# code quality is a profile

def profile(src):

feats = extract(src)

for dim in DIMENSIONS:

level = ladder(feats, dim)

yield dim, level

# infer the intended shape

shape = nearest(centroid)

gap = target - actual

return Report(levels, gap)

# the model finds, the code decides

evidence = cite_lines(src)

assert deterministic(score)

class Dimension:

name: str

rungs: list[Rung]

qualified: bool

# code quality is a profile

def profile(src):

feats = extract(src)

for dim in DIMENSIONS:

level = ladder(feats, dim)

yield dim, level

# infer the intended shape

shape = nearest(centroid)

gap = target - actual

return Report(levels, gap)

# the model finds, the code decides

evidence = cite_lines(src)

assert deterministic(score)

class Dimension:

name: str

rungs: list[Rung]

qualified: bool

# code quality is a profile

def profile(src):

feats = extract(src)

for dim in DIMENSIONS:

level = ladder(feats, dim)

yield dim, level

# infer the intended shape

shape = nearest(centroid)

gap = target - actual

return Report(levels, gap)

# the model finds, the code decides

evidence = cite_lines(src)

assert deterministic(score)

class Dimension:

name: str

rungs: list[Rung]

qualified: bool

// validated, not asserted

const report = await facet(file)

for (const d of report.dims) {

if (d.level < d.target) {

climbs.push(d)

}

}

// one judge per dimension

const judge = route(dim)

const levels = score(features)

// your source is never stored

const vec = scrub(features)

const hash = digest(src)

function coherence(sections) {

return spread(sections)

}

// a throwaway script can pass

// validated, not asserted

const report = await facet(file)

for (const d of report.dims) {

if (d.level < d.target) {

climbs.push(d)

}

}

// one judge per dimension

const judge = route(dim)

const levels = score(features)

// your source is never stored

const vec = scrub(features)

const hash = digest(src)

function coherence(sections) {

return spread(sections)

}

// a throwaway script can pass

// validated, not asserted

const report = await facet(file)

for (const d of report.dims) {

if (d.level < d.target) {

climbs.push(d)

}

}

// one judge per dimension

const judge = route(dim)

const levels = score(features)

// your source is never stored

const vec = scrub(features)

const hash = digest(src)

function coherence(sections) {

return spread(sections)

}

// a throwaway script can pass

Code quality is a profile, not a score.

Most tools hand your code a grade. That assumes there is one right answer, and there isn’t: a hot path, a throwaway script, and a security boundary all want different things. Facet measures your code across 14 dimensions that trade off against each other, works out which profile it was built for, and shows you where it drifts from that. A model reads the code and cites the evidence. Deterministic scoring, not the model, sets the levels.

Get started See a sample report Why no single “best”?

Free while in beta, no card required. See a sample report first, or where pricing is heading.

What makes it different

One judge per dimension. No model is best at everything, so each dimension is scored by the model that proved most reliable at measuring it.
The model finds, the code decides. The model reads your code and cites the lines it relies on. Scoring is deterministic, so the same code always gets the same result.
Honest about what we know. Each dimension is marked validated or provisional by whether it clears our reliability bar. All 14 are validated today; if one ever slips below the bar, we say so rather than hide it.
We never store your source. We keep secret-scrubbed measurements about your code and a content hash, scoped to your account. The code itself is never kept.