Backend Demo Seeding

These scripts seed the PostgreSQL-backed Customer Identity Resolution demo:

• init_sample_data.py creates the tenant-scoped raw-profile fixture.
• generate_fuzzy_match_demo.py adds address and company variations for fuzzy-match verification.
• seed_full_demo_data.py enriches resolved profiles with CRM, relationship, persona, and behavioral-event fixtures.
• seeding_utils.py provides validated metadata models and reusable UUID, random, URL, event-partition, and vector helpers.
• seed_demo_metadata.json holds the full-demo catalogs: CRM entities, campaign fixtures, names, data sources, agents, domain settings, and persona archetypes. Credentials and environment overrides remain outside the file.
• seeding_demo_contents.json holds the validated recommendation catalog with seven realistic items for each supported domain.
• seeding_content_items.py loads and validates seeding_demo_contents.json before seeding cdp_content_items.
• utils.py contains the shared PII hashing helper used before database insertion.

The scripts require the backend seed dependencies and an installed local leo-customer360-dao package. The normal CIR launcher installs those dependencies and invokes the scripts from the repository root:

customer360-backend/identity_resolution/run-demo.sh

They are also included in the backend Docker image at /opt/customer360-seeding for the one-shot compose, Kubernetes, and deployment seed jobs.

The content catalog can be seeded independently after the database schema and demo tenant exist:

python customer360-seeding/dev-backend-seeding/seeding_content_items.py

seed_full_demo_data.py imports the same function, so the full demo flow and the standalone content seeder use one catalog implementation.

Full-demo seeder

FullDemoSeeder coordinates database stages, S3 event generation, source statistics, and connection cleanup. It preserves the existing seed order and commits once the stages complete. A failure rolls back the database and is reported explicitly; S3 writes are external side effects and are not rolled back by PostgreSQL.

Metadata is loaded relative to seeding_utils.py, not the working directory. Missing files, unknown keys, or invalid catalog shapes fail before seeding. The Docker image and deployment bundle copy the entire dev-backend-seeding directory, including the JSON file.

The fixtures retain their existing synthetic VN/EU/US names. Master-profile names and retail contact details may be synthetic plaintext; other inherited PII remains hashed. Persona embeddings remain deterministic synthetic vectors, not production model embeddings.

Run the focused tests from the repository root using an environment with the backend seeding dependencies and local DAO installed:

python -m pytest customer360-seeding/test_seed_full_demo_data.py -q