FAQ


■ Service Overview

Q1. What is iHandify?

iHandify is an AI-powered handwriting recognition API that supports text, mathematical expressions, and scientific notation.


■ Authentication

Q2. How do I obtain an API key?

After subscribing, a Secret API key can be obtained from the dashboard.

Q3. What is the difference between a Secret API key and a Scoped Public API key?

  • Secret API key: A confidential key intended for secure use in backend systems
  • Scoped Public API key: A temporary key intended for frontend use with limited scope

Q4. How do I generate a Scoped Public API key?

A Scoped Public API key can be generated by sending a POST request to the designated endpoint using your Secret API key.


■ API Usage

Q5. Which API endpoint should I use?

Select an endpoint based on the input type, language, and intended recognition use case.

Q6. What input formats are supported?

  • Digital Ink (stroke data)
  • Image (Base64-encoded or multipart/form-data)

Q7. What is the difference between ink and image input?

Digital Ink includes stroke information and typically enables higher recognition accuracy.
Image input is static and may contain less contextual information.

Q8. What does the order parameter do?

Specifies whether stroke order should be considered during recognition.

Q9. What are specialized endpoints?

Specialized endpoints are predefined endpoints with fixed parameter values for specific use cases.

Q10. What happens if parameters conflict?

Parameters defined by the endpoint take precedence. Conflicting values in the request will be ignored.


■ Accuracy & AI Behavior

Q11. How can I improve recognition accuracy?

Accuracy may be improved by restricting recognition targets, optimizing input data, and configuring appropriate parameters.

Q12. What are targetFilters and targetCharacters?

These parameters define the scope of recognizable characters or symbols, which can help improve accuracy.

Q13. What is documentType?

Specifies the structural type of the input (e.g., character, word, sentence, expression).

Q14. What is candidateNum?

Defines the number of recognition candidates returned in the response.

Q15. What does lmWeight do?

Controls the influence of the language model on recognition results.

Q16. Will results always be consistent?

No. Results may vary due to model updates or internal processing changes.

Q17. Can I lock a specific model version?

Model version locking is not supported unless explicitly agreed upon.

Q18. Why can results change even with the same API version?

The underlying recognition engine may be updated independently to improve performance and accuracy.


■ Limits

Q19. What are the request size limits?

  • Digital Ink: up to 512 KB
  • Image: up to 2 MB (Base64) or 5 MB (multipart/form-data)

Q20. What happens if the limit is exceeded?

The request will be rejected with a 413 (Content Too Large) error.


■ Errors

Q21. What error codes may be returned?

  • 400 (Bad Request): Invalid request parameters
  • 403 (Forbidden): Authentication or authorization failure
  • 413 (Content Too Large): Request size exceeds limits

■ Data & Privacy

Q22. How is request data handled?

By default, request data may be used for service quality improvement and issue investigation.

Q23. Can I disable data usage?

Yes. By setting improveProduct=false, request data will be deleted after processing and will not be used for improvement purposes.

Q24. Does the SDK transmit data externally?

No. The SDK operates within the Customer’s environment and does not transmit data to external servers.


■ Production

Q25. Can the API be used in production environments?

Yes. The API is designed for production use, subject to the applicable terms and conditions.

Q26. Is uptime guaranteed?

No. Availability targets may be defined, but uptime is not guaranteed.


■ Legal

Q27. Can the results be used for decision-making?

Recognition results are provided for reference only. Customers are responsible for validating outputs before use.

Q28. Who is responsible for validating the results?

The Customer is solely responsible for validating and using the results.

Q29. Is recognition accuracy guaranteed?

No. The Service is provided on an "as is" basis without any guarantees of accuracy or fitness for a particular purpose.