Skip to content
MildportDocs
Get started

What is Mildport?

Mildport is the embeddable import wizard and self-hosted import engine for SaaS products and CRMs. Messy customer files — CSV, TSV, XLSX, ODS, JSON, XML, PDF tables, even OCR’d scans — become exactly the records your product expects.

It is proprietary, self-hosted-first software. Today you run the engine in your own infrastructure — Docker Compose or Kubernetes (Helm); a managed EU-hosted cloud opens with commercial access. Cloud alternatives like Flatfile, OneSchema and Dromo run on someone else’s infrastructure; Mildport is the option for when customer data must stay in your environment — regulated data, security reviews, air-gapped deployments.

Two halves, one public contract

Section titled “Two halves, one public contract”
  • The widget — <mildport-import>, a framework-agnostic custom element you embed in your app. Attributes in, DOM events out. Published as @capitality-io/mildport-widget (with a React wrapper). See Install and Frameworks.
  • The engine — a stack you run in your own infrastructure via Docker Compose or the Helm chart: the service, MongoDB, decode sidecars (XLSX, PDF, OCR) and optional object storage. See Self-hosting.

The widget talks to the engine over a public, license-keyed REST API — the same surface any host uses, including us. The API is URL-versioned with a published breaking-change policy.

What makes it different

Section titled “What makes it different”
  • Deterministic first, AI when it earns it. An explainable engine scores header and value matches: same input, same answer. An optional, evidence-gated AI judge re-ranks only the uncertain tail — bring your own model, rationale shown in the grid, off by default.
  • Runs in your infrastructure. Files never leave your VPC. Licenses verify offline against a public key you deploy — no license server, no phone-home.
  • Explainable and replayable. Visible match scores, an event-sourced normalization pipeline you can replay, and webhook deliveries with an audit trail.
  • Schema changes without releases. The import schema can live server-side: edit fields in an embeddable no-code editor — or draft them from your API spec — publish, and every embed updates on its next load. Versioned, with one-click rollback. See Target catalogs.

Next steps

Section titled “Next steps”