Extract Text - Dreep API Documentation

Extracts text using OCR. With a destination folder the text is also stored as a .txt asset; without one it is returned only. See Text Extraction.

f, err := os.Open("invoice.pdf")
if err != nil {
    return err
}
defer f.Close()

result, err := client.UploadAndExtractText(ctx, f, "invoice.pdf", nil)
fmt.Println(result.Text)

Signature

func (c *Client) UploadAndExtractText(ctx context.Context, file io.Reader, filename string, o *dreep.UploadAndExtractOptions) (*dreep.OCRResult, error)

Parameter	Type	Description
`file`	`io.Reader`	Required. Image or PDF
`filename`	`string`	Required. Name of the source file
`o`	`*dreep.UploadAndExtractOptions`	Optional. Pass `nil` to return the text only

`UploadAndExtractOptions`

Field	Type	Description
`Destination`	`dreep.Destination`	Embedded. Set `Folder` or `FolderID` to save the text
`AutoCreateFolders`	`*bool`	Set `dreep.Ptr(false)` to require the path already exists

Returns

type OCRResult struct {
    Text       string
    Blocks     []map[string]any       // line and word objects with bounding boxes
    SavedAsset *dreep.SavedTextAsset  // ID, StorageKey; nil unless a folder was given
}

Blocks carries positions, for when you need where the text sits rather than just the concatenated string.

result, err := client.UploadAndExtractText(ctx, f, "receipt.jpg", &dreep.UploadAndExtractOptions{
    Destination: dreep.Destination{Folder: "receipts/2026"},
})

result.SavedAsset.ID // the stored .txt asset

Reading an asset already in Dreep

An uploaded asset doesn’t need re-uploading. ExtractText fetches it in the txt format and returns the plain text:

func (c *Client) ExtractText(ctx context.Context, mediaID string, o *dreep.TextOptions) (string, error)

text, err := client.ExtractText(ctx, asset.ID, nil)

Results are cached, so repeat reads of the same asset don’t run OCR again. For an asset in a signed folder, set SigningTTL and the SDK attaches exp and sig using your signing secret:

text, err := client.ExtractText(ctx, asset.ID, &dreep.TextOptions{SigningTTL: time.Minute})

OCR is billed per page. See Errors & Retries for what happens when the quota runs out.