Inspecting the Files
The whole point of anarchie is that the store is not a black box. After the
previous page you have a real repository on disk. Here is how to read it back
with tools you already have - no anarchie binary required.
Walk the tree
tree my-cdr
my-cdr
├── anarchie.toml
├── ehrs
│ └── 1b4e28ba-2fa1-11d2-883f-0016d3cca427
│ ├── ehr.json
│ ├── ehr_status
│ │ └── status.json
│ ├── compositions
│ │ └── 9f1c8a3e-7c2b-4e9a-bd11-2a1f6c0e4d55
│ │ └── composition.json
│ └── contributions
│ ├── 4d8e…-contrib.json
│ └── 7b2a…-contrib.json
├── index
│ └── .gitignore
└── templates
├── adverse_reaction_list.v1.opt.json
├── attribution.md
├── encounter_note.v1.opt.json
├── immunisation_list.v1.opt.json
├── index.json
├── laboratory_result_report.v1.opt.json
├── medication_list.v1.opt.json
├── problem_list.v1.opt.json
├── procedure_list.v1.opt.json
└── vital_signs_encounter.v1.opt.json
Each EHR is a directory. Each Composition is a directory holding the canonical head file. Each Contribution leaves a manifest. It is all just files.
Read a Composition with jq
jq '.name.value, .archetype_details.template_id.value' \
ehrs/*/compositions/*/composition.json
"Blood pressure"
"vital_signs_encounter.v1"
Because the files are canonical JSON, jq, ripgrep, and friends work without
any special knowledge of anarchie.
Read the history with git
The version history is the git history. From inside the deployment directory,
cd into your EHR's directory and ask git:
cd ehrs/$EHR
git log --oneline -- compositions/*/composition.json
b2c3d4e Corrected diastolic reading
a1b2c3d Admission observations
The openEHR audit metadata is right there in the commit:
git log -1 --format='%an <%ae> %cI%n%b'
Dr Ada Lovelace <ada@example.org> 2026-01-15T09:42:11Z
anarchie-contribution-id: 7b2a…
anarchie-change-type: modification
anarchie-system-id: anarchie.example.org
Reconstruct an old version with git
git show a1b2c3d:compositions/*/composition.json | jq .name.value
"Blood pressure"
This is exactly what anarchie cat <version_uid> does under the hood. There is
no hidden state: the data lives in files, the history lives in git, and both are
open formats you can audit, back up, and grep forever.
The onion model
The canonical JSON files at the centre are the source of truth. The git history, the (future) AQL index, and the (future) REST API are all derived views. Delete a derived layer and rebuild it; the patient data is untouched because it never lived there.