Settings & Configuration

Manage all Ocular OCR system settings and configuration

My Provider/Model Preferences

Set your personal defaults for providers and models. These do not change global system settings.

Comma-separated values used for the Upload dashboard provider selection.
Comma-separated values used in the real-estates structured extraction step.
Comma-separated values for the Property OLMocr cloud provider selector.
Comma-separated values for Property OLMocr model multi-select.
Default OLMocr model value for Property OLMocr workflows.
Configuration Overview

Application Settings

Main application configuration and behavior

Basic Configuration
Display name for the application
Application description
Application version number
Current deployment environment
⚠️ Disable in production
Server Configuration
Server bind address
Server port number
Number of worker processes
File Processing
Maximum upload file size
Maximum files in batch upload
Comma-separated file extensions
Processing Configuration
Default OCR processing strategy
Comma-separated provider list
Maximum concurrent OCR requests
Static Files
Path to Jinja2 templates
Path to static assets (CSS, JS)
Disable for Cloud Functions

Database Configuration

Database connection and pool settings

SQLAlchemy database connection string
Examples:
SQLite: sqlite:///data/ocular.sqlite
PostgreSQL: postgresql://user:pass@localhost:5432/db
MySQL: mysql+pymysql://user:pass@localhost:3306/db
Number of persistent connections
Additional connections beyond pool size

Authentication Configuration

User authentication and authorization settings

Clerk Authentication
Clerk backend secret key
Clerk frontend publishable key
JWT Configuration
Secret key for JWT token signing
User session timeout duration

Payment Processing Configuration

Stripe payment integration settings

Stripe backend secret key
Stripe frontend publishable key
Webhook endpoint signing secret
Stripe Setup
  1. Create account at stripe.com
  2. Get API keys from Dashboard > Developers > API Keys
  3. Set up webhook endpoint for payment events
  4. Copy webhook signing secret

AI Models Configuration

Configure deployed AI models on cloud GPU providers (RunPod, Vast.ai, DeepInfra)

Cloud Provider Configuration
Infrastructure settings for all AI model deployments
Select your preferred cloud GPU infrastructure provider
RunPod Configuration
RunPod serverless endpoint ID
Get from RunPod Settings > API Keys
RunPod data center region
S3 Storage
RunPod S3 access key
RunPod S3 secret key
S3-compatible storage endpoint
Network volume for model storage
OLMocr (AllenAI OCR Model)
State-of-the-art OCR model from AllenAI - allenai/olmOCR-7B
About OLMocr: Advanced OCR model trained on diverse document types. Excellent for extracting text from complex layouts, handwritten documents, and multilingual content.
HuggingFace model identifier
Maximum file size for processing
Request timeout duration
Comma-separated file extensions
RoLMocr (Rotated OCR Model)
Specialized for rotated and skewed document processing
About RoLMocr: Optimized for documents with rotation, skew, or non-standard orientations. Handles mobile captures and scanned documents with varying angles.
Model identifier for RoLMocr
Maximum file size for processing
Request timeout duration
Specific endpoint for RoLMocr model
Pixtral (Vision Model)
Mistral's vision model for document understanding
About Pixtral: Multimodal vision model from Mistral AI. Excels at document analysis, table extraction, and visual question answering.
Model identifier for Pixtral
Maximum file size for processing
Request timeout duration
Specific endpoint for Pixtral model

Cloud Provider Configuration

API keys and settings for cloud OCR and AI providers

Mistral AI
Vision LLM for document OCR and analysis
Mistral vision model
Google Cloud Vision
Enterprise-grade OCR with excellent accuracy
Path to service account JSON file
GCP project identifier
Setup: Enable Vision API in GCP Console → Create service account → Download JSON key
AWS Textract
Advanced document analysis with form and table extraction
AWS access key
AWS secret key
AWS region
Setup: IAM Console → Create user with Textract permissions → Generate access keys
Azure Document Intelligence
Microsoft's AI-powered document processing
Azure Document Intelligence endpoint
Azure API key
Setup: Azure Portal → Create Document Intelligence resource → Copy endpoint and key
Tesseract OCR
Open-source OCR engine for local processing (no API key required)
Primary OCR language
Image DPI (150-600, default 300)
Installation Required: Tesseract must be installed on your system. Download here
HuggingFace
Access to gated models and private repositories
Required for gated models and private repos
Get Token: HuggingFace Settings → Create new token with "read" permissions

Logging Configuration

Application logging and monitoring settings

Minimum logging level
Log output format
Optional Logfire integration token
Write logs to file
Path to log file

Security Configuration

Security, CORS, and access control settings

Comma-separated origins (* for all)
Maximum HTTP request body size
Requests per minute per IP
Secret key for API authentication
Key for encrypting sensitive data
⚠️ Required for production
Production Security:
- Use specific CORS origins
- Enable HTTPS only
- Set strong API keys
- Disable debug mode

Cache Configuration

Caching settings for improved performance

Redis connection string (leave empty for in-memory cache)
Redis Examples:
Local: redis://localhost:6379/0
Auth: redis://:password@localhost:6379/0
Cloud: rediss://user:pass@host:port/db
Time-to-live for cached items (1 hour = 3600s)
Cache OCR results for faster repeated requests