The synthetic PostgreSQL-to-React count tracer provides an R entry point, a validated partial aggregate bundle and an exact-count React view. The observation tracer adds observed people, monthly people and unique person-days with a Recharts view. The measurement reference report adds counts, representations, raw units/operators and exact distributions with local terminology. An optional weighted strategy computes the same exact quantiles using cumulative frequencies; the reference remains the default. These are development slices, not a complete alpha report.
After installing the R package, open an existing supported result bundle with
AchillesNext::viewResults("path/to/bundle"). The installed local viewer
opens automatically with prebuilt React assets; viewing needs no Node, npm or Shiny.
Open-source OMOP data characterization and exploration with reproducible aggregate results and an embeddable analytical UI.
R package name: AchillesNext · Repository: achilles-next · License: Apache-2.0
AchillesNext is in early development. The R package installs from source and computes synthetic PostgreSQL person/observation counts and a bounded measurement reference report, with an installed local React viewer. A separate PostgreSQL batch generator now produces complete restricted core candidates with saved metadata. Checked full-refresh execution is available; production host integration remains under development. See the count-tracer guide for the tested fixture scope, installation and limitations.
- Generate descriptive summaries of an OMOP CDM within the institution.
- Preserve metric definitions, populations, units, availability states and provenance alongside aggregate results.
- Open generated results through a prebuilt viewer.
- Reuse analytical views within other applications, including ATLAS3.
Read CONTRIBUTING.md for the contribution and Git workflow. Source code, tests and documentation explaining package behavior belong in this repository.
The PostgreSQL fixture environment provides reproducible synthetic CDM 5.3/5.4/5.5 sources, optional pinned GiBleed 5.3/5.5 datasets, and separate empty schemas for later SynPUF loading. This is developer test infrastructure; the R package and analysis engine are still under development.
The alpha measurement profile specifies 15 bounded outputs with explicit observation, representation and exact-distribution semantics. Its manifests and read-only fixture references can be checked now; the production analysis engine is not implemented yet.
The v1 result/provider/embed contracts add versioned schemas, a complete synthetic bundle, TypeScript interfaces and executable conformance checks for Generation → View and host embedding.
The synthetic PostgreSQL reference benchmark checks pinned legacy counts/distributions against an independent oracle and captures separate execution/export timings on fixture and deterministic cardinality workloads.
The experimental core batch planner enumerates all 142 frozen core results across 51 dependency nodes, with explicit shared-input candidates and exact grains. It plans full-history/all-concept work; complete core generation is available through the complete candidate generator; checked full-refresh execution is available. A small synthetic harness compares reference and transaction-private materialization with independent numerical oracles.
The separate registered analysis example demonstrates a two-report contribution through generation, approval and saved readback. It preserves the core inventory and does not complete core generation.
generatePopulationObservationBaseline() computes and persists seven population
and twelve observation-period outputs from PostgreSQL sources. The
baseline guide explains exact semantics,
worker permissions and saved readback without CDM access. Its building release
leaves the other 123 core outputs explicitly ungenerated and cannot be published
as a complete core result.
generateDomainBaseline() adds the eight-domain and density baseline: 100 frozen
outputs with the same 19 population dependencies, for 119 of 142 outputs accounted
for. The domain guide covers batch generation,
exact concept/month/year selection and paged raw concept catalogues without CDM
access. Conditional unsupported fields retain explicit states. These restricted building releases cannot be
published as complete core results or opened in the v1 tracer viewer.
generateMeasurementBaseline() adds all nineteen whole-source measurement
outputs, bringing the restricted baseline to 138 of 142 outputs. The
measurement baseline guide describes
representations, units/operators, exact numeric and reference-range summaries,
and saved readback. That entry point retains its partial baseline boundary. The separate
complete candidate generator adds the four real metadata outputs and supports sealing/native validation. Full refresh execution is described below; integrated publication qualification
remains in progress. The earlier synthetic measurement tracer and its viewer stay separate.
The refresh contracts add source-free
planCharacterizationRefresh() requests and a read-only inspectResultStore()
version probe. executeCharacterizationRefresh() recomputes a checked plan into
a fresh complete candidate, retaining the previous publication through failures.
Sealing, validation, institutional approval and CAS promotion remain explicit.
The ATLAS3 integration boundary records a native v3.0.0 fixture proof, its lifecycle constraints and the experimental authenticated HTTP result transport. A production viewer/host adapter is still under development.
AchillesNext is an independent successor project. OHDSI Achilles and other ecosystem projects retain their own identities and attribution. Existing functionality is not claimed as implemented or certified here.
The optional durable execution API and CLI persist a bounded plan, operation states and measured telemetry for synthetic measurement runs. Inspect records after process exit without reconnecting.
The optional durable run API now generates all fifteen alpha outputs for marked synthetic PostgreSQL fixtures. Approved publication validates immutable restricted/distributable releases before switching a local current pointer. Cancellation and isolated recovery record targeted PostgreSQL cancellation and linked, idempotent attempts; unknown source identity forces recomputation. Institutional execution, general anonymization and cross-run reuse remain outside this implementation.
See LICENSE for the Apache License 2.0. Preserve applicable notices when reusing third-party code.