callbotagent.aiDocs
v1 · Contract draft
ENPL
Open dashboard
Preview

Custom store integration: end-to-end

Preview documentation. The interface is being rolled out — examples are indicative, and we will confirm final values at launch.

This guide takes you from an empty store to the first confirmed conversation results — in five steps. Custom store integration relies entirely on the public Commerce API: you send catalog, customers, and repurchase intents, and we return dispositions and the final order state. Base URL for all calls: https://api.callbotagent.ai/api/v1/commerce.

Before you begin

You need an API key (always masked in the dashboard and examples: cba_test_••••••••) and scopes matching the operations you perform. Sign every request (HMAC v1) — details in Authentication. Order matters: catalog and customers first, then orders — the bot will not offer a product or call a customer it cannot verify.

End-to-end flow

1

Sync the catalog

POST /products/batch — send the catalog with stable external IDs, availability, currency, and categories. The bot will not offer a product it cannot verify in your catalog, so a complete, current feed is the foundation of the integration.

2

Add customers with consent

PUT /customers/{externalCustomerId} — pass phone in E.164, locale, and consent proof for contact. Records without valid consent do not enter the calling queue — that is mandatory, not optional.

3

Submit repurchase intents

POST /orders with order_kind=repurchase_intent, expected contact date in contact_not_before, explicit line items and prices. Intents appear in the dashboard as orders — never as anonymous sales contacts.

4

Owner policy takes precedence

Countries, calling hours and days, retry policy, DNC lists, and store-owner budget always win. Partner data may narrow policy (e.g. a tighter contact window) but never relax it.

5

Receive results

Consume signed webhooks or poll GET /orders/{externalOrderId}/result. In final_order_snapshot you get the final conversation state: confirmed, modified, or declined products and prices.

Field mapping

Common mappings from your platform model to Commerce API fields.

Your platform Commerce API
Merchant / market One store + custom connector
Repurchase prediction order_kind=repurchase_intent
Expected contact date contact_not_before
Prediction confidence / score priority / confidence
Final decision after conversation final_order_snapshot

If your platform distinguishes multiple markets or sales accounts, all map to one store served by the custom connector — keep segmentation in external IDs and metadata, not separate integrations.

Partner profiles on the custom platform

Partners predicting repurchases use the same Commerce API as any other custom store. There is no separate endpoint or platform for them — only how they map their data to contract fields differs.

Typical mapping for foresight / repurchase:

  • repurchase prediction → order_kind=repurchase_intent,
  • expected next purchase date → contact_not_before,
  • prediction confidence → priority / confidence.

Catalog, customers with consent, owner policy precedence, and result delivery work identically to any other custom store. The partner profile is only a mapping layer — the rest of the flow is unchanged.

Related material