Skip to content
Hxtory

Hxtory JSON · version 1

Read your documents with your own AI

Hxtory JSON is a simple file format for your history: cases and the entries in them. Your own AI reads your documents and writes the file, and Hxtory imports it. Neither the documents nor the file are sent to us.

How it works

  1. 1

    Give your AI the prompt

    Copy the prompt below into the AI you trust, together with your documents: photos, scans or PDFs.

  2. 2

    Save its answer

    Your AI answers with one JSON file. Save it as a .json file, for example in Files on your iPhone.

  3. 3

    Import it in Hxtory

    Documents → Import Hxtory JSON, then pick the file. Hxtory reads it on your iPhone, without AI.

  4. 4

    Check every entry

    The entries open for review, as after any reading. You confirm or change each one, then they go into your cases, or into Not in a case.

Who sees your documents

  • Only the AI you choose. If it runs on your own computer, your documents never leave it. If it's an online service, its own privacy terms apply.
  • Hxtory imports the file on your iPhone and sends nothing anywhere. After that, the entries are part of your history like any others.
  • The file has no place for your name, date of birth, blood type, height or weight. Hxtory keeps those in About you, and ignores them if a file has them.

The prompt

In English, so every AI reads it the same way. The entries stay in your documents' own language.

Prepare a Hxtory JSON file (version 1) from the medical documents I give you. Hxtory is a personal record-keeping app. The file is imported into it, and I check every entry before it's saved.

Rules:
1. Copy, never interpret. Write names and values exactly as the document prints them, in its own language, with units and qualifiers. Don't explain, judge, diagnose, summarise or advise.
2. One entry per finding: a diagnosis (condition), a result or measurement (observation), a medicine (medication), an operation or procedure (procedure), a visit or hospital stay (encounter), an allergy (allergy), a vaccination (immunization).
3. "text" is the name only, as written, e.g. "Креатинін". The value with its unit goes in "value", e.g. "145 мкмоль/л".
4. Dates: YYYY, YYYY-MM or YYYY-MM-DD, never more precise than the document. Put the date as printed in "date_as_written".
5. "source_flagged": true only when the document itself marks the value (H, L, *, an arrow, bold). Never decide it yourself.
6. Codes (ICD-10, LOINC, ATC) only when they are printed or you are certain. Otherwise leave "code" out.
7. Leave out everything about who I am: no name, date of birth, age, address, phone, ID, insurance or record numbers, gender, blood type, height or weight. Hxtory keeps those separately.
8. For each document, add one "source" (title, document_type, document_date, facility) and point its entries to it with "source".
9. Group the entries into a few "cases", the stories they belong to: a chronic condition, an acute event, an investigation, routine care or a congenital condition. Not one case per diagnosis. An entry can be in several cases, or in none.
10. If you don't know a field, leave it out. Never guess.

Answer with the JSON only, valid against https://www.hxtory.app/hxtory-json/v1.schema.json:
{"format": "hxtory-json", "version": 1, "generator": "<your model's name>", "language": "<the documents' main language, e.g. uk>", "cases": [...], "sources": [...], "entries": [...]}

What's in the file

cases
The stories your entries belong to: a chronic condition, an acute event, an investigation, routine care or a congenital condition.
sources
The documents the entries were read from, described by title, type, date and place. Not the documents themselves.
entries
Diagnoses, results, medicines, procedures, visits, allergies and vaccinations, with names and values exactly as printed, and the cases and source each one belongs to.

Files

Good to know

  • Importing needs Hxtory Full. It's included in the TestFlight beta.
  • Imports are unlimited and don't use your pages or messages.
  • Imported entries have no original document attached for now: they show as imported.
  • Your AI can misread a document, as any AI can. That's why every entry is checked before it's saved.