Each incoming record is validated against Antei’s internal schema:

• Required fields (e.g. contact_id, amount, jurisdiction) must be present
• Field types must conform (e.g. number, enum, ISO date)
• Conditional rules (e.g. country required if tax_applied) are enforced

❗ Invalid records at this stage are immediately flagged and not processed further.

Valid records are decomposed into sub-entities (e.g., Transaction, Product, Contact, Invoice, TransactionOp) and mapped to Antei’s internal data registry.

• Mappings are config-driven and aligned with the source system
• Each component inherits sync metadata and classification
• Nested objects are flattened for processing

📌 Example: One invoice may contain multiple transaction ops and line-item products.

Antei applies a weighted deduplication algorithm that combines:

• Fuzzy Matching — For loosely matching names, addresses, etc.
• Structured Field Scoring — For fields like tax ID, country, and external references
• Threshold Score — ≥ 90% similarity flags a match as Unprocessed (needs user override)
• New Object Creation — If similarity is < 90%, a new record is auto-created

Deduplication is applied both:

• At the entity level (e.g., Product, Contact)
• At the final transaction compilation stage

Antei checks relationships to ensure:

• Each entity references a valid product_id, contact_id, or jurisdiction_id
• Linkages are consistent with existing org-scoped records
• Duplicate relationship mappings are avoided

Example: A transaction line referencing a deleted product will be flagged for review.

Antei auto-applies enrichment tags such as:

• Tax classification (e.g. exempt, taxable, reverse charge)
• Entity role (e.g. supplier, customer, tax authority)
• Sync metadata (e.g. source, timestamp, entity_type)

These enrichments power downstream logic like taxability rules and country-specific filings.

A final check ensures all mandatory fields and linked entities are present.

If anything is incomplete, the entity is marked as:

• Unprocessed — Requires manual resolution
• Logged in dashboards (e.g. Unprocessed Products, Unprocessed Transactions)
• Skipped from downstream flows until corrected

Users can resolve unprocessed items via the UI:

• Fill in missing fields
• Override and reassign to correct entities
• Re-run validation post-manual mapping

✔️ Once approved, these records are stored as processed and marked clean.

Successfully validated records are committed to your workspace:

• Structured by module (Transactions, Products, Contacts, etc.)
• Linked to sync payloads and original metadata
• Available for reconciliation, return generation, analytics, and audit trails