Building on Google Cloud

Vertex AI — Gemini Document Analysis

// Vertex AI — Gemini document analysis (Node.js / Edge Function)
// Terra Core document intelligence pipeline — Google Cloud Partner build
import { VertexAI } from '@google-cloud/vertexai'

const vertex = new VertexAI({
  project: process.env.GCP_PROJECT_ID,
  location: 'europe-west2',              // London — EU data residency
})

const model = vertex.getGenerativeModel({
  model: 'gemini-1.5-pro',
  generationConfig: {
    maxOutputTokens: 2048,
    temperature: 0.1,                    // Low temp for factual extraction
  },
})

async function analyseDocument(documentText, extractionSchema) {
  const prompt = 'Extract the following fields from this legal document. ' +
    'Return only valid JSON matching the schema. Do not hallucinate. ' +
    'Schema: ' + JSON.stringify(extractionSchema) + ' ' +
    'Document: ' + documentText

  const result = await model.generateContent(prompt)
  const text = result.response.candidates?.[0]?.content?.parts?.[0]?.text
  if (!text) throw new Error('Vertex AI returned no content')

  return JSON.parse(text)
}

Cloud Run — Production Deployment

# Cloud Run deployment — Two Bit Digital production pattern
# Containerised Next.js → Cloud Run (London, europe-west2)

# Build and push image to Artifact Registry
gcloud builds submit \
  --tag europe-west2-docker.pkg.dev/$PROJECT_ID/apps/twobitdigital:prod \
  --region europe-west2

# Deploy to Cloud Run with auto-scaling config
gcloud run deploy twobitdigital-prod \
  --image europe-west2-docker.pkg.dev/$PROJECT_ID/apps/twobitdigital:prod \
  --platform managed \
  --region europe-west2 \
  --allow-unauthenticated \
  --min-instances 0 \          # Scales to zero off-peak
  --max-instances 10 \         # Burst to 10 under load
  --memory 512Mi \
  --cpu 1 \
  --concurrency 80 \
  --set-env-vars NODE_ENV=production \
  --service-account app-sa@$PROJECT_ID.iam.gserviceaccount.com