Macro Market Structure Notes: Inside the Atlas Macro Radar Economic Calendar and Its Keyless FRED Spine

The Atlas Macro Radar economic calendar now reads every published US macro release from a keyless FRED feed, attaches the real previous and actual figures, and shows an honest pending marker wherever a number does not yet exist, with zero fabricated values on the public surface.

What the Atlas Macro Radar Economic Calendar Ships

The economic calendar lives as its own branded page at https://atlas.freedomcore.io/economic-calendar/, built from /var/www/freedomcore/economic-calendar/index.html and served through a dedicated nginx location block. The H1 is "Atlas Macro Radar" and the route is wired into sitemap-atlas-apex.xml, so the page is discoverable as a standalone destination rather than a cramped homepage section. The backend behind it is economic_calendar.py, fronted by atlas-backend.service.

The most important design choice is documented in the build notes from 2026-06-16: a frontend function, atlas.js renderMetric(), had been generating random fake forecast and previous percentages on the public calendar using Math.random()*2.5+0.1. That code was removed. The calendar now prints an honest "awaiting release" or "n/a" marker instead of inventing data, which is the correct behaviour for a trust surface that Google and AI crawlers read.

The Keyless FRED and ALFRED Data Spine

The macro values come from real public sources, not a paid consensus vendor. /root/Atlas_Backend/macro_values.py pulls US figures from the keyless FRED CSV endpoint (fredgraph.csv), cached in-process for six hours and fail-safe so any error returns a pending marker rather than a guess. The verified series include CPI (CPIAUCNS, year over year), PPI (PPIFID), core PCE (PCEPILFE), nonfarm payrolls (PAYEMS, month over month change in thousands), and JOLTS (JTSJOL). A practical gotcha was logged: FRED hangs for roughly twenty seconds on a custom User-Agent header, while the default urllib agent fetches in about 0.4 seconds.

A second engine, /root/Atlas_Backend/fred_calendar.py, calls the FRED releases API and filters 329 FRED releases down to a curated allowlist of roughly seventeen market-movers by release id, for example CPI (release 10), PPI (46), PCE (54), nonfarm payrolls (50), and JOLTS (192). After this merge the US window expanded from 7 to 24 events across the 21-day horizon, with seven carrying a real surprise reading.

Revisions are handled by /root/Atlas_Backend/macro_revisions.py, which uses ALFRED, the FRED vintage API, to rebuild a metric in first-vintage versus current-vintage form and report a revision only when the displayed value actually changed. Recorded examples include nonfarm payrolls revised from +99K to +179K and PPI from 6.0% to 5.7% year over year. CPI is correctly left without a revision line because index revisions there are noise.

Cross Asset Market Intelligence and Source Quality

Atlas covers more than the United States. /root/Atlas_Backend/uk_eu_values.py wires UK headline data from the Office for National Statistics JSON endpoints: UK CPI (D7G7), unemployment (MGSX), and retail sales volume (J5EK). A logged detail is that ONS rejects the default urllib agent with a 403 and needs a browser User-Agent, the opposite of FRED. Live readings captured in the build notes include UK CPI at 2.8% year over year and unemployment at 5.0%.

Source quality is treated as a first-class concern. Euro-area HICP from Eurostat is stale, with the latest point at 2025-12, so the calendar shows an honest n/a behind a 75-day freshness gate that auto-activates when the feed catches up. UK GDP series were also held back because the candidate identifiers are stale or measure gross value added rather than GDP. This is the kind of data-source methodology that belongs in macro market structure notes: stating exactly which series are live, which are gated, and why.

SEO Indicator Pages and Economic Calendar Research

The economic calendar research surface extends into fifteen dedicated indicator pages generated by /root/Atlas_Backend/atlas_indicator_pages.py, covering paths such as /economic-calendar/us-cpi/, /economic-calendar/us-nonfarm-payrolls/, and /economic-calendar/us-core-pce/. Each page carries the latest value, the previous value, a surprise reading, an 18-point sparkline, a last-six-prints table, methodology, and a link to the official source. The pages are self-canonical and described with WebPage, Thing, and BreadcrumbList JSON-LD, following the structured-data pattern set out in the Schema.org BlogPosting and related vocabularies. Regeneration runs daily via atlas-indicator-pages.timer at 07:20 so the data never goes stale. The underlying FRED documentation behind this OpenBB DefiLlama RSS market spine of public feeds sits alongside the broader OpenBB docs that the wider research stack draws on.

For the symbol side of this AI trading research dashboard, the macro layer connects to the per-asset coverage in Atlas symbol coverage, and the published research record lives in Atlas notes. The franchise context sits at FreedomCore, and the live trading surface, which is governed separately, is Maverick.

Caveats and What Is Not Yet Live

Several items remain explicitly unproven or pending. Forecast and consensus values stay absent because no free redistributable source exists, so the calendar never shows a fabricated forecast. Euro-area inflation remains honest n/a until the Eurostat feed refreshes past its 75-day gate, and several UK GDP and retail identifiers were not wired because the candidate series did not pass the freshness or definition check. A Sprint 6 substrate-sentence upgrade, mapping the 180 per-symbol substrate files to specific macro events, was deliberately not built to avoid non-sequitur reads, so the current "Atlas read" line is a generic-true category framing rather than an extraction. These boundaries are the point: the public macro surface ships only data it can verify.