Answer Preparation Steps

This section describes how to use the EXPLAIN tab under Query Inspector.

Connecty AI breaks the question into a pipeline of four steps to generate accurate and secure response from structured data: Intent -> Context -> Grammar -> Query

Here’s what each step means:

1. Intent

Goal: Understand what the user wants.

• The system retrieves and decomposes the semantic intent behind the natural language question.

• For example, if a user says, “How many users signed up last month?”, the intent might be:

  • Retrieve count of user signups

  • Filter by signup date in the previous month

This is the "what" of the request — the high-level task the system must solve.

---

2. Context

Goal: Understand where and how to apply the intent using data structure and relationships.

• Uses metadata, data lineage, and semantic layers to understand:

  • What “users” means (e.g., user_table)

  • What “last month” maps to in the data (e.g., signup_date)

• Leverages previously learned relationships between tables, columns, and business terms.

This is the "framework" of meaning based on the underlying database and previous interactions.

3. Grammar

Goal: Transform structured intent into a syntactically correct query logic.

• Converts the mapped concepts into a logical grammar structure using Connecty’s autonomous semantic graph.

• Builds the internal representation of operations: SELECT, WHERE, GROUP BY, etc.

• Ensures the query will both make sense to the database and stay true to user intent.

Think of this as the "blueprint" from the parsed meaning into executable form.

---

4. Query

Goal: Generate and execute the final SQL.

• Outputs the final query that can be run on the data warehouse.

• Executes this query and shows the results in form of a data table

This is the "deliverable" — the actual output that provides the user with an answer.

---