Day Zero Semantic Layer
AI generated semantic layer on day zero, right after data source is connected - building the foundation of automated and trusted data governance. No manual modeling or Yaml/Json upload required.
How to setup
Summary
Connect data source (Snowflake, BigQuery, Databricks, PostgreSQL, Athena etc.).
Schema sync → captures tables, joins, and data types.
Query history sync → learns past business usage.
Day Zero initialization → builds autonomous semantic relationships.
Data stats collection → captures data completeness and quality indicators.
Ready state → Connecty is now fully semantic and query-ready.
Goals and KPI recommendation
Overview
The DayZero Semantic Layer (DayZero SL) feature automatically builds a semantic model the moment data connection is setup. It discovers datasets, relationships, metrics, and joins and initializes Connecty’s Autonomous Semantic Graph (ASG) without any manual configuration.
DayZero SL uses existing metadata, query history, and data samples to form a holistic understanding of your data’s logical and business context, and allows admins to validate (human-in-the-loop).
What this means for your business: Teams can start extracting insights based on your custom definitions from day one, instead of waiting for manual upload of business definitions.
Key Capabilities
1. Auto Semantic Layer Generation
During DayZero sync, Connecty runs a Query-to-Semantic Layer and Semantic Layer Reconciliation process:
Converts discovered relationships into an internal grammar representation
Merges inferred logic with existing verified grammar (if any)
Ensures consistency and removes ambiguous mappings
Business impact: Analysts get an immediately usable data reasoning layer. The AI understands metric relationships and naming conventions without needing model scripts or dbt packages.
2. Query History Analysis
Connecty analyzes past query history (from tools like Athena or Snowflake) to identify:
Frequently used joins and filters
Metrics and business expressions used in previous SQL
Naming patterns that indicate business intent (e.g., “total_sales”, “monthly_active_users”)
Business impact: The system learns the organization's natural data language automatically, so the first AI answers already align with how your team talks about KPIs.
3. Automatic Schema Discovery
When a new data connection is added (e.g., Snowflake, Databricks, BigQuery, PostgreSQL, Athena), Connecty automatically scans:
Tables, columns, and joins
Data types and primary keys
Relationships between entities
This information is used to populate the ASG (Autonomous Semantic Graph).
Business impact: Reduces dependency on engineers to define structures manually. The semantic model starts forming immediately after connection, accelerating the time to first insight.
4. Column-Level Data Statistics
Connecty automatically computes data quality metrics during sync, including:
Null counts and percentages
Unique value counts
Minimum, maximum, and average (for numeric fields)
Business impact: Analysts and AI agents can evaluate data completeness and reliability instantly, improving the accuracy of generated queries and recommendations.
5. Day-Zero Query History Sync Workflow
Connecty executes a chained sync process:
DE Query History Sync – collects past query usage.
DW Query History Sync – builds semantic clusters from those queries.
Grammar Reconciliation – updates the ASG with query-derived logic.
Completion Event – marks the environment as ready for semantic exploration.
Business impact: Connecty’s AI doesn’t start from zero — it starts with a pre-trained understanding of your organization’s analytical behavior.
6. Parallel Semantic Reasoning
The DayZero process runs multiple semantic inference steps in parallel:
Entity extraction
Join and key relationship detection
Metric and aggregation classification
Business impact: The semantic graph is ready within minutes, even for large warehouses with hundreds of tables.
7. Context Graph
Every generated DWQuery now includes a chart specification, describing how results should be visualized (e.g., line chart, bar chart, trend).
Business impact: The semantic layer doesn’t just interpret queries — it also encodes visualization intent, allowing Connecty to produce instant charts and summaries directly in chat.
8. Verified-Entity Enforcement
When “verified-only” mode is enabled, Connecty’s DayZero-generated semantic layer:
Flags unverified entities
Restricts reasoning and query generation to approved definitions
Business impact: Ensures business consistency from day zero — only trusted metrics and dimensions are used in AI responses.
What Happens Next
Once the DayZero Semantic Layer is initialized:
Business users can start asking natural language questions immediately.
Data stewards can begin verifying and refining entities.
Connecty continues to learn and reconcile new relationships automatically.
Custom Semantic Inputs (Optional add-on)
Some customers want Connecty to reflect their existing conceptual models and dictionaries, not just what we infer from the warehouse. That’s available as a custom add-on.
Industry / Conceptual Models (Custom)
Yes - Connecty can ingest and map industry models (e.g. OSHA, banking/credit, financial, loyalty, insurance) into the semantic layer so the AI reasons in terms like Account, Policy, Claim, Incident, Credit Line, Loyalty Member instead of just table names.
You provide your existing models (diagrams, dictionaries, dbt / YAML / JSON, catalog exports); we map those concepts onto your actual schemas and plug them into the Day Zero Semantic Graph.
Corporate Dictionaries & Taxonomies (Custom)
Yes - Connecty can also ingest corporate business glossaries and taxonomies:
Business glossaries (names, descriptions, owners)
Domain taxonomies (e.g. Product → Category → Subcategory, Region → Market → Territory)
Synonyms/abbreviations (e.g. GMV, LTV, LOB)
Exports from tools like Collibra, Alation, etc.
These inputs are used to normalize your language, so “churn”, “attrition” and “cancellations” all resolve to the same governed KPI, and “verified-only” mode keeps AI answers aligned to approved definitions.
Last updated