Off-chain services for Nuvex. This repository does not deploy programs and it is not a source of protocol truth.

• api/ — HTTP read API. GET /health returns authority: none. Requests, nodes, and network counts come from the indexer store. GET /v1/prices returns a median of fresh public observations and is not a chain result. Jobs and models return 501.
• indexer/ — Decodes known account layouts from getProgramAccounts and writes PostgreSQL (or a JSON snapshot). It does not invent rows.
• data/ — price-provider adapters. A source that is stale, unsigned, or missing a timestamp is dropped. The median is omitted when too few sources remain.
• ai/ — a directory and a README. There is no model.
• infra/ — local compose file and monitoring config.

The website, the documentation site, and the on-chain protocol are separate repositories. Build the oracle node image in the protocol repository and tag it nuvex-oracle-node:local before using the node compose profile.

start needs NUVEX_RPC_URL (or NUVEX_SOLANA_RPC_URL), at least one program id (NUVEX_ORACLE_CORE_PROGRAM_ID, NUVEX_ORACLE_REGISTRY_PROGRAM_ID, NUVEX_VERIFICATION_PROGRAM_ID), and either DATABASE_URL or NUVEX_READ_MODEL_PATH. Without those, the process exits 2. --health only checks the process.

Apply the SQL in indexer/prisma/migrations/ before pointing the API at Postgres.

pnpm install
make check
cp env/.env.local.example .env.local

make check formats, lints, tests, and validates the Prisma schema. It does not start a database.