Skip to main content
The problem: You have 5,000 documents from discovery. Finding the relevant ones takes weeks of manual review. The solution: Upload to a Vault. We OCR, chunk, and index everything. Search by meaning in seconds.

​
1. Create a vault

Vaults are secure containers for your users’ documents. Each vault gets automatic OCR, chunking, and vector indexing. 
import Casedev from 'casedev';

const client = new Casedev({ apiKey: process.env.CASEDEV_API_KEY });

// Create a vault for your user's documents
const vault = await client.vault.create({
  name: 'Case Documents - User 12345',
  description: 'Discovery documents for case review'
});

console.log(`Vault created: ${vault.id}`);

​
2. Upload documents

Handle file uploads from your users and trigger automatic processing: 
import fs from 'fs';

async function uploadDocument(vaultId: string, filePath: string) {
  // Get presigned upload URL
  const upload = await client.vault.upload(vaultId, {
    filename: filePath.split('/').pop()!,
    contentType: 'application/pdf'
  });

  // Upload file to S3
  const file = fs.readFileSync(filePath);
  await fetch(upload.uploadUrl, {
    method: 'PUT',
    headers: { 'Content-Type': 'application/pdf' },
    body: file
  });

  // Trigger OCR + embedding pipeline
  await client.vault.ingest(vaultId, upload.objectId);

  return upload.objectId;
}

// Process uploads from your user
const files = fs.readdirSync('./uploads');
for (const file of files) {
  await uploadDocument(vault.id, `./uploads/${file}`);
  console.log(`Processed: ${file}`);
}

​
3. Search by meaning

Enable your users to search by meaning, not just keywords: 
// In your search endpoint or UI handler
const results = await client.vault.search(vault.id, {
  query: userQuery, // e.g., "communications about equipment failure"
  method: 'hybrid',
  topK: 10
});

// Return results to your user
for (const chunk of results.chunks) {
  console.log(`📄 ${chunk.filename} (score: ${chunk.hybridScore.toFixed(2)})`);
  console.log(`"${chunk.text.substring(0, 200)}..."`);
}

​
4. Summarize findings

Enhance results with AI-generated summaries for your users: 
const context = results.chunks.map(c => c.text).join('\n\n---\n\n');

const summary = await client.llm.v1.chat.createCompletion({
  model: 'anthropic/claude-sonnet-4.5',
  messages: [
    {
      role: 'system',
      content: 'Summarize these search results concisely. Highlight the most relevant findings.'
    },
    {
      role: 'user',
      content: `User searched for: "${userQuery}"\n\nResults:\n\n${context}`
    }
  ],
  max_tokens: 500
});

// Return summary along with search results
console.log(summary.choices[0].message.content);
Time saved: What used to take weeks of manual review now takes minutes. The AI finds relevant passages even when documents use different terminology.