Platform Capabilities

Full Feature Overview

Ocriva is organized around the five stages of the Document Automation Transformer (DAT) pipeline: Ingest, Extract, Transform, Automate, and Integrate. This page documents every major capability, organized by the stage it belongs to.

Stage 1: Ingest

The Ingest stage handles document receipt — accepting files from wherever they originate and preparing them for processing.

Single File Upload

• Drag-and-drop interface in the web application
• Click-to-browse file selection
• Immediate upload confirmation with file preview
• Supported formats: PDF, JPG, PNG, WebP, BMP

Batch Upload

• Upload up to 50 files in a single submission
• Drag-and-drop multiple files simultaneously
• Per-file status tracking during batch upload
• Automatic queueing for batch processing
• Combined export on batch completion

API Upload

• REST API endpoint for programmatic document submission
• Supports multipart/form-data for file uploads
• Returns processing ID for status tracking
• Authenticated via JWT or API token
• Full OpenAPI/Swagger documentation available

Webhook-Triggered Ingestion

• Accept documents from upstream systems via webhook events
• Event-driven processing pipelines
• Configurable per project

LINE Integration

• Mobile document capture via LINE messaging
• Users photograph documents on mobile devices and submit via LINE
• Documents are routed to the configured Ocriva project automatically
• Results can be returned to the LINE conversation

Stage 2: Extract

The Extract stage is where AI processes the document and identifies the data you need.

AI Providers

Ocriva connects to six AI providers. Each template specifies which provider and model to use.

Multi-Language Support

• Thai and English in the same document, handled in a single extraction pass
• No separate language configuration required
• Character recognition for Thai script, including common handwriting variants
• Mixed-language documents (e.g., Thai invoices with English product codes) extracted accurately

Extraction Modes

Structured Extraction (JSON Schema)

Define a JSON schema in the template. Ocriva instructs the AI to return data that conforms to the schema. Output is always valid, typed JSON.

Example schema:

{
  "invoice_number": "string",
  "vendor_name": "string",
  "total_amount": "number",
  "due_date": "date",
  "line_items": [
    {
      "description": "string",
      "quantity": "number",
      "unit_price": "number"
    }
  ]
}

Free Text Extraction

Provide a prompt describing what to extract. The AI returns a text response. Useful for summaries, contract analysis, or any extraction that does not map to a fixed schema.

Templates

Templates are reusable extraction configurations. Once created, a template processes every document of the same type in the same way.

Each template defines:

• Template Name — for identification
• AI Provider and Model — which AI to use
• Extraction Mode — structured JSON schema or free text
• Prompt / Instructions — what to extract and how to handle edge cases
• Output Format — the downstream format for results
• Project Assignment — which project the template belongs to

Preset Templates

Ocriva includes 19 preset templates for common document types:

Custom Templates

Create custom templates for document types not covered by presets:

• Define any field schema for any document type
• Write custom AI instructions tailored to your specific documents
• Configure normalization rules (date formats, number formats, name ordering)
• Test extraction on sample documents before deploying
• Iterate and improve based on extraction history

Data Files (RAG)

Attach reference documents to a template to improve extraction accuracy:

• Upload company-specific glossaries, numbering conventions, or codebooks
• The AI uses these reference documents as context during extraction
• Useful for internal document types with organization-specific conventions
• Files are scoped per template, not shared across templates

Stage 3: Transform

The Transform stage converts extracted data into the format your downstream systems expect.

Output Formats

CSV Configuration

• Configure column ordering for CSV exports
• Map extracted field names to custom column headers
• Choose delimiter (comma, semicolon, tab)
• Batch CSV combines rows from multiple documents into a single file

Image Generation Mode

• Generate a visual representation of extracted data overlaid on the source document
• Useful for review workflows where annotated documents need to be stored alongside records
• Output as image or PDF

Batch Export

• Process 50 documents in a batch and export all results as a single file
• Combined JSON array, combined CSV, or individual file per document
• Configurable at the template level

Stage 4: Automate

The Automate stage handles volume, scheduling, and reliability — turning document processing into operational infrastructure.

Batch Processing

• Submit up to 50 documents in a single batch
• Real-time progress tracking: each document shows pending, in_progress, completed, or failed status
• Parallel processing — multiple documents processed simultaneously
• Batch-level status summary: total count, completed count, failed count
• WebSocket-powered live UI updates — no page refresh required

Processing Queue

• All documents enter a managed queue on submission
• Queue processes in order with configurable concurrency
• Automatic retry on transient failures
• Failed documents are flagged with error details for review
• Queue status visible in the web interface and via API

Processing History

• Every document processed through Ocriva has a permanent history record
• Record includes: source document, applied template, AI model used, status, timestamps, result data
• Filter and search history by project, template, status, date range
• Re-process any historical document with the same or updated template
• Export history records for audit or reporting

Credit System

• Processing each document consumes credits based on document complexity and AI model used
• Credits are purchased in advance via Stripe
• Real-time credit balance displayed in the web interface
• Low-balance alerts configurable per organization
• Detailed credit usage history by project, template, and AI provider
• Credit consumption is auditable: each processing record shows credits consumed

Organization → Project → Template Hierarchy

Organization (billing entity, team management)
└── Project (logical grouping for a use case or department)
    ├── Templates (extraction configurations)
    ├── Processing History
    ├── Webhooks
    └── API Tokens

• Multiple organizations per account (for agencies or multi-entity companies)
• Multiple projects per organization (finance, HR, logistics, etc.)
• Multiple templates per project (invoice extractor, receipt scanner, etc.)
• Billing and credits managed at the organization level

Status Tracking

Every document processing record has one of four statuses:

Stage 5: Integrate

The Integrate stage delivers results to the systems that need them, automatically.

REST API

• Full CRUD API for all Ocriva resources
• Endpoints for: organizations, projects, templates, uploads, processing history, analytics
• Authentication via JWT (session-based) or API tokens (service-to-service)
• Paginated list endpoints with filtering and sorting
• OpenAPI 3.0 specification available at /api/docs
• Swagger UI for interactive API exploration

API Tokens

• Generate API tokens per project for external integrations
• Tokens are scoped to a specific project
• Multiple tokens per project (one per integration)
• Tokens can be revoked without affecting other integrations

Webhooks

Configure webhook endpoints to receive event notifications when processing events occur.

Supported Events:

Webhook Payload Structure:

{
  "event": "document.processed",
  "timestamp": "2026-03-31T10:15:00Z",
  "organizationId": "org_xyz",
  "projectId": "proj_abc",
  "processingId": "proc_123",
  "status": "completed",
  "result": {
    "fields": { ... },
    "format": "json"
  }
}

• Webhooks include a signature header for payload verification
• Failed webhook deliveries are retried with exponential backoff
• Delivery history and status visible per webhook

WebSocket Real-Time Notifications

• Connect via Socket.IO for real-time processing events
• Processing status updates pushed to connected clients as they happen
• Batch progress events streamed live
• Credit balance updates pushed in real time
• No polling required — the UI stays current automatically

LINE Integration

• Connect Ocriva to a LINE Official Account
• Users send document photos directly via LINE
• Documents are processed per the project's default template
• Results returned to the LINE conversation (configurable)
• Useful for mobile-first workflows: field staff, delivery personnel, branch operations

Notifications

• In-app notifications for processing completions, errors, and credit alerts
• Email notifications for batch completions and low-credit warnings
• LINE push notifications for connected accounts
• Notification preferences configurable per user

OpenAPI / Swagger Documentation

• Interactive API documentation at /api/docs
• Full request/response schemas for all endpoints
• Authentication integrated into the Swagger UI
• Download OpenAPI spec for code generation or API client tools

Security & Access Control

• JWT authentication — session-based auth for web application users
• API token authentication — token-based auth for programmatic integrations
• Organization-level data isolation — organizations cannot access each other's data
• Project-level scoping — API tokens are scoped to individual projects
• httpOnly cookies — session tokens stored in httpOnly cookies, not accessible to JavaScript
• Input validation — all API inputs validated with class-validator (NestJS)
• Webhook signature verification — payloads signed with HMAC for authenticity verification