@snowcone-app/chat-contracts

Types for the Snowcone AI design-chat wire contract.

Snowcone's chat endpoint (POST https://api.snowcone.app/ai-generations/chat/stream) speaks the AI SDK UI Message Stream protocol over SSE, so any AI SDK v5+/v6 client (useChat, assistant-ui, …) can consume it directly. This package is the typed contract for what streams back: the data-part payloads (generated images, product suggestions, mockup cards, committed designs, progress stages) and the chunk union.

Full integration guide: https://developers.snowcone.app/guides/ai-chat

Install

npm install @snowcone-app/chat-contracts

Typing your messages

import { useChat } from "@ai-sdk/react";
import type { UIMessage } from "ai";
import type { SnowconeChatDataParts } from "@snowcone-app/chat-contracts";

type SnowconeUIMessage = UIMessage<never, SnowconeChatDataParts>;

const { messages, sendMessage } = useChat<SnowconeUIMessage>({ /* … */ });

Every part in message.parts is then discriminated: data-images, data-products, data-mockup, data-design, plus the transient data-stage progress labels.

Data parts

| Part type | Payload | Meaning | |---|---|---| | data-images | ChatImagesData | AI-generated or processed artwork this turn (image count varies by pipeline — don't assume a fixed grid) | | data-products | ChatProductsData | Product suggestions from catalog search — price/lowestPrice are integer USD cents (3772 = $37.72) | | data-mockup | ChatMockupData | A product mockup preview card, including the variantId it was rendered for | | data-design | ChatDesignData | Result of an explicit save/commit | | data-stage | ChatStageData | Transient progress label ("generating image", …) — rendered live, never persisted |

A data-mockup part carries viewUrl/editorUrl only for a committed design; a stateless preview has neither, and your UI should offer a Save affordance instead (the orchestrator never auto-commits). When committing, pass the part's variantId along so the saved design renders the same variant (color) the user previewed.

Server helpers

UI_MESSAGE_STREAM_HEADERS, encodeSseChunk, and encodeSseDone are the SSE framing helpers Snowcone's backend uses to emit the protocol. You only need them if you re-emit or transform the stream; a pass-through proxy (the normal integration) does not.

License

MIT