Drop files into the context folder and Copilot can read them, extract the data model inside, and help you build it into your project. This page covers what Copilot can work with, what it does with it, and how to get the most out of it.

Supported file types

• Excel spreadsheets (.xlsx, .xls)
• CSV files
• Text files (SQL, JSON, YAML, and others)
• Images (diagrams, screenshots, ERDs)
• PDF documents

What Copilot can extract and analyze

1.  Data model documentation.  Table definitions, column specifications, relationship mappings, primary and foreign key structures, data types, and constraints.

2.  SQL and DDL.  Parse CREATE TABLE statements, extract schema definitions, identify relationships from DDL, and interpret view definitions.

3.  Configuration files.  dbt YAML configurations, Snowflake tag definitions, governance metadata, and naming conventions.

4.  Spreadsheets and matrices.  Table-to-table mappings, column lineage documents, data governance matrices, and business glossaries.

5.  Visual analysis.  ERD diagrams, with entities and relationships identified; data flow diagrams; and architecture documentation.

6.  Transformation specs.  Mapping documents, transformation logic, join specifications, and data lineage.

What Copilot does with it

• Creates database objects — tables, views, relationships — in your project.
• Generates Mermaid diagrams for visualization.
• Identifies missing relationships and suggests mappings.
• Extracts metadata for governance fields.
• Validates against your existing project structure.
• Recommends schema organization.

Limitations

• Copilot does not bulk-import a file into your project as a single automatic step. It extracts the model from the file and helps you build the objects using the modeling tools.
• Large files over 5 MB may hit processing limits.

How to work with Copilot using files

The process

1.  Upload your file.  Drag and drop, or click attach in the chat. Copilot can handle multiple files at once, and there is no need to explain what is in the file upfront.

2.  Copilot analyzes it.  It examines the file structure and contents, extracts the relevant data model information, and summarizes what it found.

3.  You state your goal.  For example: create these tables in my project; show me the relationships as a diagram; map this to my existing schema; or extract the governance metadata.

4.  Copilot takes action.  It uses the modeling tools to create objects, generates visualizations, makes recommendations, and asks clarifying questions where needed.

Best practices

• Be specific about your goal.  “Create the Party domain tables from this model” works far better than “what do you think about this file?”
• Give context when it helps.  Tell Copilot whether a file is source system documentation, the target tables you want to build, or the regulatory subset to focus on.
• Iterate in steps.  Start with one domain or subset, review what Copilot created, then move to the next piece.

What makes this powerful

1.  No manual translation.  You do not type out table definitions by hand.

2.  Relationship detection.  Copilot spots foreign-key relationships in documentation.

3.  Governance extraction.  It pulls out PII flags, data classifications, and similar metadata.

4.  Validation.  It checks extracted structure against your existing project.

5.  Multiple formats.  Excel, CSV, SQL files, even screenshots of ERDs.

Example workflow

You:  [upload financial-model.xlsx]
Copilot:  “Found 35 tables across 9 domains. Here is the summary...”
You:  “Create the Account domain tables first”
Copilot:  [creates dim_account, fact_account_balance, etc.]
You:  “Now show me how Party connects to Account”
Copilot:  [generates Mermaid diagram of relationships]

The key: upload first, then state your specific goal. Copilot handles the heavy lifting of parsing and translation.

In practice

Step one — choosing the files to upload into the context folder.

The files attached in SqlDBM Copilot, ready to analyze.