Frequently Asked Questions
Find answers to common questions about Entity Detector API.
Jump to:
Getting Started
How do I get an API key?
Sign up for a free account at entitydetector.com/register, then create an API token from your dashboard. The token is shown only once, so save it securely.
Is there a free tier?
Yes! Every account gets 100 free API requests per day. No credit card required to start. The free tier includes full access to all API features.
How long does it take to get started?
You can make your first API call within minutes. Sign up, create a token, and send a request. Check our Quick Start guide for step-by-step instructions.
API Usage
What languages are supported?
Entity Detector supports text in any language. The AI model can detect entities across most major languages including English, Spanish, French, German, Chinese, Japanese, and more. You can provide a language hint for better accuracy.
What is the maximum text length?
Text must be between 200 and 30,000 characters. For longer documents, split them into chunks and process separately.
How accurate is the entity extraction?
Accuracy varies by content type and language, but typically ranges from 85-95% for well-structured text. The API provides confidence scores for relations to help you filter results.
Can I process HTML content?
The API expects plain text. Strip HTML tags before sending content. Most programming languages have libraries for this (e.g., cheerio for Node.js, BeautifulSoup for Python).
Are results cached?
Yes, identical text inputs are cached for 24 hours. Cached responses are nearly instant and still count toward your quota. The response includes a "cached" field indicating if the result was from cache.
Entities & Relations
What types of entities are extracted?
The API extracts: Persons (people names), Organizations (companies, institutions), Locations (places, addresses, countries), and Objects (physical or conceptual items mentioned in the text).
What relation types are detected?
Common relation types include: works_for, located_in, owns, created_by, part_of, interacts_with, and many more. The specific types depend on the context of the text.
What do the confidence scores mean?
Confidence scores range from 0.0 to 1.0 and indicate how certain the model is about a detected relation. Scores above 0.8 are generally reliable. Consider filtering out low-confidence results for production use.
How is sentiment determined?
Sentiment (positive, negative, neutral) is analyzed in the context of each relation. It reflects the tone of how the source and target entities are related in the specific text passage.
Pricing & Limits
How does pricing work?
We offer monthly subscription plans based on daily request quotas. The Free tier includes 100 requests/day. Paid plans start at $29/month for 1,000 requests/day. See our pricing page for details.
What happens if I exceed my quota?
You'll receive a 429 (Too Many Requests) error with a "retryAfter" field indicating when to retry. Quotas reset at midnight UTC. Consider upgrading your plan if you consistently hit limits.
Can I upgrade or downgrade my plan?
Yes, you can change plans anytime from your dashboard. Upgrades take effect immediately. Downgrades apply at the start of your next billing cycle.
Do failed requests count toward my quota?
No, requests that result in 4xx or 5xx errors do not count against your quota. Only successful (200) responses count.
Security & Privacy
Is my data stored?
Text content is processed in memory and cached (hashed) for 24 hours to speed up repeat requests. We do not permanently store or use your text data for training. Cache entries are automatically purged.
Is the API secure?
Yes, all API communication is encrypted via HTTPS/TLS. API tokens are stored as secure hashes. We recommend using environment variables and rotating tokens periodically.
Can I use this for sensitive documents?
While we take security seriously, evaluate your compliance requirements. For highly sensitive data, contact us about dedicated infrastructure options.
Troubleshooting
Why am I getting a 401 error?
This means authentication failed. Check that: 1) Your Authorization header uses "Bearer " prefix, 2) The token is correct and not revoked, 3) There are no extra spaces around the token.
Why am I getting a 400 error?
This indicates invalid input. Common causes: text too short (<200 chars) or too long (>30,000 chars), invalid JSON syntax, or missing required fields. Check the error message for specifics.
The API is slow. What can I do?
First-time requests for new text take 1-5 seconds for analysis. Subsequent identical requests are cached and nearly instant. Consider implementing your own cache layer for frequently processed content.
Results seem inaccurate. How can I improve them?
Try: 1) Providing a language hint if text isn't in English, 2) Using clean, well-formatted text, 3) Removing boilerplate/navigation content, 4) Filtering by confidence score.