mirror of
https://github.com/bulwarkmail/webmail.git
synced 2026-11-04 08:15:41 +00:00
Linus Rath
bda99e1df8
api.email.getAttachmentText reads PDF (pdf.js), Word, Excel, PowerPoint and OpenDocument files (own reader over fflate and DOMParser, unpacking capped at 50 MB), text, CSV, HTML, calendar and address files, and answers with a status for what it can't read. The extraction lives in lib/attachment-text so Files can use it too. The sandbox runtime now copies every flag the host sends into api.host.features; before, it kept only the three core flags, so the assistant's flags never reached a plugin. The host bridge assertions in plugin-host-features.test.ts compare against the core flags plus the assistant's (withAssistant), since the init payload and host-change now carry attachmentText too.
55 lines
2.1 KiB
TypeScript
55 lines
2.1 KiB
TypeScript
// Shapes and limits of the text extraction for attachments and files.
|
|
// Pure: no imports from the app, so the plugin host and Files can share it.
|
|
|
|
/** What kind of file the text came from. */
|
|
export type TextFormat =
|
|
| 'text' | 'csv' | 'html' | 'ics' | 'vcf' | 'pdf'
|
|
| 'docx' | 'xlsx' | 'pptx' | 'odt' | 'ods' | 'odp' | 'other';
|
|
|
|
/**
|
|
* ok: `text` holds the text. too_big: over the size limit, not read.
|
|
* unsupported: a format without a reader (images, legacy .doc/.xls/.ppt,
|
|
* archives). no_text: nothing to read, e.g. a scanned PDF. encrypted:
|
|
* protected by a password. failed: the file is damaged or reading broke.
|
|
*/
|
|
export type TextStatus = 'ok' | 'too_big' | 'unsupported' | 'no_text' | 'encrypted' | 'failed';
|
|
|
|
export interface ExtractedText {
|
|
format: TextFormat;
|
|
status: TextStatus;
|
|
/** The text, '' unless status is ok. */
|
|
text: string;
|
|
/** True when the text was cut at maxChars, or a PDF or sheet had more than was read. */
|
|
truncated: boolean;
|
|
/** PDFs: the number of pages in the file. */
|
|
pages?: number;
|
|
/** Spreadsheets: the names of the sheets read. */
|
|
sheets?: string[];
|
|
}
|
|
|
|
export interface ExtractOptions {
|
|
/** File name; its extension picks the format first. */
|
|
name?: string;
|
|
/** Media type, used when the name says nothing. */
|
|
type?: string;
|
|
/** Longest text returned (default DEFAULT_TEXT_CHARS, at most MAX_TEXT_CHARS). */
|
|
maxChars?: number;
|
|
}
|
|
|
|
/** Text returned when the caller doesn't ask for a length. */
|
|
export const DEFAULT_TEXT_CHARS = 20_000;
|
|
/** Longest text one read returns. */
|
|
export const MAX_TEXT_CHARS = 100_000;
|
|
/** Largest file read for its text. */
|
|
export const MAX_TEXT_SOURCE_BYTES = 20 * 1024 * 1024;
|
|
/** Pages of a PDF read at most. */
|
|
export const MAX_PDF_PAGES = 50;
|
|
/** Fewer characters per page than this on average: the PDF has no text layer. */
|
|
export const MIN_PDF_CHARS_PER_PAGE = 20;
|
|
/** Sheets of a spreadsheet read at most. */
|
|
export const MAX_SHEETS = 5;
|
|
/** Rows per sheet read at most. */
|
|
export const MAX_SHEET_ROWS = 2_000;
|
|
/** Bytes an Office file may unpack to, all entries read together (zip bombs). */
|
|
export const MAX_INFLATED_BYTES = 50 * 1024 * 1024;
|