SpeedyF

Documents are hard. Especially in serverless environments. SpeedyF ("sPDF") is a one-click deployable solution for file manipulation, with support for using object-storage URLs as input.

Contrary to the name, SpeedyF is not just for PDFs, but can be used to manipulate a variety of files. The details for that are below.

Supported Document Formats

Format	Extension	MIME Type
PDF	.pdf	application/pdf
Word	.docx	application/vnd.openxmlformats-officedocument.wordprocessingml.document
PowerPoint	.pptx	application/vnd.openxmlformats-officedocument.presentationml.presentation
EPUB	.epub	application/epub+zip
RTF	.rtf	application/rtf
OpenDocument Text	.odt	application/vnd.oasis.opendocument.text
XML DocBook / JATS XML / OPML	.xml or .opml	application/xml
FictionBook	.fb2	application/x-fictionbook+xml
Troff man pages	.1, .man	application/x-troff-man
Jupyter notebooks	.ipynb	application/x-ipynb+json

Supported Image Formats

Format	Extension	MIME Type
JPEG	.jpg, .jpeg	image/jpeg
PNG	.png	image/png
AVIF	.avif	image/avif
TIFF	.tiff	image/tiff
GIF	.gif	image/gif
HEIC / HEIF	.heic, .heif	image/heic, image/heif
BMP	.bmp	image/bmp
WebP	.webp	image/webp

Development and Deployment

To develop locally:

Fill the environment variables:

cp .env.example .env.local 
# or 
vc env pull

A MISTRAL_API_KEY is required for the markdown endpoint. You can make an API account and keys here.

bun install
vc dev

open http://localhost:3000

To build locally:

npm install
vc build

To deploy:

npm install
vc deploy

API Documentation

Authentication

All API endpoints require bearer token authentication. The token is configured via the INTERNAL_API_KEY environment variable.

Include the token in the Authorization header:

curl -X GET "http://localhost:3000/truncate?url=..." \
  -H "Authorization: Bearer YOUR_API_KEY"

Example (TypeScript):

const response = await fetch('http://localhost:3000/truncate', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'Authorization': `Bearer ${process.env.INTERNAL_API_KEY}`
  },
  body: JSON.stringify({ url: 'https://example.com/document.pdf' })
})

Requests without a valid bearer token will receive a 401 Unauthorized response.

---

File Truncation

POST /truncate

Truncate files to fit within specified page count and file size limits. For PDFs, uses binary search optimization. For DOCX files, truncates based on paragraph count with a configurable heuristic for pages-to-paragraphs conversion.

Accepts both JSON with URL and multipart file upload:

Option 1: JSON with URL

Request:

curl -X POST http://localhost:3000/truncate \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/document.pdf",
    "maxPages": 1000,
    "maxSizeBytes": 52428800
  }'

Request Body:

Field	Type	Required	Default	Description
url	string	Yes	-	URL of the file to truncate
maxPages	number	No	1000	Maximum number of pages
maxSizeBytes	number	No	52428800 (50MB)	Maximum file size in bytes
paragraphsPerPage	number	No	15	(DOCX only) Paragraphs per page heuristic for estimating page count

Option 2: Multipart File Upload

Request:

curl -X POST http://localhost:3000/truncate \
  -F "file=@/path/to/document.pdf" \
  -F "maxPages=1000" \
  -F "maxSizeBytes=52428800"

Form Fields:

Field	Type	Required	Default	Description
file	File	Yes	-	File to truncate
maxPages	string	No	"1000"	Maximum number of pages
maxSizeBytes	string	No	"52428800"	Maximum file size in bytes
paragraphsPerPage	string	No	"15"	(DOCX only) Paragraphs per page heuristic for estimating page count

Response:

• Content-Type: Matches input file MIME type
• Body: Binary file data (truncated)
• Headers:
  • X-Was-Truncated: "true" or "false"
  • X-Original-Size: Original file size in bytes
  • X-Final-Size: Final file size in bytes

Example (TypeScript):

// With URL
const response = await fetch('http://localhost:3000/truncate', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    url: 'https://example.com/document.pdf',
    maxPages: 500
  })
})
const fileBlob = await response.blob()

// With File Upload
const formData = new FormData()
formData.append('file', Bun.file('/path/to/document.docx'))
formData.append('maxPages', '10')
formData.append('paragraphsPerPage', '20')  // Adjust if document has dense formatting

const response = await fetch('http://localhost:3000/truncate', {
  method: 'POST',
  body: formData
})
const docxBlob = await response.blob()

---

GET /truncate

Truncate a file from a URL using query parameters.

Request:

curl -X GET "http://localhost:3000/truncate?url=https://example.com/document.pdf&maxPages=1000&maxSizeBytes=52428800"

Query Parameters:

Field	Type	Required	Default	Description
url	string	Yes	-	URL of the file to truncate
maxPages	number	No	1000	Maximum number of pages
maxSizeBytes	number	No	52428800 (50MB)	Maximum file size in bytes
paragraphsPerPage	number	No	15	(DOCX only) Paragraphs per page heuristic for estimating page count

Response: Same as POST /truncate above.

Example (TypeScript):

const url = 'https://example.com/document.pdf'
const response = await fetch(`http://localhost:3000/truncate?url=${encodeURIComponent(url)}&maxPages=500`)
const fileBlob = await response.blob()

// For DOCX with custom paragraphs per page
const docxUrl = 'https://example.com/document.docx'
const response2 = await fetch(`http://localhost:3000/truncate?url=${encodeURIComponent(docxUrl)}&maxPages=10&paragraphsPerPage=20`)
const docxBlob = await response2.blob()

---

PDF to Markdown (OCR)

Convert PDFs to markdown using Mistral's OCR API. PDFs are automatically truncated to 999 pages or 49MB before processing.

POST /markdown

Convert a PDF to markdown using binary search optimization.

Accepts both JSON with URL and multipart file upload:

Option 1: JSON with URL

Request:

curl -X POST http://localhost:3000/markdown \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/document.pdf",
    "maxPages": 999,
    "maxSizeBytes": 51380224
  }'

Request Body:

Field	Type	Required	Default	Description
url	string	Yes	-	URL of the PDF to convert
maxPages	number	No	999	Maximum pages before OCR
maxSizeBytes	number	No	51380224 (49MB)	Maximum size before OCR

Option 2: Multipart File Upload

Request:

curl -X POST http://localhost:3000/markdown \
  -F "file=@/path/to/document.pdf" \
  -F "maxPages=999" \
  -F "maxSizeBytes=51380224"

Form Fields:

Field	Type	Required	Default	Description
file	File	Yes	-	PDF file to convert
maxPages	string	No	"999"	Maximum pages before OCR
maxSizeBytes	string	No	"51380224"	Maximum size before OCR

Response Fields:

Field	Type	Description
markdown	string	Extracted markdown content with pages separated by ---
remainingFields	object	The remaining fields from the Mistral OCR Response, which are detailed here.

Example (TypeScript):

// With URL
const response = await fetch('http://localhost:3000/markdown', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    url: 'https://example.com/document.pdf',
    maxPages: 500
  })
})
const { markdown, pages, usage } = await response.json()

// With File Upload
const formData = new FormData()
formData.append('file', Bun.file('/path/to/document.pdf'))
formData.append('maxPages', '500')

const response = await fetch('http://localhost:3000/markdown', {
  method: 'POST',
  body: formData
})
const { markdown, ...remainingFields } = await response.json()

---

GET /markdown

Convert a PDF from a URL to markdown using query parameters.

Request:

curl -X GET "http://localhost:3000/markdown?url=https://example.com/document.pdf&maxPages=999&maxSizeBytes=51380224"

Query Parameters:

Field	Type	Required	Default	Description
url	string	Yes	-	URL of the PDF to convert
maxPages	number	No	999	Maximum pages before OCR
maxSizeBytes	number	No	51380224 (49MB)	Maximum size before OCR

Response: Same as POST /markdown above.

Example (TypeScript):

const url = 'https://example.com/document.pdf'
const response = await fetch(`http://localhost:3000/markdown?url=${encodeURIComponent(url)}&maxPages=500`)
const { markdown, ...remainingFields } = await response.json()
console.log(markdown)