# `PhoenixKitAgent.Attachments.Detect`
[🔗](https://github.com/BeamLabEU/phoenix_kit_agent/blob/v0.1.0/lib/phoenix_kit_agent/attachments/detect.ex#L1)

What an uploaded file is, told by its content (spec §9: png/jpg/webp/heic,
pdf, xlsx/csv, txt/md).

The first bytes decide for everything that has a signature — a picture, a
PDF, a workbook (a zip with `xl/workbook.xml`); the name and the type the
browser claimed are never believed against them. Only a file without any
signature may be text, and then it must really be text (no NUL byte, valid
UTF-8 at the start) and its extension picks `.csv`, `.txt` or `.md`. A
file whose bytes are another known format (a GIF, an SVG, a Word document,
an executable) or a binary named `.txt` is `{:error, :unsupported_type}`.

A file named like a text file that is text in another encoding — what Excel
writes as "CSV" or "Unicode text" in many locales (Windows-1251, Windows-1252,
UTF-16) — is `{:error, :unsupported_encoding}`, so that the user is told to
save it as UTF-8 and not that the file is of a kind that is not accepted.

# `detected`

```elixir
@type detected() :: %{kind: String.t(), content_type: String.t(), ext: String.t()}
```

The kind of an attachment (`Attachment.kinds/0`), its type and the extension to store it by.

# `detect`

```elixir
@spec detect(Path.t(), String.t()) ::
  {:ok, detected()} | {:error, :unsupported_type | :unsupported_encoding}
```

Tells the kind of the file at `path`, called `filename` by the uploader.

---

*Consult [api-reference.md](api-reference.md) for complete listing*