries 3, loops 2 (nesting depth added 6). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
EvaluationDataset.get_annotation_ids (cognitive 21) src/supervision/metrics/mean_average_precision.py:361— EvaluationDataset.get_annotation_ids has cognitive complexity 21 (threshold 15). Drivers by points: if/else 16, ternaries 4, boolean chains 1 (nesting depth added 8). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
yolo.detections_to_yolo_annotations (cognitive 20) src/supervision/dataset/formats/yolo.py:308— yolo.detections_to_yolo_annotations has cognitive complexity 20 (threshold 15). Drivers by points: if/else 14, loops 4, boolean chains 2 (nesting depth added 8). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
_BaseVertexEllipseAnnotator._iter_ellipse_params (cognitive 20) src/supervision/key_points/annotators.py:343— _BaseVertexEllipseAnnotator._iter_ellipse_params has cognitive complexity 20 (threshold 15). Drivers by points: if/else 13, loops 6, boolean chains 1 (nesting depth added 12). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
detection._split_detections_by_outcome (cognitive 20) src/supervision/metrics/detection.py:222— detection._split_detections_by_outcome has cognitive complexity 20 (threshold 15). Drivers by points: if/else 17, loops 2, boolean chains 1 (nesting depth added 5). To reduce it, split the body: most of this score is breadth rather than depth — checks laid out side by side rather than stacked — so group the statements between the checks into named steps and move each step into its own function. Some of it IS depth: where a check sits inside another whose only job is to reach it, merge the two into one condition, and where an else follows a branch that already returns, drop the trailing else and let the rest of the body continue at one level.
TraceAnnotator.annotate (cognitive 19) src/supervision/annotators/core.py:2188— TraceAnnotator.annotate has cognitive complexity 19 (threshold 15). Drivers by points: if/else 12, error handling 4, ternaries 2, loops 1 (nesting depth added 9). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
Detections.select (cognitive 19) src/supervision/detection/core.py:2645— Detections.select has cognitive complexity 19 (threshold 15). Drivers by points: ternaries 12, if/else 7 (nesting depth added 5). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
vlm.from_florence_2 (cognitive 19) src/supervision/detection/vlm.py:509— vlm.from_florence_2 has cognitive complexity 19 (threshold 15). Drivers by points: if/else 14, loops 5 (nesting depth added 6). To reduce it, split the body: most of this score is breadth rather than depth — checks laid out side by side rather than stacked — so group the statements between the checks into named steps and move each step into its own function. Some of it IS depth: where a check sits inside another whose only job is to reach it, merge the two into one condition.
vlm._recover_gemini_json_objects (cognitive 19) src/supervision/detection/vlm.py:610— vlm._recover_gemini_json_objects has cognitive complexity 19 (threshold 15). Drivers by points: if/else 12, error handling 4, boolean chains 2, loops 1 (nesting depth added 11). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
detection.detections_to_tensor (cognitive 19) src/supervision/metrics/detection.py:63— detection.detections_to_tensor has cognitive complexity 19 (threshold 15). Drivers by points: if/else 18, boolean chains 1 (nesting depth added 8). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
file.list_files_with_extensions (cognitive 19) src/supervision/utils/file.py:114— file.list_files_with_extensions has cognitive complexity 19 (threshold 15). Drivers by points: if/else 11, loops 7, boolean chains 1 (nesting depth added 10). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
check_doctest_fences._check_content (cognitive 18) .github/scripts/check_doctest_fences.py:15— check_doctest_fences._check_content has cognitive complexity 18 (threshold 15). Drivers by points: if/else 14, boolean chains 3, loops 1 (nesting depth added 8). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
Detections.from_azure_analyze_image (cognitive 18) src/supervision/detection/core.py:981— Detections.from_azure_analyze_image has cognitive complexity 18 (threshold 15). Drivers by points: if/else 14, loops 3, boolean chains 1 (nesting depth added 8). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
LineZoneAnnotatorMulticlass.annotate (cognitive 18) src/supervision/detection/line_zone.py:809— LineZoneAnnotatorMulticlass.annotate has cognitive complexity 18 (threshold 15). Drivers by points: if/else 10, loops 8 (nesting depth added 8). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
internal.get_data_item (cognitive 18) src/supervision/detection/utils/internal.py:653— internal.get_data_item has cognitive complexity 18 (threshold 15). Drivers by points: if/else 17, loops 1 (nesting depth added 10). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
_color._cvt_color (cognitive 17) src/supervision/_cv2/_color.py:21— _color._cvt_color has cognitive complexity 17 (threshold 15). Drivers by points: if/else 14, boolean chains 3 (nesting depth added 5). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
core._paint_masks_by_area (cognitive 17) src/supervision/annotators/core.py:412— core._paint_masks_by_area has cognitive complexity 17 (threshold 15). Drivers by points: if/else 13, ternaries 2, boolean chains 1, loops 1 (nesting depth added 7). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
coco.coco_annotations_to_masks (cognitive 17) src/supervision/dataset/formats/coco.py:101— coco.coco_annotations_to_masks has cognitive complexity 17 (threshold 15). Drivers by points: if/else 12, loops 3, ternaries 2 (nesting depth added 9). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
CSVSink.parse_detection_data (cognitive 17) src/supervision/detection/tools/csv_sink.py:148— CSVSink.parse_detection_data has cognitive complexity 17 (threshold 15). Drivers by points: loops 7, ternaries 6, if/else 4 (nesting depth added 9). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
JSONSink.parse_detection_data (cognitive 17) src/supervision/detection/tools/json_sink.py:150— JSONSink.parse_detection_data has cognitive complexity 17 (threshold 15). Drivers by points: loops 7, ternaries 6, if/else 4 (nesting depth added 9). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
internal.merge_metadata (cognitive 17) src/supervision/detection/utils/internal.py:599— internal.merge_metadata has cognitive complexity 17 (threshold 15). Drivers by points: if/else 12, loops 3, boolean chains 2 (nesting depth added 8). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
masks._masks_to_roi (cognitive 17) src/supervision/detection/utils/masks.py:585— masks._masks_to_roi has cognitive complexity 17 (threshold 15). Drivers by points: if/else 10, boolean chains 4, ternaries 3 (nesting depth added 5). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
_image._resize (cognitive 16) src/supervision/_cv2/_image.py:131— _image._resize has cognitive complexity 16 (threshold 15). Drivers by points: if/else 10, boolean chains 5, ternaries 1 (nesting depth added 1). To reduce it, split the body: this score is breadth rather than depth — many checks laid out side by side rather than nested inside one another, so inverting conditions into early returns has nothing left to flatten. Group the statements between the checks into named steps and move each step into its own function, so the body reads as a short sequence of named stages.
LineZone.trigger (cognitive 16) src/supervision/detection/line_zone.py:148— LineZone.trigger has cognitive complexity 16 (threshold 15). Drivers by points: if/else 13, boolean chains 1, loops 1, ternaries 1 (nesting depth added 5). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
internal.merge_data (cognitive 16) src/supervision/detection/utils/internal.py:538— internal.merge_data has cognitive complexity 16 (threshold 15). Drivers by points: if/else 11, loops 5 (nesting depth added 5). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
iou_and_nms.compact_mask_iou_batch (cognitive 16) src/supervision/detection/utils/iou_and_nms.py:600— iou_and_nms.compact_mask_iou_batch has cognitive complexity 16 (threshold 15). Drivers by points: if/else 8, ternaries 6, boolean chains 1, loops 1 (nesting depth added 7). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
vlm.from_qwen_2_5_vl (cognitive 16) src/supervision/detection/vlm.py:311— vlm.from_qwen_2_5_vl has cognitive complexity 16 (threshold 15). Drivers by points: if/else 9, error handling 4, boolean chains 2, loops 1 (nesting depth added 4). To reduce it, split the body: most of this score is breadth rather than depth — checks laid out side by side rather than stacked — so group the statements between the checks into named steps and move each step into its own function. Some of it IS depth: where a check sits inside another whose only job is to reach it, merge the two into one condition, and where an else follows a branch that already returns, drop the trailing else and let the rest of the body continue at one level.
VertexLabelAnnotator.annotate (cognitive 16) src/supervision/key_points/annotators.py:760— VertexLabelAnnotator.annotate has cognitive complexity 16 (threshold 15). Drivers by points: if/else 10, loops 4, ternaries 2 (nesting depth added 7). To reduce it, split the body into named stages: move each independent step or branch into its own named function so the body reads as a short sequence of named calls rather than one long body.
KeyPoints._get_by_2d_bool_mask (cognitive 16) src/supervision/key_points/core.py:829— KeyPoints._get_by_2d_bool_mask has cognitive complexity 16 (threshold 15). Drivers by points: if/else 11, boolean chains 3, loops 1, ternaries 1 (nesting depth added 2). To reduce it, split the body: most of this score is breadth rather than depth — checks laid out side by side rather than stacked — so group the statements between the checks into named steps and move each step into its own function. Some of it IS depth: where a check sits inside another whose only job is to reach it, merge the two into one condition.
Medium: use-defused-xml src/supervision/dataset/formats/pascal_voc.py:4— The Python documentation recommends using `defusedxml` instead of `xml` because the native Python `xml` library is vulnerable to XML External Entity (XXE) attacks. These attacks can leak confidential data and "XML bombs" can cause denial of service. This is a semgrep security-AUDIT rule: it reports that a sensitive construct is present, not that it is exploitable here. Confirm whether this site handles untrusted input or is reachable across a trust boundary — and apply the change where it is; where the construct is required by the platform or protocol it calls into, and carries no untrusted data (a syscall/FFI shim, a build- or debug-gated tool, a fixed local surface), record the review and leave the code as it is.
Unpinned build actions — CI references GitHub Actions by a floating ref (@main / @tag) rather than a pinned commit SHA, weakening build integrity. 10 floating ref(s) across 7 workflow file(s). Each floating ref is itemized at file:line by the SAST (D29) lens.
Duplicated block (44 lines × 2) src/supervision/annotators/core.py:1244— src/supervision/annotators/core.py:1244-1287 | src/supervision/annotators/core.py:2587-2630 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/annotators/core.py:1244` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (43 lines × 4) src/supervision/annotators/core.py:228— src/supervision/annotators/core.py:228-270 | src/supervision/annotators/core.py:952-994 | src/supervision/annotators/core.py:1048-1090 | src/supervision/annotators/core.py:2703-2746 — all 4 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/annotators/core.py:228` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (43 lines × 2) src/supervision/detection/vlm.py:651— src/supervision/detection/vlm.py:651-695 | src/supervision/detection/vlm.py:744-786 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/detection/vlm.py:651` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (40 lines × 2) src/supervision/annotators/core.py:2059— src/supervision/annotators/core.py:2059-2100 | src/supervision/annotators/core.py:2471-2510 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (31 lines × 4) src/supervision/metrics/f1_score.py:429— src/supervision/metrics/f1_score.py:429-459 | src/supervision/metrics/mean_average_recall.py:579-611 | src/supervision/metrics/precision.py:427-457 | src/supervision/metrics/recall.py:381-411 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (24 lines × 2) src/supervision/metrics/f1_score.py:259— src/supervision/metrics/f1_score.py:259-282 | src/supervision/metrics/recall.py:222-245 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/metrics/f1_score.py:259` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (23 lines × 3) src/supervision/metrics/mean_average_recall.py:642— src/supervision/metrics/mean_average_recall.py:642-664 | src/supervision/metrics/precision.py:485-507 | src/supervision/metrics/recall.py:439-461 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/metrics/mean_average_recall.py:642` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (23 lines × 2) examples/count_people_in_zone/inference_example.py:102— examples/count_people_in_zone/inference_example.py:102-124 | examples/count_people_in_zone/ultralytics_example.py:101-123 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/count_people_in_zone/inference_example.py:102` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (19 lines × 5) src/supervision/metrics/f1_score.py:100— src/supervision/metrics/f1_score.py:100-118 | src/supervision/metrics/mean_average_precision.py:1426-1444 | src/supervision/metrics/mean_average_recall.py:339-357 | src/supervision/metrics/precision.py:102-120 | src/supervision/metrics/recall.py:103-121 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/metrics/f1_score.py:100` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (19 lines × 3) src/supervision/metrics/f1_score.py:727— src/supervision/metrics/f1_score.py:727-745 | src/supervision/metrics/mean_average_recall.py:166-185 | src/supervision/metrics/recall.py:681-699 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (18 lines × 3) src/supervision/metrics/f1_score.py:522— src/supervision/metrics/f1_score.py:522-539 | src/supervision/metrics/mean_average_recall.py:674-691 | src/supervision/metrics/recall.py:471-488 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (18 lines × 2) src/supervision/annotators/core.py:1510— src/supervision/annotators/core.py:1510-1528 | src/supervision/annotators/core.py:1859-1876 — both copies are in the same file, so extract the block into one function there and call it from each site — the copies drift apart the first time only one of them is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/annotators/core.py:1510` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (17 lines × 2) src/supervision/metrics/f1_score.py:364— src/supervision/metrics/f1_score.py:364-380 | src/supervision/metrics/precision.py:360-376 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (16 lines × 2) examples/time_in_zone/rfdetr_file_example.py:73— examples/time_in_zone/rfdetr_file_example.py:73-88 | examples/time_in_zone/rfdetr_naive_stream_example.py:73-88 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (15 lines × 4) src/supervision/metrics/f1_score.py:127— src/supervision/metrics/f1_score.py:127-141 | src/supervision/metrics/mean_average_recall.py:366-380 | src/supervision/metrics/precision.py:129-143 | src/supervision/metrics/recall.py:130-144 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/metrics/f1_score.py:127` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (14 lines × 3) examples/time_in_zone/inference_naive_stream_example.py:43— examples/time_in_zone/inference_naive_stream_example.py:43-56 | examples/time_in_zone/rfdetr_naive_stream_example.py:122-135 | examples/time_in_zone/ultralytics_naive_stream_example.py:43-56 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/time_in_zone/inference_naive_stream_example.py:43` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (13 lines × 6) examples/time_in_zone/inference_file_example.py:65— examples/time_in_zone/inference_file_example.py:65-77 | examples/time_in_zone/inference_naive_stream_example.py:75-87 | examples/time_in_zone/rfdetr_file_example.py:142-154 | examples/time_in_zone/rfdetr_naive_stream_example.py:152-164 | examples/time_in_zone/ultralytics_file_example.py:69-81 | examples/time_in_zone/ultralytics_naive_stream_example.py:79-91 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `examples/time_in_zone/inference_file_example.py:65` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (13 lines × 5) src/supervision/metrics/f1_score.py:806— src/supervision/metrics/f1_score.py:806-818 | src/supervision/metrics/mean_average_precision.py:277-289 | src/supervision/metrics/mean_average_recall.py:256-268 | src/supervision/metrics/precision.py:801-813 | src/supervision/metrics/recall.py:760-772 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once.
Duplicated block (12 lines × 3) src/supervision/metrics/mean_average_precision.py:1682— src/supervision/metrics/mean_average_precision.py:1682-1693 | src/supervision/metrics/mean_average_precision.py:1695-1706 | src/supervision/metrics/mean_average_precision.py:1708-1720 — all 3 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/metrics/mean_average_precision.py:1682` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (11 lines × 4) src/supervision/metrics/f1_score.py:563— src/supervision/metrics/f1_score.py:563-573 | src/supervision/metrics/mean_average_recall.py:718-728 | src/supervision/metrics/precision.py:553-563 | src/supervision/metrics/recall.py:517-527 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (11 lines × 3) src/supervision/metrics/f1_score.py:397— src/supervision/metrics/f1_score.py:397-407 | src/supervision/metrics/precision.py:393-405 | src/supervision/metrics/recall.py:347-359 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/metrics/f1_score.py:397` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (10 lines × 5) examples/time_in_zone/rfdetr_file_example.py:32— examples/time_in_zone/rfdetr_file_example.py:32-41 | examples/time_in_zone/rfdetr_naive_stream_example.py:32-41 | src/supervision/detection/utils/iou_and_nms.py:41-50 | src/supervision/detection/vlm.py:60-69 | src/supervision/detection/vlm.py:105-114 — there are 5 copies across 4 file(s) — more copies than files, so at least one file holds the block twice. Extract it once into a single shared function every call site can reach and call it from all 5 sites; resolving a subset leaves the remainder to drift apart. Read the line range as the matched WINDOW rather than a finished unit: at `examples/time_in_zone/rfdetr_file_example.py:32` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (9 lines × 4) src/supervision/detection/utils/iou_and_nms.py:1240— src/supervision/detection/utils/iou_and_nms.py:1240-1251 | src/supervision/detection/utils/iou_and_nms.py:1403-1411 | src/supervision/detection/utils/iou_and_nms.py:1493-1501 | src/supervision/detection/utils/iou_and_nms.py:1670-1678 — all 4 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/detection/utils/iou_and_nms.py:1493` it does not close everything it opens, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it. Note that the copies do not run to the end of the range shown: their LAST lines are different code, not the same code under different names — the matched region ends inside that line. Extract the lines above it, and read the last line of each site separately.
Duplicated block (7 lines × 4) src/supervision/metrics/f1_score.py:711— src/supervision/metrics/f1_score.py:711-717 | src/supervision/metrics/mean_average_recall.py:150-156 | src/supervision/metrics/precision.py:704-711 | src/supervision/metrics/recall.py:664-671 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. Read the line range as the matched WINDOW rather than a finished unit: at `src/supervision/metrics/precision.py:704` it begins part-way through the construct above it, so those exact lines cannot be lifted as they stand — widen the region to the smallest complete statement or declaration that contains it, and extract that.
Duplicated block (6 lines × 4) src/supervision/metrics/f1_score.py:551— src/supervision/metrics/f1_score.py:551-556 | src/supervision/metrics/mean_average_recall.py:703-710 | src/supervision/metrics/precision.py:538-545 | src/supervision/metrics/recall.py:500-507 — the copies sit in sibling files of one directory: extract the block into a single shared function in that directory and call it from each site, so a change lands once. The matched lines also transfer control out of the body holding them, which cannot survive a move into a called unit unchanged: have the extracted unit return that decision and let each site act on it.
Duplicated block (5 lines × 4) src/supervision/metrics/utils/object_size.py:123— src/supervision/metrics/utils/object_size.py:123-127 | src/supervision/metrics/utils/object_size.py:158-162 | src/supervision/metrics/utils/object_size.py:205-209 | src/supervision/metrics/utils/object_size.py:251-255 — all 4 copies are in the same file, so extract the block into one function there and call it from every one of those sites — resolving only two of them leaves the rest to drift apart the first time one is edited.
Recommendation — 5 finding(s)
D11 · Test Reliability· Test reliability not included · ×1
Test reliability not included — Test source is present (.py) but the built-in reliability runner does not support this repository's ecosystem, so flakiness couldn't be assessed. Not scored — this is a gap in the analyzer's language coverage, not a finding about this repository.
Off-boarding risk: anonymized user #1 — If anonymized user #1 becomes unavailable, 4 significant file(s) lose their only recent owner: src/supervision/detection/compact_mask.py, src/supervision/geometry/core.py, src/supervision/utils/notebook.py, src/supervision/utils/iterables.py. Pair on, review, or document these before any departure.
D28 · Secrets (history)· Rotate the exposed credentials · ×1
Rotate the exposed credentials — git history can't be un-committed — Some of these secrets are in git HISTORY: deleting the file does not remove them (the commit persists on every clone, fork and backup). The remediation is to ROTATE each historically-exposed credential and treat it as compromised — not to delete the file. Rewriting history is disruptive and unreliable across existing forks. (Working-tree-only secrets — no commit — can instead be removed from the file and moved to a secret store.)
D36 · Supply-chain Provenance & Signing· No SBOM · ×1
No SBOM — No SBOM generation or committed SBOM found — produce one with what your ecosystem ships (`cyclonedx-py` over the resolved Python environment/lockfile, `syft` (or `anchore/sbom-action` in CI) over the source tree or released image). Publish it as a release asset (`*.spdx.json` / `*.cdx.json`) so consumers can see what they are installing.
Coverage not included — suite not readable by the collector — Coverage NOT READ here — but this repository measures it: a Codecov configuration (.codecov.yml, target 95%) and a coverage step in CI (`pytest src/ tests/ --cov=supervision --cov`) shows that coverage is collected and tracked in your own CI. The built-in collector has no runner for this ecosystem (.py), so the analyzer could not read the number — a gap in the analyzer's language coverage, not an unmeasured repo. Not scored. To have the real number read, produce a coverage report in a standard format (`coverage run -m pytest` then `coverage xml`) into the repo working tree before the scan — a CI step is the usual place, since the artefact is commonly gitignored.
Info — 2 finding(s)
D12 · Dependency Hygiene· Dependency hygiene not measured · ×1
Dependency hygiene not measured — dependency manifest found but not parsed for hygiene — This repository's dependency manifest (a Python pyproject.toml/requirements.txt (pip/uv/Poetry)) was found, but this pass cannot parse it for hygiene, so no package was assessed. Zero packages read is NOT a clean dependency tree, so this is NOT SCORED — a gap in the analyzer, not a verdict about this repository. This row is about dependency HYGIENE — outdated, deprecated or unmaintained direct dependencies; known CVEs in the same dependency graph are a separate question, reported under D38 wherever the manifest is OSV-readable.
D22 · Internal API Consistency· No exposed public API · ×1
No exposed public API — No intentionally-exposed types (IsPackable or .Contracts) to evaluate.
Appendix B — Reproduction & audit trail
Every external tool invocation behind a deep-scan dimension — the tool, its captured version, the exact command, how many findings it yielded, and a link to the retained raw output. To reproduce any finding: check out the same commit and run the command shown (repo-relative — never an absolute scratch path). The complete raw scanner output is retained verbatim under artifacts/raw/ (indexed in artifacts/raw/index.json); per-invocation exit codes and wall-clock durations are in sidecar.json — kept out of this table so the rendered report stays byte-identical across runs of the same commit.
trivy: not applicable — No Infrastructure-as-Code or container manifests found (Dockerfile, Terraform, Kubernetes/Helm, CloudFormation); nothing to scan.
semgrep: not applicable — Data compliance (PII/GDPR) was not assessed in this scan — no ruleset is currently available for it. This says nothing about how this repository handles personal data, in either direction.
trivy: not applicable — No JS/npm manifest or lockfile found outside build output (package.json, package-lock.json, yarn.lock, pnpm-lock.yaml, bun.lockb); no JS dependencies to scan.
disclosure: not applicable — No vulnerability-disclosure policy file found (SECURITY.md/.markdown/.rst/.txt at root or under .github/.forgejo/.gitea/docs, .well-known/security.txt). A coordinated-disclosure policy may live off-repo, so this is not evidenced rather than failed.
runtime-hardening: not applicable — No Kubernetes/orchestration workloads found in the repository manifests; network egress policy is a cluster-native control that may live at the platform/firewall layer, so there is nothing to assess here.
runtime-hardening: not applicable — No Kubernetes/orchestration workloads found in the repository manifests; seccomp/AppArmor/SELinux confinement is a workload-level control, so there is nothing to assess here.
runtime-hardening: not applicable — No Kubernetes/orchestration workloads found in the repository manifests; runtime threat-detection and admission-control policy are cluster-level controls, so there is nothing to assess here.
0
—
Run 019fc3ff-3d37-7972-8a08-31887f4bfbab · every finding is also locatable in findings.md, and the complete scoring record (with exit codes + durations) in sidecar.json.
Issues: 14 · Warnings: 186 · Recommendations: 5 · Info: 2 — Appendix A · all findings · full markdown report.
Generated by Watchdog — deterministic code-health analysis. 02-08-2026 @ 19:41 UTC.
Downloadable artifacts
Machine-readable and reproducible from this commit + frozen rubric — drop them straight into a contract appendix, a CRA dossier, or a downstream SCA / VEX tool.