veridist

English | فارسی | Deutsch

Fit lifetime data with an inspectable result

Veridist is for reliability engineers and analysts who need a defined lifetime fit from a strict CSV, plus the information needed to review how it ran. It validates the input, returns a typed fit or failure, and keeps the execution facts visible.

See it work · Choose a workflow · Read known limits

Install and first success

Install the released package:

python -m pip install veridist

See it work

Run this complete CSV fit after installation:

from pathlib import Path
from tempfile import TemporaryDirectory

from veridist import (
    CsvLifetimeLimits,
    CsvLifetimeSchema,
    PublicSourceId,
    fit_exponential_csv,
)
from veridist.families import ExponentialFitSuccess

with TemporaryDirectory() as directory:
    path = Path(directory) / "lifetimes.csv"
    path.write_text("time,event_observed\n1,1\n1,0\n", encoding="utf-8")
    fit = fit_exponential_csv(
        path,
        schema=CsvLifetimeSchema("time", "event_observed"),
        source_id=PublicSourceId("src_0123456789abcdef0123456789abcdef"),
        limits=CsvLifetimeLimits(32768, 32768),
    ).fit
assert isinstance(fit, ExponentialFitSuccess)
assert fit.rate == 0.5
assert fit.inference == "not_provided"
assert fit.censoring_assumption == "independent_right_censoring"
print(f"rate={fit.rate}; events={fit.event_count}; censored={fit.censored_count}")

rate=0.5; events=1; censored=1

The first row is an observed event. The second had not occurred by the end of observation, so it is independently right-censored. rate is expressed in the inverse of the time unit in the CSV. A successful fit is not, by itself, proof that the exponential model is appropriate; use the model guidance below before making a decision.

Pick the right path

Supported work

Capability and evidence

The fitting surface contains fixed-location Exponential, Weibull-minimum, and Lognormal MLE cells for exact and independently right-censored lifetimes. A finite solution yields a point estimate; invalid statistical or operational conditions yield typed failures. Inference is restricted to that declared cell: the uncensored exponential cell supports refit Monte Carlo KS, AD, and CvM, AIC/BIC, and adequacy-gated selection with a caller-owned generator.

The public CSV path is strict: UTF-8 with exactly time,event_observed, event token 1, and right-censoring token 0. It uses one iterator pass. It is not a generic CSV reader. A successful call should be read with its execution record and model assumptions.

The CI gate checks supported Python versions, at least 95% global line and branch coverage, quality checks, package installation, and documentation. The Coverage ≥95% badge shows the pass/fail status of that enforced contract on main; this page does not claim a static percentage. The mutation and release- evidence badges link to their own verifiable workflows.

Scale and resume

Retained evidence covers a measured 10k/100k/1m by 32KiB/64KiB/128KiB matrix for the declared paths. It does not establish general big-data support, throughput, portable RSS, dataframe, Parquet, Arrow, database, distributed execution, broad censoring, vectorized operations, or universal model choice. SQLiteCheckpointStore is durable local state, not a distributed service. Use it with the same source revision and compatible store to continue a committed prefix after interruption. The executable SQLite recipe shows the necessary initialization and a compatible second pass.

Production readiness

Read KNOWN_LIMITS.md and the repository evidence ledger before production use.

Documentation, contribution, and support

Use the API reference to integrate, the family guide to assess the statistical surface, and the documentation toolchain to work on docs. For changes, start with the repository contribution guide and engineering conventions. Report reproducible defects through GitHub Issues and vulnerabilities through SECURITY.md.

The package uses BUSL-1.1 with an Apache-2.0 additional-use grant for personal, non-commercial use; see LICENSE.