OCR Engine Status
OCR support lets AI workflows read scanned PDFs and images when the deployment has selected an approved OCR engine. Novantra treats OCR as an optional deployment capability inside the AI feature bundle. It is not required for every AI deployment, and it is physically absent from no-AI profiles.
The OCR status view shows whether OCR is configured, which connection mode is used, whether authorization material is present, which content types are eligible, and whether an extractor is registered through the artifact text extraction seam.
What The Status Means
| Status | Meaning |
|---|---|
| Configured | The deployment has selected an OCR endpoint or engine and the configuration shape is valid. |
| Not configured | No deployment OCR endpoint is configured. Built-in native text extraction can still handle supported text-based files. |
| Misconfigured | OCR configuration exists, but Novantra rejected the configuration shape before use. |
| Registered | The deployment has registered an artifact text extractor that AI can call through the governed artifact seam. |
The status view never displays OCR endpoint URLs, tokens, authorization headers, artifact storage paths, or extracted document text.
Configuration Responsibilities
Cloud and Sovereign deployments may use different OCR operating models:
- Cloud deployments can use a Novantra-approved or customer-approved OCR service when the AI package is enabled.
- Sovereign deployments can point to a customer-hosted OCR service, including a local endpoint where the install policy allows it.
- No-AI Sovereign profiles exclude the AI feature bundle and therefore exclude OCR routes and runtime code shipped by the AI package.
Operators configure OCR outside the workspace UI. The workspace UI only shows sanitized readiness posture and evidence.
Supported deployment settings include:
| Setting | Purpose |
|---|---|
AI_OCR_HTTP_ENDPOINT | OCR endpoint selected by the deployment. |
AI_OCR_HTTP_AUTHORIZATION or AI_OCR_HTTP_TOKEN | Optional authorization material resolved by deployment configuration. |
AI_OCR_HTTP_SUPPORTED_CONTENT_TYPES | Optional comma-separated content types that OCR should accept. |
AI_OCR_HTTP_TIMEOUT_MS | Optional request timeout. |
AI_OCR_HTTP_ENGINE_KIND and AI_OCR_HTTP_ENGINE_VERSION | Optional engine provenance labels. |
Use placeholder values in runbooks and tickets. Do not paste production endpoint URLs, tokens, headers, or document content into support notes.
Evidence Handling
When OCR is used during a governed AI run, Novantra records sanitized provenance such as extractor kind, content type, engine kind or version, confidence, and page or region citations where the selected engine provides them.
OCR does not make AI the owner of the source artifact. The source artifact stays with DMS, artifacts, storage, or the owning source module. AI receives text only through the approved artifact text extraction seam and stores source refs and sanitized evidence, not storage keys or raw source files.