> ## Documentation Index
> Fetch the complete documentation index at: https://docs.memic.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Ingesting documents

> Upload files, batch imports, metadata, and file processing lifecycle.

Memic's ingestion pipeline takes a file, chunks it, generates embeddings,
and indexes it for semantic search. This guide walks through the flow and
common patterns.

## The upload flow

Every file upload is a three-step process:

1. **Init** — call `POST /files/init` with the filename and size. Memic
   returns a presigned upload URL.
2. **Upload** — `PUT` the file contents directly to the presigned URL. This
   bypasses the Memic API entirely so large files don't hit request limits.
3. **Confirm** — call `POST /files/{id}/confirm` to signal upload is complete.
   Memic queues the file for processing.

The Python SDK wraps all three steps into a single `client.files.upload(...)`
call, but if you're calling the API directly (or uploading from a browser),
you'll do each step yourself. See the [cURL quickstart](/quickstart-curl) for
the raw HTTP flow.

## Processing lifecycle

After confirm, a file goes through these statuses:

```
pending → processing → ready
                    ↘ failed
```

* **`pending`** — queued, not yet picked up by a worker
* **`processing`** — being chunked, embedded, and indexed
* **`ready`** — searchable
* **`failed`** — processing failed (check `/files/{id}/status` for the reason)

Typical times: a PDF under 10MB is usually `ready` within 30 seconds. Larger
files or spreadsheets can take longer.

## Polling vs waiting

For interactive flows, poll `/files/{id}/status` every 2–5 seconds until
`ready`. The endpoint has a higher rate limit than search/chat specifically
for this use case.

For batch uploads where you don't need immediate feedback, upload everything
and check status in a single pass at the end.

## Supported file types

* **PDF** — with full text extraction and table detection
* **Plain text** (`.txt`, `.md`)
* **Microsoft Office** (`.docx`, `.pptx`, `.xlsx`)
* **Web formats** (`.html`)
* **Image formats** (`.png`, `.jpg`) — text extracted via OCR

Files outside this list are rejected at `init` time with `invalid_file_type`.

## Maximum file size

The current limit is **100 MB per file**. For larger files, split them into
logical sections (e.g. chapters) and upload as separate files.

## Batch uploads

The Python SDK provides a batch helper that uploads multiple files in
parallel:

```python theme={null}
results = client.files.upload_batch([
    "./doc1.pdf",
    "./doc2.pdf",
    "./doc3.pdf",
])
```

Under the hood this calls `init` + `PUT` + `confirm` for each file with
controlled concurrency.

## Replacing a file

To update the content of a file, upload a new version with the same
filename. Memic treats filenames as human labels, not unique keys — the new
upload creates a new `file_id`.

To replace and remove the old version: upload the new file, delete the old
one with `DELETE /files/{old_id}`.

## Related

<CardGroup cols={2}>
  <Card title="Search guide" icon="magnifying-glass" href="/guides/search">
    Once files are ready, query them.
  </Card>

  <Card title="API reference" icon="code" href="/api-reference/introduction">
    Full file endpoint docs.
  </Card>
</CardGroup>