Objective

Make the imported legacy content not the final migration result, but the raw material for systematic AI editing.

Basic Scenario

The AI agent receives:

• original record data from the legacy DB;
• the newly created Concept;
• general project context and rules.

After that, the agent updates the Concept content: reworks the structure, formatting, and presentation while preserving confirmed original facts.

What to Do

• Define a unified interface for AI processing of Concepts.
• Pass the original legacy data to the agent as factual context, and the new Concept as the editable result.
• At the first stage, work primarily with content, without complicating the DB schema with new fields for specific editorial structures.
• Allow the agent to structure the material, add formatting, and internal linking.
• Prohibit inventing unverified facts and commercial results.
• Add result validation before saving: markup correctness, validity of internal links, and other deterministic checks.
• Make the processing repeatable and suitable for mass execution across the imported database.

Result

After migration, old content is systematically transformed into a modern knowledge base without manual rewriting of every page and without expanding the schema for every type of material.

Definition of Done

• A single AI pipeline is applicable to different types of Concepts.
• Original facts are preserved.
• The result passes programmatic validation before saving.
• The imported database can be processed in bulk.
• No separate AI code is required for each legacy table.