Introducing Profiles: Structured User Data for AI Applications

The Problem: Memories vs. Structured Data

Mnexium's memory system is incredibly powerful for storing free-form facts about users: "User prefers dark mode", "User is learning Rust", "User has a dog named Max". But what about structured data that every application needs?

Things like a user's name, email, timezone, or preferred language don't fit neatly into the free-form memory model. You need to query them by key, update them reliably, and ensure only one value exists at a time.

Why Businesses Need Structured User Data

In today's AI-first landscape, businesses are racing to deliver hyper-personalized experiences. But personalization at scale requires more than just remembering random facts — it requires structured, queryable, reliable user data. Here's why:

The bottom line: free-form memories are perfect for capturing the richness of human conversation — preferences, stories, context. But structured profiles are essential for the operational data that powers business processes, integrations, and compliance.

That's where Profiles come in.

What Are Profiles?

Profiles are business-defined schemas that capture the structured information your organization needs about each user. Think of it as your company's custom data model for user attributes — tailored to your specific industry, use case, and integration requirements.

Unlike free-form memories that capture the richness of conversation ("User mentioned they love hiking on weekends"), profiles capture the operational data your business runs on: the fields your CRM expects, the attributes your support team needs, the segments your marketing relies on.

Your Schema, Your Rules

Every business is different. A B2B SaaS company needs company_name,job_title, andteam_size. A healthcare app needspreferred_pharmacy andinsurance_provider. An e-commerce platform needsshipping_preference andsize.

With Mnexium Profiles, you define exactly what fields matter to your business. The AI then knows what to look for in conversations and extracts those specific attributes automatically.

How Profiles Work

• Schema-Driven Extraction. Define your schema once, and the LLM knows exactly what to extract. When a user mentions "I work at Acme Corp as a product manager", Mnexium automatically populates company_name andjob_title — because those fields are in your schema.
• Automatic Updates. When information changes, the new value supersedes the old one. "Actually, I just switched to the Enterprise plan" automatically updatessubscription_tier. No duplicate data, no conflicts — just the current truth.
• Confidence & Provenance. Every field tracks where it came from (conversation, manual edit, API sync) and how confident the extraction was. Your team can trust AI-extracted data while maintaining full visibility into its source.
• Human Override. AI extraction isn't perfect. Support agents and users can correct values at any time, and manual edits always take priority over AI extractions. The system learns from corrections to improve future accuracy.

How It Works

1. Automatic Extraction from Conversations

The simplest way to populate profiles is to let Mnexium do it automatically. When you make a chat completion request with learn: true, the LLM analyzes the conversation and extracts both free-form memories AND structured profile fields.

From this single exchange, Mnexium automatically extracts:

• ✓Profile: name = "Sarah Chen" (confidence: 0.95)
• ✓Profile: timezone = "America/New_York" (confidence: 0.85)
• ✓Memory: "User recently moved from San Francisco to New York"

The Memories view shows all extracted facts and preferences, with superseded memories grouped under their replacements.

2. Direct API Access

You can also read and write profile fields directly via our new Profiles API. This is perfect for:

• • Pre-populating profiles from your existing user database
• • Building profile editing UIs in your application
• • Syncing profile data with external systems
• • Correcting or overriding extracted values

curl -G "https://www.mnexium.com/api/v1/profiles" \
  -H "Authorization: Bearer $MNX_KEY" \
  --data-urlencode "subject_id=user_123" \
  --data-urlencode "format=full"

{
  "data": {
    "name": {
      "value": "Sarah Chen",
      "confidence": 0.95,
      "source_type": "chat",
      "updated_at": "2026-01-03T10:30:00Z"
    },
    "timezone": {
      "value": "America/New_York",
      "confidence": 0.85,
      "source_type": "chat",
      "updated_at": "2026-01-03T10:30:00Z"
    }
  }
}

3. Smart Superseding

One of the most powerful features of Profiles is automatic superseding. When a user's information changes, Mnexium intelligently handles the update:

The superseding logic considers:

• Confidence scores: Higher confidence values supersede lower ones
• Source type: Manual edits (confidence: 1.0) always win
• Recency: Newer extractions supersede older ones at the same confidence

Default Profile Schema

Every project comes with a default set of system fields that cover the most common use cases:

You can also define custom fields specific to your application through the Profile Schema editor in the dashboard. Add fields like company,role,subscription_tier, or anything else your AI needs to know about users.

Profiles in the Dashboard

The Profile Schema builder lets you define exactly what fields your AI should extract and track. Configure your schema once, and Mnexium handles the rest:

• • AI Extraction Mode: Choose between auto-fill from conversations, explicit API-only updates, or manual-only edits
• • Custom Profile Fields: Add fields specific to your business — company name, favorite fruit, whatever you need
• • Default System Fields: Name, Timezone, Email, and Language come pre-configured
• • Field Types: Mark fields as required or optional based on your data needs

The Profile Schema builder lets you configure extraction modes and define custom fields for your business.

API Reference

The Profiles API includes three endpoints:

For complete documentation with request/response examples, see the Profiles API documentation.

Use Cases

Profiles unlock powerful personalization capabilities:

Getting Started

Profiles are available now for all Mnexium users. Here's how to get started:

1. 1Enable learning: Add learn: true to your chat completion requests. Profile fields will be extracted automatically.
2. 2View profiles: Open the Memories dashboard and select a subject to see their profile card.
3. 3Customize schema: Visit the Profile page in your dashboard to add custom fields specific to your application.
4. 4Use the API: Integrate the Profiles API to read, update, and sync profile data programmatically.

We're excited to see what you build with Profiles. As always, we'd love to hear your feedback —

Happy building! 🚀