Regulated & GxP Environments
Reproducible, auditable, self-describing figures for pharma, clinical and other regulated work — and an honest account of what CanvasXpress provides versus what your validation must cover.
Why CanvasXpress fits regulated work
CanvasXpress came out of biopharma, and three of its core properties are exactly what a GxP or 21 CFR Part 11 workflow needs: reproducibility (a figure serializes to one portable specification that renders the same result), a replayable audit trail (every interaction is a recorded, re-playable operation), and self-describing exports (the data and the specification travel inside the exported figure). Together these make a figure something you can regenerate, inspect, and defend rather than a static picture whose provenance is lost.
How the features map to 21 CFR Part 11 themes
Part 11 is a regulation about your system, not about a charting library — compliance is a property of your validated application and its controls. What a visualization component can do is support those controls. Here is an honest mapping:
| Part 11 theme | What CanvasXpress provides | What remains your responsibility |
|---|---|---|
| Accurate, reproducible records | A figure serializes to one portable JSON specification that re-renders the same result; exports embed the underlying data. | Storing, versioning and controlling those records in your validated system. |
| Audit trail | A replayable, in-figure operation log (the audit trail) capturing the interactions that produced the view. | User-attributed, time-stamped, tamper-evident audit records at the system level (identity, retention). |
| Access & authority controls | Client-side rendering with no hidden network calls in the free, attribution-visible tier; data connectors keep credentials server-side. | Authentication, authorization, and role controls in your application. |
| Record integrity / provenance | Self-describing exports carry data + spec + version together; optional signed provenance — a verifiable hash over data, specification and version — is available via CanvasXpress.provenanceHash. |
Electronic-signature binding and long-term record protection. |
| System validation | A public, versioned build with a large automated visual-regression suite and a stated quality and license position. | Your IQ/OQ/PQ of the deployed system in its intended use. |
CanvasXpress does not, by itself, make a system Part 11 compliant, and no such claim is made here. It provides features that support a compliant, validated workflow.
AI assistant and data egress
The built-in AI assistant never has to leave your perimeter. llmServiceURL points every request at a canvasxpress-mcp server you host, which holds the provider key (or runs fully local models through Ollama); and a single page-level policy — <meta name="canvasxpress-llm-outbound" content="off"> or CanvasXpress.llmOutbound = false — blocks every outbound inference request from every chart on the page. The block is enforced at the two request paths and verified by a network-intercepting regression test, so a validation file can cite it as a control rather than a setting. See Self-hosted inference and the outbound policy.
Validation support
To help teams validate a deployment, CanvasXpress offers, today, the evidence a validation effort draws on: a public versioned source and build, a documented visual-regression suite, a clear version history, and reproducible figure specifications. Formal IQ/OQ documentation templates and a validation package are in development; if you need them for an active program, contact us to discuss availability and scope.
Submission-grade figures
Because a CanvasXpress figure carries its data and specification, a figure in an analysis or a filing can be regenerated exactly and inspected by a reviewer — the opposite of a flattened image whose numbers can no longer be checked. Kaplan–Meier curves, volcano plots, and clustered heatmaps — common in clinical and translational work — are first-class chart types, not add-ons.
On the roadmap
Two items previously listed here have shipped: a versioned, published specification schema with a stated long-term compatibility policy (a 2026 figure still renders in 2031), and optional signed provenance (a verifiable hash over data, specification and version). What remains planned:
- Formal IQ/OQ documentation and a validation package.
- A VPAT and deeper accessibility for public-sector and pharma procurement.
These are stated as planned, not shipped — we would rather be precise than imply a capability that is not yet in your hands.
