Skip to main content

​
Overview

The Vision Actions module provides AI-powered image analysis for industrial parts and equipment using Google’s Gemini Vision model. It identifies, classifies, and evaluates the physical condition of items in images.

​
analyzePartImage

Analyzes an image of an industrial part or equipment using Gemini Vision to extract structured information including type, condition, brand, model, and recommendations. 
export async function analyzePartImage(
  formData: FormData
): Promise<{ 
  result: PartAnalysisResult | null; 
  success: boolean; 
  error?: string 
}>

​
Parameters

​
formData
FormData
required
Form data containing the image file and optional prompt
​
formData.file
File
required
Image file to analyze (JPG, PNG, etc.)
​
formData.prompt
string
Custom analysis prompt (defaults to INVENTORY_PROMPT)

​
Response

​
result
PartAnalysisResult | null
Structured analysis result from the AI model
​
success
boolean
Whether the analysis succeeded
​
error
string
Error message if analysis failed

​
PartAnalysisResult Schema

​
tipo_articulo
'mobiliario' | 'equipo'
General classification of the item
​
codigo
string
Visible identification code on the part
​
descripcion
string
Detailed description of the part or equipment
​
marca
string
Manufacturer brand if visible
​
modelo
string
Equipment model if visible
​
cantidad_detectada
number
Number of parts of this type detected in the image
​
estado_fisico
'nuevo' | 'usado' | 'dañado' | 'requiere_inspeccion'
Visual condition of the part
​
recomendacion
string
Brief recommendation for handling or maintenance
​
nivel_confianza
'alta' | 'media' | 'baja'
AI’s confidence level in the identification

​
Features

  • Size Validation: Enforces maximum image size (configured in MAX_IMAGE_SIZE_MB)
  • Structured Output: Uses Zod schema validation for consistent results
  • Low Temperature: Uses temperature 0.1 for strict schema adherence
  • Detailed Analysis: Extracts multiple attributes including condition and recommendations

​
Example

const formData = new FormData();
formData.append('file', imageFile);
formData.append('prompt', 'Analiza esta pieza y determina si requiere mantenimiento');

const result = await analyzePartImage(formData);

if (result.success && result.result) {
  const part = result.result;
  console.log(`Tipo: ${part.tipo_articulo}`);
  console.log(`Descripción: ${part.descripcion}`);
  console.log(`Estado: ${part.estado_fisico}`);
  console.log(`Cantidad: ${part.cantidad_detectada}`);
  console.log(`Confianza: ${part.nivel_confianza}`);
  
  if (part.marca) console.log(`Marca: ${part.marca}`);
  if (part.modelo) console.log(`Modelo: ${part.modelo}`);
  if (part.codigo) console.log(`Código: ${part.codigo}`);
  
  console.log(`Recomendación: ${part.recomendacion}`);
} else {
  console.error("Error:", result.error);
}

​
Example Response

{
  "result": {
    "tipo_articulo": "equipo",
    "codigo": "EQ-2024-1523",
    "descripcion": "Compresor de aire industrial marca Atlas Copco",
    "marca": "Atlas Copco",
    "modelo": "GA 55",
    "cantidad_detectada": 1,
    "estado_fisico": "usado",
    "recomendacion": "Verificar filtros y nivel de aceite. Programar mantenimiento preventivo",
    "nivel_confianza": "alta"
  },
  "success": true
}

​
Configuration

​
Environment Variables

  • GOOGLE_GENERATIVE_AI_API_KEY: Required for vision analysis
  • MAX_IMAGE_SIZE_MB: Maximum image file size (defined in config/limits)

​
Constants

  • INVENTORY_PROMPT: Default prompt for inventory analysis (defined in config)

​
Model Configuration

  • Model: gemini-2.5-flash
  • Temperature: 0.1 (low temperature for consistent structured output)
  • Schema: Zod validation for type-safe results

​
Supported Image Formats

The function accepts standard image formats including:
  • JPEG/JPG
  • PNG
  • WebP
  • GIF

​
Size Limits

Images are validated against MAX_IMAGE_SIZE_MB before processing. Large images are rejected with a descriptive error message.

​
Use Cases

  1. Inventory Management: Automatically catalog parts with descriptions and codes
  2. Quality Inspection: Assess physical condition of equipment
  3. Maintenance Planning: Get AI-powered recommendations for part handling
  4. Asset Tracking: Detect quantities and identify brands/models
  5. Documentation: Generate structured records from visual inspections

Build docs developers (and LLMs) love

Get started for freeTalk to us