Turbo / Nx parity map

What a Turborepo or Nx user relies on, the vx spelling of it, and the test that proves it. Two end-to-end suites run the real CLI on one four-package fixture (tests/helpers/parity.ts: app → ui → lib, app → lib, docs alone) and name each case for the upstream contract: tests/parity-turbo.test.ts (turborepo.dev, 2.10) and tests/parity-nx.test.ts (nx.dev, 23). They are the scan surface; the right column names the suite that pins the behaviour in depth. A deliberate divergence is marked ≠ and explained.

Turbo

Turbo	vx	Deep pin
`"dependsOn": ["^build"]`	`dependsOn: ['^build']`	`tests/task-graph.test.ts`
`"dependsOn": ["build"]` (same package)	`dependsOn: ['build']`	`tests/task-graph.test.ts`
`"dependsOn": ["pkg#task"]`	`dependsOn: ['pkg#task']`	`tests/dependency-spec.test.ts`
`--only`	`--exclude-dependencies[=<names>]`	`tests/cli.test.ts`
`--dry`, `--dry=json`	`--dry`, `--dry=json`	`tests/plan-format.test.ts`, `tests/plan-predict.test.ts`
`--graph=file.dot`	`--graph[=<path>]` (DOT; edges point dependency → dependent)	`tests/cli.test.ts`
`--filter=pkg`, `pkg...`, `...pkg`, `./dir`, `!pkg`	the same DSL (`docs/cli.md` § Filter DSL)	`tests/filter.test.ts`
`--filter=[ref]`, `--affected`	`--filter '[<ref>]'`, `--affected[=<ref>]`; `...[<ref>]` adds dependents	`tests/affected.test.ts`
`--filter` + `--affected` intersect	≠ they union (`docs/comparison.md` § Filter DSL)	`tests/filter.test.ts`
a filter matching nothing exits 1	the same	`tests/task-selection.test.ts`
second run hits; missing `outputs` restored; logs replayed	the same, with outputs cleaned before exec and restore (≠ additive)	`tests/orchestrator.test.ts`, `tests/output-dirs.test.ts`
`"inputs": [...]` narrows the hash	`cache.inputs.files` (required; nothing is inferred)	`tests/inputs.test.ts`, `tests/task-hash.test.ts`
`"inputs": ["src/"]`, `"outputs": ["dist"]` (a directory literal)	the same: a literal is the file or its whole tree	`tests/inputs-resolution.test.ts`
`--filter=./apps/*` (a path glob)	`--filter './apps/'`, `{apps/*}` — over root-relative project dirs	`tests/filter.test.ts`
`--affected` diffs from the merge base	the same	`tests/affected.test.ts`
`TURBO_CACHE=remote:rw` with no remote → “Remote caching disabled”	`--cache=remote:rw` with no remote layer → one status line	`tests/parity-turbo.test.ts`
a dependency’s change re-hashes dependents	upstream input keys are folded (never outputs)	`tests/task-hash-derive.test.ts`, `tests/upstream.test.ts`
`"env": ["MODE"]`	`cache.inputs.env: ['MODE']`	`tests/env.test.ts`
`"passThroughEnv"`	`exec.env.passThrough` (reaches the command, never the hash)	`tests/env.test.ts`
`"cache": false`	no `cache` block (`noCache: true` in `--summarize`)	`tests/no-cache-word.test.ts`, `tests/run-artifacts.test.ts`
`--force`	`--force` (reads off, writes on)	`tests/cli.test.ts`
`--cache=local:r,remote:` / `--no-cache`	`--cache=<spec>` / `--no-cache`	`tests/cli.test.ts`, `tests/download-policy.test.ts`
`turbo run test -- --watch`	`vx run test -- --watch` (forwarded args are in the hash)	`tests/cli.test.ts`
the global hash covers the lockfile	the workspace fingerprint, in every key; `@vzn/vx-lockfile` narrows it	`tests/fingerprint.test.ts`, `tests/lockfile-claim.test.ts`
a package’s `package.json` is in its hash	the same	`tests/task-hash.test.ts`
`--continue=dependencies-successful` (default)	`--continue=deps-ok` (default)	`tests/continue-taint.test.ts`
`--continue=always`	`--continue` (a task built on a failure runs but is never saved)	`tests/continue-taint.test.ts`
`--continue=never`	`--continue=never`	`tests/scheduler.test.ts`
graceful shutdown (`graceful_shutdown_test.rs`)	the signal received, `VX_KILL_GRACE_MS`, SIGKILL; a second one skips the grace	`tests/signal-handling.test.ts`
recursive `turbo run` refused (`recursive_turbo_test.rs`)	`VX_RUN_WORKSPACE` marker; `vx run` in its own workspace is refused	`tests/recursive-run.test.ts`
failures are never cached	the same	`tests/execute-task.test.ts`
`--output-logs=errors-only`, `hash-only`, `none`	the same (`full` too; the default follows the run’s flow)	`tests/output-flow.test.ts`
`--summarize`	`--summarize[=<path>]`	`tests/run-artifacts.test.ts`
`--profile`	`--profile[=<path>]`	`tests/run-artifacts.test.ts`
`--concurrency`	`--concurrency <n \| n%>`	`tests/scheduler.test.ts`
`"persistent": true`	`exec.persistent: { readyWhen }` (readiness gates dependents)	`tests/persistent.test.ts`, `tests/dev.test.ts`
`turbo watch`	`vx watch <task>`	`tests/watch-rules.test.ts`, `tests/watch-loop.test.ts`, `tests/watch-signals.test.ts`
`turbo prune`	≠ removed 2026-09-11 (was `@vzn/vx-prune`): copy the workspace and `--filter` the build
`--summarize` + Vercel Remote Cache	`turboCache()` in `@vzn/vx-migrate` (same wire, self-hosted or Vercel)	`packages/vx-migrate/tests/turbo-cache.test.ts`
`$TURBO_ROOT$/tsconfig.json` input	`cache.inputs.workspaceFiles: ['tsconfig.json']` (outputs too)	`tests/workspace-files.test.ts`
`globalDependencies`, `globalEnv`, `extends`	≠ rejected: configs are TypeScript and compose by import	`docs/comparison.md` § Likely-worth-adding item 4 (global inputs and env), § Config schema comparison “Target defaults / inheritance” (`extends`)

Nx

Nx	vx	Deep pin
`nx run app:build`	`vx run app#build`	`tests/task-selection.test.ts`
`nx run app:build:production`	`vx run app#build:production` — under `nx()` or the migration a configuration is a task of its own; the default configuration is the base task	`packages/vx-migrate/tests/nx.test.ts`
`nx run-many -t build test`	`vx run build test --all`	`tests/task-selection.test.ts`
`nx run-many -t lint -p app ui`	`vx run lint --filter app --filter ui`	`tests/filter.test.ts`
`--exclude docs`	`--filter '!docs'`	`tests/filter.test.ts`
`nx affected` (changed and dependents)	the same: `--affected` is changed and dependents; `--filter '[<ref>]'` is the changed set	`tests/affected-dependents.test.ts`
`nx affected --base=main`	`--affected=main`	`tests/affected.test.ts`
`packages: ['packages/*', '!packages/fixtures']` (a negated workspace glob)	the same: negations subtract in discovery	`tests/workspace.test.ts`
`nx affected` with nothing changed exits 0	the same (`nothing affected since <ref>`)	`tests/affected.test.ts`
`--parallel=3`	`--concurrency 3`	`tests/scheduler.test.ts`
`--skipNxCache`	`--no-cache`	`tests/cli.test.ts`
`--nxBail`	`--continue=never`	`tests/scheduler.test.ts`
`nx show projects`, `nx show project app --json`	`vx show`, `vx show app --format json`	`tests/show-info.test.ts`
`nx graph`	`vx run <task> --graph` (DOT)	`tests/cli.test.ts`
`"dependsOn": ["^build"]` / `{ projects: "dependencies" }`	`dependsOn: ['^build']`	`tests/task-graph.test.ts`
`{ projects: ["ui"], target: "build" }`	`dependsOn: ['ui#build']`	`tests/dependency-spec.test.ts`
`"dependsOn": ["build-*"]` (wildcards)	`dependsOn: ['check.']`, `['^build.']`	`tests/wildcard-depends.test.ts`
a target that only aggregates	a task with no `exec` — a group; its members run, it is not a task	`tests/task-graph.test.ts`, `tests/tally.test.ts`
`{ "runtime": "node -v" }` input	`cache.inputs.runtime: ['node -v']` (`workspaceRuntime` at the root)	`tests/runtime-inputs.test.ts`
`{ "env": "NODE_ENV" }` input	`cache.inputs.env: ['NODE_ENV']`	`tests/env.test.ts`
`{workspaceRoot}/tsconfig.base.json` input	`cache.inputs.workspaceFiles: ['tsconfig.base.json']`	`tests/workspace-files.test.ts`
`"outputs"` restored on a hit	`cache.outputs.files` (cleaned before exec and restore, ≠ additive)	`tests/orchestrator.test.ts`
`project.json` / `package.json` changes re-hash	the resolved config and the project `package.json` are in every key	`tests/config-staleness.test.ts`, `tests/task-hash.test.ts`
failures are never cached	the same	`tests/execute-task.test.ts`
`"continuous": true`	`exec.persistent: { readyWhen }`	`tests/persistent.test.ts`
`nx watch`	`vx watch <task>`	`tests/watch-rules.test.ts`, `tests/watch-loop.test.ts`
Nx Cloud’s flaky-task flag	local: same key, both outcomes → the footer, `--summarize` `flaky`, `vx info`	`tests/flaky.test.ts`, `tests/failure-mode.test.ts`
lockfile-aware affected / hashing (the daemon’s pruned lockfile)	`@vzn/vx-lockfile`: per-project dependency closures, `--affected` follows	`packages/vx-lockfile/tests`, `tests/lockfile-claim.test.ts`
`nx reset`	`vx cache prune`	`tests/cache.test.ts`, `tests/cli.test.ts`
`namedInputs`, `targetDefaults`, `{projectRoot}` tokens, `.env` auto-loading, tags	≠ rejected: TypeScript configs compose by import	`docs/comparison.md` § Explicitly rejected (`namedInputs`, `targetDefaults`), § Deliberate divergences › Config schema (tokens), › Env handling (`.env`), › Task graph (tags)
Nx Cloud (agents, dashboard, DTE)	≠ not in this repo: the `executor` / `cache` / `telemetry` seams, `@vzn/vx-reapi`	`packages/vx-reapi/tests`

Reading the map

Every row’s vx spelling is documented in docs/cli.md or docs/schema.md; the parity suite asserts the row against the real CLI, the deep pin asserts the edge cases.
The two ≠ rows that change what a command selects or leaves on disk are decisions recorded in docs/comparison.md with the reasoning: union not intersection under § Deliberate divergences, cleaned-not-additive outputs under § Where vx is ahead.
What neither runner does and vx does — the resolved-config hash, vx lock + vx run --frozen, vx why, restore-ahead scheduling — is docs/comparison.md § Where vx is ahead; the parity map lists only what a user of the other runner would look for.