API referenceDatasets
Download a RAG snapshot
GET /v1/datasets/{dataset}/snapshots/{version}
GET https://api.healthcare.dev/v1/datasets/{dataset}/snapshots/{version}Team plan (403 plan-upgrade-required below Team); never counts as a billable call. One snapshot's manifest plus a signed download URL for each file, valid for 15 minutes (expires_at) and needing no API key. latest is the newest snapshot. Check each file against the manifest's sha256 after downloading.
Requires an API key: Authorization: Bearer hdev_live_.... See Authentication.
Path parameters
| Name | Type | Required | Description | Example |
|---|---|---|---|---|
dataset | string | yes | Dataset id: one of hcpcs, icd10cm, icd10pcs, ndc, rxnorm | icd10cm |
version | string | yes | A release (data_version) from the list, or latest | latest |
Examples
curl
curl "https://api.healthcare.dev/v1/datasets/icd10cm/snapshots/latest" \
-H "Authorization: Bearer $HDEV_API_KEY"TypeScript
const res = await fetch("https://api.healthcare.dev/v1/datasets/icd10cm/snapshots/latest", {
headers: { Authorization: `Bearer ${process.env.HDEV_API_KEY}` },
});
const { data, meta } = await res.json();Python
import os
import requests
res = requests.get("https://api.healthcare.dev/v1/datasets/icd10cm/snapshots/latest", headers={"Authorization": f"Bearer {os.environ['HDEV_API_KEY']}"})
body = res.json()Try it
Response
The snapshot with download URLs
{
"data": {
"format": "hdev-rag-snapshot/1",
"dataset": "string",
"name": "string",
"version": "string",
"effective_date": "2026-10-01",
"published_at": "2026-10-24T12:00:00Z",
"created_at": "2026-10-24T12:00:00Z",
"records": 0,
"chunks": 0,
"embedding": {
"id": "string",
"dims": 1,
"dtype": "float32",
"byte_order": "little-endian",
"normalized": true,
"metric": "cosine",
"file": "string",
"rows": 0
},
"files": [
{
"path": "string",
"bytes": 0,
"sha256": "string",
"content_type": "string",
"gzip": true
}
],
"source": {
"source": "string",
"source_url": "string",
"source_sha256": "string",
"pipeline_version": "string"
},
"license": {
"license": "string",
"terms": {
"display": "string",
"attribution": "string",
"url": "string"
},
"file": "string"
},
"latest": true,
"downloads": [
{
"path": "string",
"url": "string"
}
],
"expires_at": "2026-10-24T12:00:00Z"
},
"meta": {
"source": "cms-icd10",
"data_version": "2026.1",
"retrieved_at": "2026-10-24T12:00:00Z",
"license": "Public domain (CMS). Reference data, not medical advice."
},
"links": {
"self": "/v1/datasets/icd10cm/snapshots/latest",
"next": null
}
}Fields
| Field | Type | Always present | Description |
|---|---|---|---|
format | "hdev-rag-snapshot/1" | yes | |
dataset | string | yes | Dataset id, e.g. icd10cm |
name | string | yes | Dataset name |
version | string | yes | The release (data_version) |
effective_date | string | null (date) | yes | When the release takes effect |
published_at | string (date-time) | yes | When we published the release |
created_at | string (date-time) | yes | When the snapshot was built |
records | integer | yes | Active records in the release (retired records are left out) |
chunks | integer | yes | Lines in the chunks file: one per active record with text |
embedding | SnapshotEmbedding | yes | |
files | SnapshotFile[] | yes | Every file except the manifest itself, with its size and SHA-256 |
source | object | yes | Where the release came from, as in each response's meta |
license | object | yes | What you may do with the data; the same text is in LICENSE.txt |
latest | boolean | yes | True for the newest snapshot of the dataset |
downloads | SnapshotDownload[] | yes | One signed URL per file, the manifest included |
expires_at | string (date-time) | yes | When the download URLs stop working |
Errors
Errors are application/problem+json (RFC 9457). Branch on type, not on title. See Errors.
| Status | type | When |
|---|---|---|
| 400 | invalid-request | A parameter or the body failed validation; errors lists each field. |
| 401 | unauthorized | Missing, malformed or revoked API key. |
| 403 | plan-upgrade-required | The plan doesn't include this feature; required_plan names the plan that does. |
| 404 | not-found | No record with that id (retired records are still returned). |
| 429 | rate-limited / quota-exceeded | Too many requests per second (retry after Retry-After), or the monthly quota is used up. |
| 503 | dataset-unavailable / unavailable | The dataset has no published version yet, or the service is degraded. |