Schema Promotion

---

What Is Schema Promotion?

Schema Promotion formalizes structure and intent. When Azi sees a new pattern in your data, she proposes a schema. Only you can promote it - confirming that structure, naming the fields, and setting it in motion.

You're not labeling raw data. You're defining the truth your system will operate on.

---

Why It Deepens the Moat

Schema Promotion extends governance from actions to structure:

Today	With Schema Promotion
Actions require approval	Structure requires approval too
AI sees your data	AI only acts on promoted structure
Queries work on raw data	Queries work on governed schemas

---

How It Will Work

1. Pattern Recognition

Azi sees incoming data and proposes a candidate schema:

// Azi proposes:
{
  "field": "tempRaw",
  "observed_type": "number",
  "sample_values": [72.4, 73.1, 71.8]
}

2. Your Review

You review, name, and confirm types:

// You promote:
{
  "field": "temperature",
  "type": "number",
  "unit": "celsius",
  "promoted": true
}

3. Activation

Once promoted, the schema becomes active:

• Logic can bind to it
• Queries resolve against it
• Agents can respond to it

---

What Changes After Promotion

Until you promote a schema:

• Logic can't bind to it
• Queries don't resolve against it
• Agents ignore it

Promotion is the activation point. It's when data becomes usable, and when surfaces can respond.

---

Example: SensorReading

Azi sees this stream:

{
  "device_id": "abc-123",
  "temperature": 72.4,
  "location": "north-lab"
}

Azi proposes:

"Promote this as SensorReading?"

You approve. Now:

• A rule can trigger when temperature > 75
• Dashboards can load all SensorReading entries
• Azi routes future data with confidence

All from a single yes: you promoted it.

---

Schema Packs (Future)

Once promoted, schemas become portable through Schema Packs:

• Share structure across teams
• Reuse logic across surfaces
• Version and export your knowledge

Promoted schemas become part of your shared language.

---

Benefits When Available

Benefit	What It Means
Explicit governance	Nothing is assumed - you approve structure
Version history	Track every schema change
Safe experimentation	Fork schemas without risk
Composable structure	Build complex schemas from simple ones
Audit trail	Know who promoted what, and when

---

Current Status

Capability	Status
Basic schema inference	Available via connections
Explicit schema promotion	In development
Schema versioning	Planned
Schema forking	Planned
Schema packs	Future

---

What You Can Do Today

While full schema promotion is in development, the foundation is working:

• Create Connections - Connections include schema configuration
• Warm Queries - Work with structured data today
• Working with Azi - AI-assisted query building