KeXtract™
Cost and operating parameters
- The service is billed per token, and usage is tracked against the customer's API key.
- Documents can be uploaded as PDFs or images (JPG, PNG), up to 100 pages or 50 MB per file.
- Input and output tokens are priced differently: output tokens generally cost more per token, but tend to be fewer in number than input.
How billing works:
1
Token billing
Each API call consumes input tokens (text or images sent) and output tokens (the generated response). Token usage is tracked per API key and aggregated for each billing period.
2
Token input
Includes all text from the document, any content sent to the model, and metadata required for processing.
3
Token output
Includes the exact response.
Use the following formula to estimate costs of a single processing operation:
Cost = Token input x Cost Token input + Token output x Cost Token output
- Input token = estimated number of tokens derived from the submitted document
- Token output = estimated number of tokens in the JSON/extract response
- Token input cost and Token output cost = price per input and output token, as specified provided in the pricing plan
Tip:
Run a few test files in the Playground to measure the actual average token consumption (input/output), then multiply by the expected monthly volume.
Structured JSON extraction: output tokens (the response) are significantly fewer than input tokens (document content plus the KeXtract™ schema and instructions).
Markdown or Plain text extraction: output tokens include the entire text extracted from the document.
Do you have any questions? Write to us at info@kextract.it
