Truss is a CLI tool that detects and enforces architectural boundaries in JavaScript and TypeScript projects, especially when integrated into CI pipelines.
It prevents unintended dependencies between layers, such as controllers importing database modules directly, by analyzing import/export declarations, CommonJS require calls, dynamic imports, and building a dependency graph.
Truss detects both direct and transitive violations and is designed for deterministic local runs and CI-based enforcement.
As applications grow, architectural boundaries are often violated unintentionally:
- controllers start importing database modules directly
- routes bypass services
- dependencies become tangled and hard to reason about
Truss enforces these boundaries automatically when used in CI, helping prevent architectural drift before problematic code is merged.
Run without installing:
npx truss-lint init
npx truss-lint checkOr install globally:
npm install -g truss-lint
truss-lint init
truss-lint checkTruss can render a dependency graph of your project with layer grouping and highlighted architectural violations.
Install locally in your project:
npm install -D truss-lintOr run directly with npx:
npx truss-lint init
npx truss-lint checkCreate a starter configuration file in your project:
npx truss-lint initThis generates a truss.yml file in your project root.
Edit the generated truss.yml to define your architecture:
version: "1"
layers:
client:
- "client/**/*.ts"
- "client/**/*.tsx"
server:
- "server/**/*.ts"
rules:
- name: no-client-to-server
from: client
disallow: [server]npx truss-lint checknpx truss-lint graph > graph.dot
dot -Tsvg graph.dot -o graph.svgThis helps visualize architectural structure and identify problematic dependencies.
Truss performs deterministic static analysis using a dependency graph pipeline:
- Load and validate
truss.yml - Discover source files (
.ts,.tsx,.js,.jsx) - Parse import/export declarations, CommonJS require calls, and dynamic imports
- Build dependency edges
- Assign files to layers
- Evaluate architectural rules
- Apply suppressions
- Render human-readable or JSON output
- Exit with a CI-friendly status code
Type-only imports are excluded from runtime dependency analysis to avoid false-positive cycles from TypeScript import type declarations.
- Graph-based dependency analysis
- Detection of direct and transitive architectural violations via graph traversal
- Dependency cycle detection integrated into analysis diagnostics
- Layer-based architecture enforcement via configuration
- Deterministic CLI and JSON outputs
- CI integration for automated architectural checks
- Visual graph rendering with violation highlighting
- Suppression support for intentional exceptions
0No unsuppressed violations1One or more unsuppressed architectural violations or blocking graph diagnostics2Configuration or CLI usage error3Internal error
Truss: Architectural violations found (2)
Direct Violations (1)
--------------------------------
[VIOLATION] no-api-to-db
Layers: api -> db
src/api/user.ts:15
import { db } from "../db/client";
Reason: API layer must not depend directly on DB layer.
Transitive Violations (1)
--------------------------------
[TRANSITIVE VIOLATION] no-api-to-db
Layers: api -> db
src/api/user.ts:0
[transitive]
Path: src/api/user.ts -> src/service/status.ts -> src/db/client.ts
Reason: API layer must not depend on DB layer.
Summary:
Unsuppressed: 2
Suppressed: 0
Total: 2
Truss: No Architectural violations found
Checked 9000 files
Truss guarantees stable and deterministic output across runs:
- consistent sorting of violations
- stable JSON schema
- snapshot-safe CLI output
This ensures reliable CI checks and predictable developer experience.
When --format json is provided, Truss prints exactly one JSON object to stdout.
All JSON output includes:
schemaVersionkind(reportorerror)
{
"schemaVersion": "1.1.0",
"kind": "report",
"exitCode": 0,
"checkedFiles": 42,
"edges": 137,
"unsuppressed": [],
"suppressed": [],
"parserIssues": [],
"analysis": {
"diagnostics": [],
"categories": {
"parser": 0,
"graph": 0,
"validation": 0,
"suppression": 0
}
},
"summary": {
"unsuppressedCount": 0,
"suppressedCount": 0,
"parserIssueCount": 0,
"diagnosticCount": 0,
"totalCount": 0
}
}{
"schemaVersion": "1.1.0",
"kind": "error",
"exitCode": 2,
"error": "Failed to load truss.yml"
}Use check --report-dir <path> to write both reports to a directory resolved relative to the repository root (or an absolute path). The directory is created if needed. Reports are written when analysis produces a report, including when violations are found.
truss-report.json: the versioned JSON report with the exit code, file and dependency counts, unsuppressed and suppressed violations, parser issues, diagnostics, and summary counts.truss-report.html: a standalone HTML summary with a generation timestamp, PASS/FAIL status, file and dependency counts, unsuppressed violations, and diagnostics. Unsuppressed violations or error diagnostics produce FAIL.
From the Truss repository, run:
npx tsx bin/truss.ts check --report-dir reportsS3 publishing is opt-in. Add --upload-s3 together with --report-dir <path> to upload the generated JSON and HTML files after they are written locally.
TRUSS_S3_BUCKET: required destination bucket when uploading.TRUSS_S3_PREFIX: optional object-key prefix. Leading and trailing slashes are removed; without a prefix, each filename is used as its object key.
With AWS credentials and region already configured, replace the example bucket with your destination bucket:
TRUSS_S3_BUCKET=your-report-bucket TRUSS_S3_PREFIX=truss/reports \
npx tsx bin/truss.ts check --report-dir reports --upload-s3The publisher uses S3Client({}), so credentials must be supplied through the AWS SDK's standard Node.js credential provider chain. Truss does not provide a credential CLI option. Keep credentials outside the repository; do not commit secrets.
Truss integrates with CI pipelines to enforce architectural constraints automatically.
When Truss is run inside CI, unsuppressed architectural violations or blocking diagnostics cause the command to exit with a non-zero status code. This allows pull requests to fail before architecture-breaking changes are merged.
Build the image:
npm run docker:buildDisplay CLI help:
npm run docker:helpRun an architecture check against the current project:
npm run docker:checkOr run the image directly:
docker run --rm -v "$PWD:/workspace" -w /workspace truss-lint checkname: Truss
on:
pull_request:
push:
branches: [main]
jobs:
truss:
runs-on: ubuntu-latest
steps:
- name: Checkout repository
uses: actions/checkout@v4
- name: Setup Node.js
uses: actions/setup-node@v4
with:
node-version: 20
cache: npm
- name: Install dependencies
run: npm ci
- name: Run Truss architecture check
run: npx truss-lint checkRun the full test suite locally:
npm testRun the suite with source-only coverage thresholds:
npm run test:coverageCI enforces source-only minimum coverage of 80% for lines, branches, and functions. tests/**, dist/**, and node_modules/** are excluded from the coverage calculation.
The integration suite uses fixture repos and committed snapshots to keep the CLI contract explicit.
- Snapshot tests for human-readable and JSON output
- Validation of exit codes (
0–3) - Coverage of clean, violation, suppressed, and error scenarios
- Coverage of parser diagnostics and graph diagnostics
- Coverage of deterministic cycle detection
Truss can emit OpenTelemetry metrics for analysis runs. Telemetry is disabled by default so normal CLI output stays clean.
Enable console metrics output:
TRUSS_OTEL_CONSOLE=1 truss-lint checkOr during local development:
TRUSS_OTEL_CONSOLE=1 npx tsx bin/truss.ts checkTruss currently records:
| Metric | Description |
|---|---|
truss.analysis.duration_ms |
Duration of a Truss analysis run in milliseconds |
truss.files.scanned |
Number of source files scanned during analysis |
truss.dependency_edges.count |
Number of dependency edges detected during analysis |
truss.violations.count |
Number of unsuppressed architecture violations detected |
truss.diagnostics.count |
Number of diagnostics produced during analysis |
Telemetry is opt-in and intended for CI observability, debugging, and tracking analysis performance as repositories grow.