@mostlyrightmd/econ

Economic-indicator data for the mostlyright TypeScript SDK — CPI (headline / core / YoY), PPI, nonfarm payrolls (NFP), U3 unemployment, initial jobless claims, GDP, and Fed decisions, sourced from FRED/ALFRED + BLS/BEA/DOL/Federal Reserve and joined to Kalshi + Polymarket econ markets for leakage-free settlement pairs. Mirrors the Python mostlyrightmd-econ distribution surface. Declares @mostlyrightmd/core as a peer dependency.

Install

pnpm add @mostlyrightmd/econ @mostlyrightmd/core

Surface

Surface-equivalent to the Python mostlyright.econ vertical (camelCase TS signatures):

• history(indicator, fromDate, toDate, { vintages }) — observation rows. vintages: "settlement" (default) returns the settlement-grade first print (the value as-of the Kalshi expiration); vintages: "all" returns every vintage for feature engineering.
• releases(indicator) — the release calendar / schedule for an indicator.
• researchEcon(seriesOrContract, fromDate, toDate, { asOf }) — leakage-free settlement pairs (market outcome + first-print value + as-of features). When asOf is given, any pair whose knowledge_time (= vintage_date) is after it throws LeakageError — a future revision can never leak into a backtest.

Load-bearing invariants (parity with Python)

• First-print / ALFRED vintage discipline. Kalshi settlements contractually exclude post-expiration revisions; the backtest layer serves as-released vintages, never the revised series. Vintage selection happens at read time via vintages.
• Per-series settlement routing. 48 of the Kalshi Economics series settle to Trading Economics, not the government agency (incl. KXUSPPI PPI-MoM, KXUSNFP). Routing is genuinely per-series: KXUSPPI (MoM) → TradingEconomics, but KXUSPPIYOY (YoY) → BLS. The routing table is codegen-shared from schemas/kalshi-econ-settlement.json (the Python↔TS anti-drift gate).
• TE-settled series are honest, never fabricated. For a Trading-Economics-settled series, researchEcon ships the agency first print labeled settlementGrade=false and emits a divergence warning — it NEVER fabricates a TE value.
• IndicatorNotYetReleasedError, not []/null. "Not yet released" is a normal, branchable control state distinct from "unavailable". The error is re-exported from @mostlyrightmd/core (Python↔TS lockstep taxonomy).
• Parity firewall. Econ never registers into the TS research() / merge / live-source equivalents — an econ row must never reach the weather settlement join (the same isolation the Python vertical and CWOP use).

Browser / Node constraints

The fetch layer uses the standard fetch API and JSON — no Node-only API on a browser-viable path (mirrors the weather TS browser-safe boundary). API keys (FRED_API_KEY / BLS_API_KEY for ALFRED / BLS-v2) are read from the Node environment only; they are never bundled into a browser build.

Schema

The EconObservationRow (schema.econ.observations.v1) column types are codegen-derived from the shared JSON via @mostlyrightmd/codegen — never hand-written in TS. Regenerate with pnpm codegen.

Docs

See https://mostlyright.md/docs/sdk/ for quickstart and the full API reference.