Memory Banks

Fresh

Source: docs.hindsight.vectorize.io/memory-banks

Overview

Memory Banks are the foundational organizational structure in Hindsight. Each bank functions as an isolated container for memories with its own profile, settings, and data.

Bank Profile

Configure your bank's identity through:

• Name — Display identifier
• Background — Contextual purpose description
• Disposition Traits — Reasoning characteristics:
  • Skepticism — How critically the agent evaluates claims
  • Literalism — How literally vs. figuratively content is interpreted
  • Empathy — How much emotional context influences reasoning

Bank Views

View	Purpose
Profile	Configure bank identity and disposition
Recall	Debug search queries across all retrieval methods with relevance scores
Reflect	Interactive chat interface for reasoning over memories
Memories	Browse stored memories in World Facts, Experience, Observations tabs
Mental Models	Create, edit, refresh, and toggle auto-refresh for cached reflections
Documents	Upload source documents (PDF, Word, PowerPoint, Excel, images, text)
Entities	View extracted entities (people, organizations, places) with relationships

Creating a Memory Bank

Via UI

1. Access Memory Bank Selector in navigation
2. Click Create New Bank
3. Enter name, optional background and traits
4. Click Create

Via API

Python

bank = client.create_bank(
    bank_id="customer-support-agent",
    name="Customer Support Agent",
    background="Handles customer inquiries for an e-commerce platform",
    disposition={
        "skepticism": 3,
        "literalism": 2,
        "empathy": 4
    }
)

TypeScript

const bank = await client.createBank('customer-support-agent', {
  name: 'Customer Support Agent',
  background: 'Handles customer inquiries for an e-commerce platform',
  disposition: {
    skepticism: 3,
    literalism: 2,
    empathy: 4
  }
});

cURL

curl -X GET https://api.hindsight.vectorize.io/v1/default/banks \
  -H "Authorization: Bearer your-api-key"

Document Upload

Add source documents via text input or file upload.

Extraction Methods

Method	Cost	Use Case
Standard	Free	Text-based documents
Enhanced (Iris)	Per-token charge	Images, scanned PDFs, complex layouts

Track processing status via the Document Operations panel.

Best Practices

Organization:

• Create separate banks for different users, projects, or agents
• Use descriptive names identifying purpose
• Provide contextual backgrounds

Memory Quality:

• Store concrete, actionable information
• Include contextual details for retrieval
• Add temporal references for time-based searches

Performance:

• Batch operations when storing multiple memories
• Use filters to narrow search scope
• Monitor Usage Analytics for cost optimization