Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.docex.dev/llms.txt

Use this file to discover all available pages before exploring further.

​
Run

Run vision analysis on an image or PDF. The SDK handles upload, polling, and returns a structured envelope. The CLI is useful for one-off analysis and debugging.

​
SDK

import { createDocex } from "docexdev";

const docex = createDocex({
  apiKey: process.env.DOCEX_API_KEY,
  baseUrl: process.env.DOCEX_BASE_URL,
});

const result = await docex.run({
  file: "./invoice.pdf",
  prompt: "Extract line items, totals, and vendor details",
  outputFormat: "json",     // optional, default: "json"
  schemaId: "invoice/global-accounts-payable",  // optional
  workflow: "fast",         // optional
  metadata: { source: "uploads-api" },  // optional
});

​
Parameters

ParameterTypeRequiredDescription
filestring | { fileName, data }YesLocal file path or object with fileName and data (Uint8Array/Buffer)
promptstringYesPlain-English description of the analysis task
outputFormat"json" | "text"NoOutput format. Default: "json"
schemaIdstringNoBuilt-in registry schema identifier
workflowstringNoNamed workflow policy
metadataobjectNoArbitrary metadata attached to the job

​
Response

The SDK polls automatically and returns the completed run envelope: 
{
  "status": "completed",
  "result": { ... },
  "usage": { "requests": 1, "costUsd": 0.0321, "priceUsd": 0.0642 },
  "chargedUsd": 0.0642,
  "confidence": 0.95,
  "metadata": {}
}

​
CLI

# Analyze a local file
docex run invoice.pdf --prompt "Extract line items and totals"

# Specify output format
docex run screenshot.png --prompt "is_phishing, confidence_score" --format json

# Use a built-in schema
docex run license.heic --schema-id "trade-license/uae"

# Override env vars
docex run image.png --prompt "..." --api-key dx_live_... --base-url https://api.docex.dev

​
Backward-compatible alias

docex.extract() is an alias for docex.run(): 
// These are equivalent
await docex.run({ file, prompt });
await docex.extract({ file, prompt });