predge-imessage

Predge in your iMessage bubble. Text a question about Polymarket whale trading to a Linq number and a card-carrying AI agent answers — it plans the query with OpenAI, buys the data from Predge over x402 (pays per query), and replies inside iMessage.

Built for the Agentic Commerce Hackathon — Linq (iMessage Agent) track. Reuses the same Predge x402 data + OpenAI brain as the x402-prava-bridge submission.

Flow

iPhone ──iMessage──▶ Linq number ──message.received webhook──▶ predge-imessage
                                                                a) OpenAI tool-calling: what to buy?
                                                                b) x402 buy from Predge (USDC, per query)
                                                                c) OpenAI: write the reply
                                                                d) send back via Linq  ──iMessage──▶ iPhone

Run

npm install
cp .env.example .env    # fill LINQ_API_TOKEN + OPENAI_API_KEY

# 1) prove the brain + data pipeline with NO iPhone/Linq needed:
npm run simulate "is wallet 0x8dab6cf42d7e4d1cd6cca897b9e12a8cbe6d69e8 smart money?"

# 2) run the webhook server, expose it over HTTPS (ngrok/Railway), point the Linq
#    webhook at https://<host>/webhook, then text the Linq number from your iPhone:
npm run server

For a 0-real-money demo set PREDGE_BASE=http://localhost:4021 (local Base Sepolia Predge). Otherwise it buys from prod on Base mainnet (~$0.005/query, real).

Linq track fit

The agent is a genuine two-way iMessage agent on Linq's Partner API (message.received webhook in, /chats/{id}/messages out, with read receipts + typing indicators). What makes it more than a chatbot: every reply is paid for on-chain per query — the agent buys the Predge data it sends via x402, so it's real agentic commerce living inside an iMessage thread.

Env

LINQ_API_TOKEN, LINQ_API_BASE_URL, OPENAI_API_KEY, OPENAI_MODEL (default gpt-4o-mini), PREDGE_BASE, FLOAT_PRIVATE_KEY, PORT. See .env.example.