Deployment

Start from an identified build

Use a packaged release for routine engineering work and retain its verification evidence with controlled records. Build from source when implementation review, development, or internal packaging is part of the workflow.

Download the current release

Select the package for your engineering workstation. The links resolve to assets from the latest tagged release and remain current as new versions are published.

Need release notes, verification reports, checksums or a previous controlled version? View the latest release on GitHub ↗

Retain and verify the release evidence

Each release includes a verification bundle, standalone HTML and JSON reports, platform environment manifests and SHA-256 checksums. The report identifies the exact commit, test and build components, reference checks, API contract, and packaged application files represented by that release.

  1. Archive the application package, verification bundle and published checksum together.
  2. Compare downloaded files with the release SHA-256 records.
  3. For a controlled workflow, preserve the acceptance decision and intended-use validation with the release evidence.
Scope: Build evidence shows what was executed for a specific revision. It does not establish regulatory approval or validate the suitability of a model for an organization's particular application. Review the full assurance approach →

Build from source

Use the source build to review the implementation, contribute changes, produce an internally controlled package or support another platform. It requires:

  • Python 3.11 or 3.12
  • uv for the locked Python environment
  • Node.js 24 LTS and npm

Clone the repository, then run the three steps below from the project root:

bash
# 1. Exact Python environment from the checked-in lock
uv sync --frozen --extra test

# 2. Exact front-end dependencies
npm ci --prefix gui/frontend

# 3. Launch app — API on :8000, UI on :5173
bash gui/start.sh

Then open http://localhost:5173 in your browser. The application requires no central project database or user-account store.

Deploy for an engineering team

Perdura’s stateless backend can serve an engineering team from one container behind a TLS-terminating, authenticating reverse proxy. Project files remain under user control rather than in a shared application database. Central deployment does not alter the underlying calculation methods.

bash
cp .env.example .env      # set domain + reverse-proxy password hash
docker compose up -d --build
Deployment control: Perdura has no built-in authentication. Place a shared deployment behind the provided reverse proxy, a VPN or a network allowlist. See docs/DEPLOYMENT.md for TLS, SSO, scaling and an nginx alternative.

Prepare reliability data

Life-data analyses accept a simple CSV representation of failures and suspensions:

data.csv
value,type
100,F
150,F
200,S

Set type to F for a complete failure time or S for a suspension (right-censored observation). If the column is omitted, every row is treated as a failure. Statistical Modeling, MSA and Hypothesis Tests also accept CSV/TSV data with auto-detected delimiters, manual entry and spreadsheet paste.

Licensing for engineering use

Personal, academic and other non-commercial analysis is covered by the PolyForm Noncommercial License 1.0.0. Internal organizational work, consulting and other commercial reliability use require a separate license. See the license page or email djtroyal@gmail.com.