Skip to main content

Documentation Index

Fetch the complete documentation index at: https://labs.prompthon.io/llms.txt

Use this file to discover all available pages before exploring further.

Need help?
Ask in DiscordOpen a GitHub issueSupport guide

​
Summary

This starter shows a messaging-native transaction assistant in the smallest useful shape: bounded business context, intent capture, plan selection, user confirmation, and payment handoff.

​
Status

starter Source code: ecosystem/examples/messaging-transaction-assistant-starter

​
Why It Exists

Messaging surfaces are becoming places where users complete real tasks without switching apps. The assistant pattern is not “let the model pay for something.” It is a bounded flow where the assistant collects intent, presents a reviewable choice, and hands off to a trusted payment surface after explicit confirmation. Meta’s May 2026 Business AI on WhatsApp launch sharpens the business-side version of that pattern: a small business can ground the assistant in its profile, catalog, and support context, let it answer customer questions after hours, and still step in directly whenever needed. This starter keeps that shape vendor-neutral and repo-native.

​
Folder Structure

messaging-transaction-assistant-starter/
├── README.md
├── SOURCE_NOTES.md
├── index.mdx
└── src/
    ├── run_demo.py
    └── transaction_flow.py

​
Included Sample Files

  • src/transaction_flow.py: typed helpers for bounded business context, intent capture, plan selection, confirmation, and payment handoff
  • src/run_demo.py: tiny command-line demo of the flow
  • SOURCE_NOTES.md: source lineage and attribution boundary

​
Flow Boundaries

The assistant may:
  • answer from a local business profile, catalog, and support notes
  • infer a recharge-like intent from a message
  • ask for missing plan or recipient details
  • show a confirmation summary
  • keep business takeover available when a human seller wants to step in
  • prepare a handoff payload for a payment surface
The assistant must not:
  • execute payment
  • store card or bank credentials
  • bypass user confirmation
  • imply vendor-specific integration that does not exist

​
Current Source Signal

The current signal for this refresh is Meta’s May 2026 Business AI on WhatsApp launch for Indian small businesses. Meta frames the assistant around catalog-grounded product questions, key business information, after-hours coverage, and seller control. That makes the reusable handbook lesson more specific than a generic chatbot: the assistant should stay inside a visible business context boundary while a human still owns final overrides and payment.

​
Next Steps

  • Add a richer state machine for missing details.
  • Add localization fixtures for different messaging markets.
  • Add a fake payment adapter that only records handoff state for tests.