Crucix

Your own intelligence terminal. 27 sources. One command. Zero cloud.

Visit The Live Site: crucix.live

Enter The Signal Network

More screenshots

Boot Sequence	World Map

3D Globe View

Live website: https://www.crucix.live/ Explore the public demo first, then clone the repo to run Crucix locally.

Crucix pulls satellite fire detection, flight tracking, radiation monitoring, satellite constellation tracking, economic indicators, live market prices, conflict data, sanctions lists, and social sentiment from 27 open-source intelligence feeds — in parallel, every 15 minutes — and renders everything on a single self-contained Jarvis-style dashboard.

Hook it up to an LLM and it becomes a two-way intelligence assistant — pushing multi-tier alerts to Telegram and Discord when something meaningful changes, responding to commands like /brief and /sweep from your phone, and generating actionable trade ideas grounded in real cross-domain data. Your own analyst that watches the world while you sleep.

Try the live demo first at https://www.crucix.live/, then clone the repo when you want the full local stack.

No cloud. No telemetry. No subscriptions. Just node server.mjs and you're running.

Most of the world's real-time intelligence — satellite imagery, radiation levels, conflict events, economic indicators, flight tracking, maritime activity — is publicly available. It's just scattered across dozens of government APIs, research institutions, and open data feeds that nobody has time to check individually.

Crucix brings it all into one place. Not behind a paywall, not locked in an enterprise platform, not requiring a security clearance. Just open data, aggregated and cross-correlated on your own machine, updated every 15 minutes.

It was built for anyone who wants to understand what's actually happening in the world right now — researchers, journalists, traders, OSINT analysts, or just curious people who believe access to information shouldn't depend on your budget.

# 1. Clone the repo
git clone https://github.com/calesthio/Crucix.git
cd Crucix

# 2. Install dependencies (just Express)
npm install

# 3. Copy env template and add your API keys (see below)
cp .env.example .env

# 4. Start the dashboard
npm run dev

If npm run dev fails silently (exits with no output), run Node directly instead:

node --trace-warnings server.mjs

This bypasses npm's script runner, which can swallow errors on some systems (particularly PowerShell on Windows). You can also run node diag.mjs to diagnose the exact issue — it checks your Node version, tests each module import individually, and verifies port availability. See Troubleshooting for more.

The dashboard opens automatically at http://localhost:3117 and immediately begins its first intelligence sweep. This initial sweep queries all 27 sources in parallel and typically takes 30–60 seconds — the dashboard will appear empty until the sweep completes and pushes the first data update. After that, it auto-refreshes every 15 minutes via SSE (Server-Sent Events). No manual page refresh needed.

Requirements: Node.js 22+ (uses native fetch, top-level await, ESM)

git clone https://github.com/calesthio/Crucix.git
cd Crucix
cp .env.example .env    # add your API keys
docker compose up -d

Dashboard at http://localhost:3117. Sweep data persists in ./runs/ via volume mount. Includes a health check endpoint.

A self-contained Jarvis-style HUD with:

• 3D WebGL globe (Globe.gl) with atmosphere glow, star field, and smooth rotation — plus a classic flat map toggle
• 9 marker types across both views: fire detections, air traffic, radiation sites, maritime chokepoints, SDR receivers, OSINT events, health alerts, geolocated news, conflict events
• Animated 3D flight corridor arcs between air traffic hotspots and global hubs
• Region filters (World, Americas, Europe, Middle East, Asia Pacific, Africa) — rotates the globe or zooms the flat map
• Live market data — indexes, crypto, energy, commodities via Yahoo Finance (no API key needed)
• Risk gauges — VIX, high-yield spread, supply chain pressure index
• OSINT feed — English-language posts from 17 Telegram intelligence channels (expandable)
• News ticker — merged RSS + GDELT headlines + Telegram posts, auto-scrolling
• Sweep delta — live panel showing what changed since last sweep (new signals, escalations, de-escalations with severity)
• Cross-source signals — correlated intelligence across satellite, economic, conflict, and social domains
• Nuclear watch — real-time radiation readings from Safecast + EPA RadNet
• Space watch — CelesTrak satellite tracking: recent launches, ISS, military constellations, Starlink/OneWeb counts
• Leverageable ideas — AI-generated trade ideas (with LLM) or signal-correlated ideas (without)

The VISUALS FULL / VISUALS LITE button in the top bar only changes rendering behavior - it does not remove data sources or reduce sweep coverage.

When you switch to VISUALS LITE, the dashboard:

• Disables decorative background effects such as the radial/grid overlays and scanlines
• Removes expensive blur/backdrop-filter effects on panels and overlays
• Stops non-essential animations like the logo ring blink, conflict rings, and corridor flow effects
• Disables globe auto-rotation and turns off animated flight-arc dashes
• Converts the horizontal news ticker and OSINT stream into static, scrollable lists instead of continuously animated marquees

Mobile-specific behavior:

• On mobile, VISUALS LITE also forces the dashboard into flat map mode if you are currently on the globe
• Future mobile loads will continue to start flat while low-perf mode is enabled

The preference is saved in browser local storage, so the UI will remember your last setting.

The server runs a sweep cycle every 15 minutes (configurable). Each cycle:

1. Queries all 27 sources in parallel (~30s)
2. Synthesizes raw data into dashboard format
3. Computes delta from previous run (what changed, escalated, de-escalated) — visible in the Sweep Delta panel on the dashboard
4. Generates LLM trade ideas (if configured)
5. Evaluates breaking news alerts — multi-tier (FLASH / PRIORITY / ROUTINE) with semantic dedup. Sends to Telegram and/or Discord if configured. Works with LLM evaluation or falls back to rule-based alerting when LLM is unavailable.
6. Pushes update to all connected browsers via SSE

Crucix doubles as an interactive Telegram bot. Beyond sending alerts, it responds to commands directly from your chat:

This requires TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID in .env. The bot polls for messages every 5 seconds (configurable via TELEGRAM_POLL_INTERVAL).

Crucix also supports Discord as a full-featured bot with slash commands and rich embed alerts. It mirrors the Telegram bot's capabilities with Discord-native formatting.

Alerts are delivered as rich embeds with color-coded sidebars: red for FLASH, yellow for PRIORITY, blue for ROUTINE. Each embed includes signal details, confidence scores, and cross-domain correlations.

Setup requires: DISCORD_BOT_TOKEN, DISCORD_CHANNEL_ID, and optionally DISCORD_GUILD_ID for instant slash command registration. See API Keys Setup for details.

Webhook fallback: If you don't want to run a full bot, set DISCORD_WEBHOOK_URL instead. This enables one-way alerts (no slash commands) with zero dependencies — no discord.js needed.

Optional dependency: The full bot requires discord.js. Install it with npm install discord.js. If it's not installed, Crucix automatically falls back to webhook-only mode.

Connect any of 6 LLM providers for enhanced analysis:

• AI trade ideas — quantitative analyst producing 5-8 actionable ideas citing specific data
• Smarter alert evaluation — LLM classifies signals into FLASH/PRIORITY/ROUTINE tiers with cross-domain correlation and confidence scoring
• Providers: Anthropic Claude, OpenAI, Google Gemini, OpenRouter (Unified API), OpenAI Codex (ChatGPT subscription), MiniMax, Mistral
• Graceful fallback — when LLM is unavailable, a rule-based engine takes over alert evaluation. LLM failures never crash the sweep cycle.

Copy .env.example to .env at the project root:

cp .env.example .env

These three unlock the most valuable economic and satellite data. Each takes about 60 seconds to register.

Set LLM_PROVIDER to one of: anthropic, openai, gemini, codex, openrouter, minimax, mistral

For Codex, run npx @openai/codex login to authenticate via your ChatGPT subscription.

Discord bot setup:

1. Go to Discord Developer Portal and create a new application
2. Go to Bot → click Reset Token → copy the token to DISCORD_BOT_TOKEN
3. Under Privileged Gateway Intents, enable Message Content Intent
4. Go to OAuth2 → URL Generator → select bot + applications.commands scopes → select Send Messages + Embed Links permissions
5. Copy the generated URL and open it in your browser to invite the bot to your server
6. Install the dependency: npm install discord.js

Alerts work with or without an LLM on both Telegram and Discord. With an LLM configured, signal evaluation is richer and more context-aware. Without one, a deterministic rule engine evaluates signals based on severity, cross-domain correlation, and signal counts.

Crucix still works with zero API keys. 18+ sources require no authentication at all. Sources that need keys return structured errors and the rest of the sweep continues normally.

crucix/
├── server.mjs                 # Express dev server (SSE, auto-refresh, LLM, bot commands)
├── crucix.config.mjs          # Configuration with env var overrides + delta thresholds
├── diag.mjs                   # Diagnostic script — run if server fails to start
├── .env.example               # All documented env vars
├── package.json               # Runtime: express | Optional: discord.js
├── docs/                      # Screenshots for README
│
├── apis/
│   ├── briefing.mjs           # Master orchestrator — runs all 27 sources in parallel
│   ├── save-briefing.mjs      # CLI: save timestamped + latest.json
│   ├── BRIEFING_PROMPT.md     # Intelligence synthesis protocol
│   ├── BRIEFING_TEMPLATE.md   # Briefing output structure
│   ├── utils/
│   │   ├── fetch.mjs          # safeFetch() — timeout, retries, abort, auto-JSON
│   │   └── env.mjs            # .env loader (no dotenv dependency)
│   └── sources/               # 27 self-contained source modules
│       ├── gdelt.mjs          # Each exports briefing() → structured data
│       ├── fred.mjs           # Can run standalone: node apis/sources/fred.mjs
│       ├── space.mjs          # CelesTrak satellite tracking
│       ├── yfinance.mjs       # Yahoo Finance — free live market data
│       └── ...                # 23 more
│
├── dashboard/
│   ├── inject.mjs             # Data synthesis + standalone HTML injection
│   └── public/
│       └── jarvis.html        # Self-contained Jarvis HUD
│
├── lib/
│   ├── llm/                   # LLM abstraction (5 providers, raw fetch, no SDKs)
│   │   ├── provider.mjs       # Base class
│   │   ├── anthropic.mjs      # Claude
│   │   ├── openai.mjs         # GPT
│   │   ├── gemini.mjs         # Gemini
│   │   ├── openrouter.mjs     # OpenRouter (Unified API)
│   │   ├── codex.mjs          # Codex (ChatGPT subscription)
│   │   ├── minimax.mjs        # MiniMax (M2.5, 204K context)
│   │   ├── mistral.mjs        # Mistral AI
│   │   ├── ideas.mjs          # LLM-powered trade idea generation
│   │   └── index.mjs          # Factory: createLLMProvider()
│   ├── delta/                 # Change tracking between sweeps
│   │   ├── engine.mjs         # Delta computation — semantic dedup, configurable thresholds, severity scoring
│   │   ├── memory.mjs         # Hot memory (3 runs, atomic writes) + cold storage (daily archives)
│   │   └── index.mjs          # Re-exports
│   └── alerts/
│       ├── telegram.mjs       # Multi-tier alerts (FLASH/PRIORITY/ROUTINE) + two-way bot commands
│       └── discord.mjs        # Discord bot (slash commands, rich embeds) + webhook fallback
│
└── runs/                      # Runtime data (gitignored)
    ├── latest.json            # Most recent sweep output
    └── memory/                # Delta memory (hot.json + cold/YYYY-MM-DD.json)

• Pure ESM — every file is .mjs with explicit imports
• Minimal dependencies — Express is the only runtime dependency. discord.js is optional (for Discord bot). LLM providers use raw fetch(), no SDKs.
• Parallel execution — Promise.allSettled() fires all 27 sources simultaneously
• Graceful degradation — missing keys produce errors, not crashes. LLM failures don't kill sweeps.
• Each source is standalone — run node apis/sources/gdelt.mjs to test any source independently
• Self-contained dashboard — the HTML file works with or without the server

All settings are in .env with sensible defaults:

Delta engine thresholds (how sensitive the system is to changes between sweeps) can be customized in crucix.config.mjs under the delta.thresholds section. The defaults are tuned to filter out noise while catching meaningful moves.

When running npm run dev:

This is a known issue where npm's script runner can swallow errors, particularly on Windows PowerShell. Try these in order:

1. Run Node directly (bypasses npm):

node --trace-warnings server.mjs

This is functionally identical to npm run dev but gives you full error output.

2. Run the diagnostic script:

node diag.mjs

This tests every import one by one, checks your Node.js version, and verifies port 3117 is available. It will tell you exactly what's failing.

3. Check if port 3117 is already in use:

A previous Crucix instance may still be running in the background.

# Windows PowerShell
netstat -ano | findstr 3117
taskkill /F /PID <the_PID_from_above>

# Or kill all Node processes
taskkill /F /IM node.exe

# macOS / Linux
lsof -ti:3117 | xargs kill

Then try starting again. You can also change the port by setting PORT=3118 in your .env file.

4. Check Node.js version:

node --version

Crucix requires Node.js 22 or later. If you have an older version, download the latest LTS from nodejs.org.

This is normal — the first sweep takes 30–60 seconds to query all 27 sources. The dashboard will populate automatically once the sweep completes. Check the terminal for sweep progress logs.

Expected behavior. Sources that require API keys will return structured errors if the key isn't set. The rest of the sweep continues normally. Check the Source Integrity section in the dashboard (or the server logs) to see which sources failed and why. The 3 most impactful free keys to add are FRED_API_KEY, FIRMS_MAP_KEY, and EIA_API_KEY.

OpenSky can also return HTTP 429 when its public hotspots are queried too aggressively. Crucix does not try to evade that limit. Instead, it surfaces the throttle/error in source health and preserves the most recent non-empty air traffic snapshot from runs/ so the dashboard flight layer does not suddenly go blank on a throttled sweep.

Make sure both TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID are set in .env. The bot only responds to messages from the configured chat ID (security measure). You should see [Crucix] Telegram alerts enabled and [Crucix] Bot command polling started in the server logs on startup. If not, double-check your token with curl https://api.telegram.org/bot<YOUR_TOKEN>/getMe.

Check these in order:

1. Make sure DISCORD_BOT_TOKEN and DISCORD_CHANNEL_ID are set in .env
2. Verify discord.js is installed: npm ls discord.js. If missing, run npm install discord.js
3. If slash commands don't appear, set DISCORD_GUILD_ID — without it, global commands can take up to 1 hour to propagate. Guild-specific commands register instantly
4. Confirm the bot was invited with bot + applications.commands scopes and has Send Messages + Embed Links permissions in the target channel
5. Check server logs for [Discord] Bot logged in as ... on startup. If you see [Discord] discord.js not installed, install it and restart
6. Webhook-only fallback: If you just want alerts without slash commands, set DISCORD_WEBHOOK_URL instead of the bot token. No discord.js needed.

The docs/ folder contains dashboard screenshots referenced by this README:

To update them: run the dashboard, wait for a sweep to complete, then use your browser's DevTools (F12 → Ctrl+Shift+P → "Capture full size screenshot") or a tool like LICEcap for GIFs.

Found a bug? Want to add a 28th source? PRs welcome. Each source is a standalone module in apis/sources/ — just export a briefing() function that returns structured data and add it to the orchestrator in apis/briefing.mjs.

If you find this useful, a star helps others find it too.

For contribution guidelines, review expectations, and source-add rules, see CONTRIBUTING.md. For security reports, see SECURITY.md.

For partnerships, integrations, or other non-issue inquiries, you can reach me at [email protected].

For bugs and feature requests, please use GitHub Issues so discussion stays visible and actionable.

AGPL-3.0