Building Conversational Commerce with Multi-Agent Swarms on GCP: A Working Prototype

HappyDelivery is a functional conversational-commerce prototype built with LangGraph, Gemini, FastAPI, and Google Cloud that turns a three-layer architecture into an automated grocery-shopping workflow. Its ingress layer validates platform-specific webhook signatures from channels such as WhatsApp, RCS, Apple Messages, and SMS, normalizes messages, and queues them through Pub/Sub for fast acknowledgement; its Redis-based session layer preserves customer profiles, carts, preferences, addresses, and tokenized payment references across conversations; and its LangGraph agent layer routes requests among profile, inventory, and transaction specialists. Deterministic tools handle catalog operations, pricing, taxes, delivery fees, and payment execution to avoid language-model calculation errors, while Gemini Flash performs low-latency routing and Gemini Pro supports more complex shopping and checkout tasks. Deployment tests reported successful webhook validation, sub-millisecond Redis lookups, a simulated three-turn order flow, and complete automated test success, while estimated model costs for a three-turn shopping session were below one cent. The implementation emphasizes asynchronous processing, secure payment tokenization, specialist-agent design, externalized state, and progress messages to manage the latency of multi-agent reasoning.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.