Market Monitor

A source-aware decision dashboard that connects four technology-and-markets questions in one interface: dollar liquidity, Bitcoin, AI infrastructure, and nuclear energy.

Screenshot uses public market data. AI interpretation is optional; the underlying dashboard remains usable without model credentials.

Recruiter quick read

Signal	Evidence in this repository
Analytics / BI	Multi-source metrics are normalized into status, trend, freshness, and provenance fields before presentation.
Energy + technology	Dedicated nuclear and AI-infrastructure views connect operating data, market structure, and source notes.
Product engineering	Responsive Next.js interface, server-side API routes, error states, Docker deployment, and scheduled snapshot refreshes.
Decision discipline	Static snapshots provide a reproducible baseline; live overlays and AI summaries are clearly separated from source data.

What the product does

• Home — one-screen regime summary across macro, BTC, AI, and nuclear signals.
• Dollar & Macro — a dated snapshot of dollar-system metrics with source notes and explainable status thresholds.
• Bitcoin — price, cycle, macro, and network context in a research-oriented terminal view.
• AI Infrastructure — company and model snapshots, public-market overlays, and source-level confidence metadata.
• Nuclear Energy — reactor, uranium, and public-company signals with dated evidence and a separate live-source layer.
• Optional interpretation — ask/explain routes and daily briefings run server-side; credentials never enter the client bundle.

Architecture

flowchart LR
    A[Public market and macro sources] --> B[Server-side fetchers]
    C[Versioned JSON snapshots] --> D[Validation and metric builders]
    B --> D
    D --> E[Next.js dashboards]
    E --> F[Status, trend, freshness, provenance]
    G[Anthropic / xAI optional] --> H[Explain, ask, briefing routes]
    D --> H
    H --> E

Loading

The key design choice is the boundary between facts and interpretation. Metric values, dates, and sources are structured data; model output is an optional narrative layer.

Local desktop app

The snappy local client is a native Rust app in desktop/. It opens the same five desks — home, Bitcoin, dollar, AI, and nuclear — from a disk cache, then refreshes public sources in the background. No Node server and no API keys.

cd desktop
cargo run --release

1–5 switch desks, R refreshes, and Esc clears the selection. cargo run --release -- --fetch prints one snapshot and exits.

AI and nuclear evidence stays on the dated snapshot shipped with the repo. Equity, rates, dollar, and Bitcoin prints are live public feeds. Rule readings are explicit thresholds, not a model call.

The Next.js app below is the browser version of the same monitor.

Run locally

Requires Node.js 20.9 or newer.

npm install
npm run dev

Open http://localhost:3000. Public-data and snapshot views render without LLM credentials.

Optional server-side integrations:

Variable	Purpose
ANTHROPIC_API_KEY	Explain, ask, interpretation, and snapshot-refresh routes
FRED_API_KEY	Live FRED overlays and yield-curve history
XAI_API_KEY	Optional recent-X summary on AI and nuclear views
EXPLAIN_MODEL / ASK_MODEL	Override Anthropic models
XAI_SUMMARY_MODEL	Override the xAI summary model
ENABLE_WEB_SEARCH=true	Explicit gate for the snapshot refresh script

Quality gates

npm run lint
npm run build

The weekly snapshot workflow opens a reviewable PR only when ANTHROPIC_API_KEY is configured. Without that secret it exits cleanly with an explicit skip instead of producing a red scheduled run.

Snapshot refresh

Slow-moving AI-infrastructure data lives in lib/aiSnapshot.json. Each value carries an asOf date, source, and confidence. The refresh script re-verifies that snapshot and supports a dry run:

ENABLE_WEB_SEARCH=true npm run refresh:snapshots -- --dry-run

Writing a refreshed snapshot additionally requires the server-side Anthropic key. Generated changes are reviewed as a diff before they become the new baseline.

VPS deployment

docker compose up -d --build
curl http://127.0.0.1:3001/api/health

The production image uses Next.js standalone output and exposes container port 3000; the included Compose file maps it to host port 3001 by default.

Limitations

• This is an analytical monitor, not investment advice or an execution system.
• Source calendars differ, so freshness is shown per metric rather than hidden behind a single “live” label.
• Provider failures degrade individual overlays; they should not invalidate dated snapshots.
• AI summaries can be wrong and are never treated as the source of record.