Skip to main content

​
What is CaseMind?

CaseMind is Victoria AI’s persistent memory system that stores, organizes, and recalls all information about your cases. Unlike traditional chatbots that forget context between sessions, CaseMind remembers everything across all conversations, documents, and interactions.

​
Key Features

289+ Fact Types

Automatically extracts and categorizes facts into 289+ structured types (parties, assets, dates, etc.)

Intelligent Deduplication

3-tier deduplication ensures no duplicate facts, even when information appears multiple times

Temporal Tracking

Tracks when facts were learned, updated, and from which source

Confidence Scoring

Each fact has a confidence score based on source quality and consistency

​
How It Works

1

Information Extraction

When you interact with Victoria or upload documents, CaseMind automatically extracts relevant facts:
  • Party information (names, ages, occupations)
  • Financial data (assets, income, debts)
  • Dates and deadlines
  • Case events and history
  • Legal issues and claims
2

Deduplication

Before storing, CaseMind checks for duplicates using a 3-tier system:
  1. Exact match: Identical facts are merged
  2. Semantic similarity: Similar facts are reconciled
  3. Temporal ordering: Newer information updates older facts
3

Storage & Indexing

Facts are stored in PostgreSQL with:
  • Full-text search indexing
  • Vector embeddings for semantic search
  • Relationships between facts
  • Source attribution
4

Retrieval

When Victoria needs context, CaseMind:
  • Searches relevant facts by keyword or semantic meaning
  • Ranks by confidence and recency
  • Returns contextual information to the AI agent

​
Extractable Fact Types

CaseMind can extract 289+ different fact types, organized into categories:

​
Party Information

  • Full names, aliases, maiden names
  • Ages, birthdates
  • Occupations, employers
  • Contact information
  • Residency and domicile

​
Financial Facts

  • Assets (real estate, vehicles, accounts)
  • Income sources and amounts
  • Debts and liabilities
  • Business interests
  • Retirement accounts

​
Case Timeline

  • Marriage date
  • Separation date
  • Filing date
  • Important events
  • Court dates and deadlines

​
Children

  • Names and ages
  • Custody arrangements
  • Support orders
  • School and medical info
  • Special needs
  • Contested issues
  • Settlement positions
  • Court orders
  • Pending motions
  • Discovery status
You can view all extracted facts for a case in the Case Details panel in Case Workspace.

​
Deduplication System

CaseMind’s 3-tier deduplication prevents information overload:

​
Tier 1: Exact Match

Input: "John Smith is 45 years old"
Existing: "John Smith is 45 years old"
Result: No new fact created (exact duplicate)

​
Tier 2: Semantic Similarity

Input: "Husband earns $150,000 annually"
Existing: "John Smith's annual income is $150,000"
Result: Facts merged, confidence score increased

​
Tier 3: Temporal Updates

Input: "John now earns $160,000/year" (dated today)
Existing: "John earns $150,000/year" (dated 6 months ago)
Result: New fact stored, old fact marked as historical

​
Source Attribution

Every fact in CaseMind tracks its source:
Source TypeExamplesConfidence Weight
Court DocumentsFiled motions, ordersHigh (90-100%)
Client DocumentsFinancial affidavits, pay stubsHigh (85-95%)
AI ConversationsVictoria chat sessionsMedium (70-85%)
Manual EntryUser-entered dataHigh (90-100%)
Third-PartyPublic recordsMedium-High (80-90%)
Facts from AI conversations have lower confidence scores and should be verified against source documents before relying on them in court.

​
Memory Retrieval

When Victoria needs information, CaseMind uses hybrid search:
  1. Keyword Search: Full-text search on fact content
  2. Semantic Search: Vector similarity for conceptual matches
  3. Relationship Traversal: Following connections between facts
  4. Temporal Filtering: Prioritizing recent or historically relevant facts

​
Example Query Flow

User asks: "What is the husband's income?"

CaseMind retrieves:
1. "John Smith's current income: $160,000/year" (source: latest pay stub)
2. "Previous income: $150,000/year" (source: financial affidavit, 6mo ago)
3. "Income includes $10,000 bonus" (source: Victoria conversation)

Victoria responds with current income and notes the increase.

​
Privacy & Security

CaseMind memory is protected by:
  • Firm Isolation: Each firm’s CaseMind data is completely separate
  • Case Isolation: Facts are scoped to specific cases
  • Encryption: All data encrypted at rest (AES-256) and in transit (TLS 1.3)
  • Access Control: Only authorized users can access case facts
  • Audit Logs: All fact access and modifications are logged

​
Managing CaseMind

​
Viewing Extracted Facts

In Case Workspace, click Case Details → Extracted Facts to see:
  • All facts organized by category
  • Source attribution for each fact
  • Confidence scores
  • Last updated timestamps

​
Editing Facts

You can manually edit or remove incorrect facts:
  1. Navigate to Case Details → Extracted Facts
  2. Click on any fact to edit or delete
  3. Changes are logged in the audit trail

​
Exporting Facts

Export all case facts to:
  • JSON: For data analysis or backup
  • PDF: For court filing or client review
  • CSV: For spreadsheet analysis

​
Best Practices

The more documents CaseMind processes, the richer the memory:
  • Upload financial affidavits first
  • Add court orders and pleadings
  • Include correspondence when relevant
Periodically review extracted facts for accuracy:
  • Check financial figures against source documents
  • Verify dates and deadlines
  • Correct any AI extraction errors
Help CaseMind by using full names consistently:
  • “John Smith” vs. “the husband”
  • “Acme Corp” vs. “his business”
  • “2023 Toyota Camry” vs. “the car”
When information changes over time, mention dates:
  • “As of January 2025, John earns…”
  • “Before the separation, they owned…”
  • “The most recent appraisal shows…”

​
Memory Optimization

CaseMind automatically optimizes memory usage:
  • Fact Pruning: Very old, low-confidence facts may be archived
  • Consolidation: Related facts are linked and summarized
  • Prioritization: Frequently accessed facts get higher retrieval priority
CaseMind memory is unlimited—you don’t need to worry about storage limits. The system automatically manages memory efficiently.

​
Integration with Victoria AI

Every Victoria AI agent has full access to CaseMind:
  • Co-Counsel: Uses facts for legal strategy and motion drafting
  • Financial Analyst: Retrieves financial data for calculations
  • Discovery Manager: References known facts to identify gaps
  • Case Manager: Tracks deadlines and case timeline
  • Client Liaison: Recalls case details for client communications

​
Next Steps

Explore LexVault

Learn how documents feed into CaseMind

Victoria AI Agents

See how agents use CaseMind memory