How it works

BiomAPI selects a processing method from your file type. It validates the result and returns measurements for you to review. Processing and BiomPIN retrieval return the same response structure.

From file to result

Input	Processing	Result
PDF, JPG/JPEG, PNG, GIF, BMP	BiomAI reads the report with Gemini	Core biometry measurements and extraction provenance
JSON	BiomJSON validates the supplied data	A standardized response with preserved or direct-entry provenance
Manual entry or a web-app edit	The app submits JSON through BiomJSON	Validated data attributed to BiomDIRECT

File routing uses the extension. PDF and image uploads also receive basic content checks. Unsupported formats or mismatched content are rejected before extraction.

The three processing paths

BiomAI — LLM Extraction

BiomAPI prepares the report for Gemini, requests structured measurements, and validates the returned data. Preparation can change the format sent to the model to improve processing. The web app keeps your original available for comparison during the session.

BiomAI extracts the core report fields. For IOLMaster700 reports, it also reads printed posterior and total keratometry into extra_data. Add free-text notes through manual entry or JSON. See supported devices and extraction limits.

The service uses a managed Gemini credential unless you supply your own Gemini key. BYOK always uses AI Studio and does not fall back to managed credentials if your key fails.

Standard processing costs 1 credit per PDF/image. Slow costs 0.5 credit and can take much longer. BiomAPI may retry temporary provider failures within the request deadline. See credits and limits for what counts toward your allowance.

BiomJSON — validation and round-trips

BiomJSON accepts a report inside a top-level data object or a saved BiomAPI response. It checks the structure, allowed values, and declared schema version without calling Gemini. Numeric values must use the units defined in the schema.

You can extract a PDF, download its JSON, and upload that JSON again later. BiomJSON preserves the declared BiomAI provenance, which records the source of the data. It does not compare the JSON with the original report or detect edits to a downloaded file.

If you edit extracted values outside the web app, declare the result as BiomDIRECT rather than retaining BiomAI as its current provenance. The web app makes that change automatically when you save an edit.

BiomDIRECT — direct data entry

BiomDIRECT labels the source of the data. BiomJSON validates it. Manual transcription, web-app edits, and JSON without BiomAI provenance receive the BiomDIRECT label. Integrations can also declare their source application and version.

Start with JSON input / BiomDIRECT for required objects, examples, and the metadata BiomAPI adds. The response schema describes the completed output.

Read the result

Field	Contents
`data`	Device, examination date, patient context, and right/left eye measurements
`extra_data`	Optional notes, posterior keratometry, or total keratometry
`metadata`	Request ID, schema/app versions, and extraction provenance
`biompin`	Sharing code and expiry, or `null` if no share was created

A missing measurement is null. It does not mean zero, normal, or confirmed absent. Validation checks whether a value fits the schema. It cannot confirm that the value came from the correct eye or is clinically correct.

A BiomPIN points to an encrypted, temporary copy of a result. It expires after 31 days by default. Patient initials and ID are removed before storage; other report fields remain.

The web app can include initials and ID separately in a shared link, or omit them when you select Anonymous. That choice does not anonymize all the remaining clinical data. See BiomPIN sharing and identity before sharing a result.

For processing destinations, storage, and retention details, read the Privacy policy.