# Bulk import

> Backfill history you already have.

Send us the conversations you already have, so a score has history behind it from the first day.

## The request

A file upload, not a JSON body. Send it as a form, with an ingest key.

```http
POST /ingest/v1/import
Authorization: Bearer ak_live_...
Content-Type: multipart/form-data
```

| | |
|---|---|
| file | The export itself. One JSON document, or one JSON record per line. |
| adapter | Which contract to read the file with. Use generic for a file you built yourself. |
| client_id | Optional. Which client the history belongs to. |

A key limited to one client may name only that client. Naming another gets 403. A key with no limit may name any client of the workspace, and an id we cannot find gets 404.

## How it differs from sending events

- The file is stored first and read by a background job, so the answer comes back before any of it is parsed.
- The ceiling is 200 MiB per file, against 8 MiB for a live batch.
- A line we cannot read is skipped and counted, so one bad record does not cost you the export. A file where no line parses at all fails outright, because that is the wrong file rather than a partial one.

## What we send back

202, with an import id and the key the file was stored under.

```json
{ "status": "accepted", "import_id": "...", "object_key": "..." }
```

| | |
|---|---|
| 202 | Stored and queued. |
| 401 | The key is missing, revoked, or does not carry the ingest scope. |
| 403 | The key is limited to a client other than the one named. |
| 404 | The client named is not one of this workspace's. |
| 413 | The file is above 200 MiB. Split it. |
| 422 | A form field is missing, or the client id is not an id. |

---

Source: https://evidova.com/docs/api/import