AGENT00

Dashboard

Loading…

–

Ask AGENT00

soul prompt + memory + all agents active

← Type a question and hit Ask.

Open To-Dos

–

Today's Events

–

H100 Hrs (MTD)

–

API Cost (MTD)

–

Memory Entries

–

Soul prompt active

▦ Agenda — Upcoming

Loading…

■ To-Do

H H100.ST — NAV

Loading…

◷ H100 Advisory

◈ Memory

⚙ System

Bot

–

PID

–

DB Size

–

API Today

–

API MTD

–

Memory

–

Ask Anything

Full conversation with soul prompt + memory

Ask AGENT00

soul prompt + memory active

← Ask a question above.

Trading Intelligence

–

☀ Daily Brief

Click ▶ Run to generate daily brief.

▶ Multi-Asset Brief

Click ▶ Run Brief to generate analysis across all assets.

Performance Stats

Click refresh to load performance stats.

📡 TradingView Alerts

TradingView Webhook Setup

1. In TradingView, create an alert on any chart
2. Under Notifications, enable Webhook URL
3. Paste your webhook URL: https://honeybees-mac-mini.tail56e8bc.ts.net/api/tv/webhook
4. Set the Message to JSON format:

{"ticker":"{ {ticker} }","action":"{ {strategy.order.action} }","price":"{ {close} }","interval":"{ {interval} }","message":"{ {strategy.order.comment} }"}

5. For simple alerts use: {"ticker":"{ {ticker} }","action":"alert","price":"{ {close} }","interval":"{ {interval} }","message":"Your alert name"}

No alerts yet — set up webhook in TradingView to receive real-time signals.

ICT Setups — Today

Signal Monitor — Live Indicators by Asset & Timeframe

Click refresh to load…

Asset Analysis

LLM · live data

Select an asset above to run analysis…

Signal Performance — Last 30 Days

Click refresh to load…

Key S/R Levels — All Assets

Click refresh to load key levels…

Market Data

Global Markets · Indices · FX · Commodities

Loading market data…

⏳ Loading ETF data…

Loading…

📰 H100 Headlines

Loading headlines…

Quick Snapshot →

Select an asset above

Bitcoin Intelligence

AGENT03 · H100.ST · Peer Analysis · Research

Bitcoin · BTC/USD · CoinGecko

–

Loading…

24H High

–

24H Low

–

24H Volume (USD)

–

Market Cap

–

Circulating Supply

–

All-Time High

–

24H Range –

Memory Bank

Loading…

By Category

To-Do

–

Agenda

Next 7 days

H100 Advisory

Hours by Category

–

Expenses by Category

–

Hours Logged

Expenses

Contacts

Network · Relationships

API Usage

OpenRouter · loading…

Today

–

MTD

–

All Time

–

Calls MTD

–

By Model — MTD

Loading…

Daily Spend — Last 14 Days

Loading…

Life Sciences & Healthcare

AGENT04 · Oxford · Etcembly · VX Pharma · Neuro

Loading…

All Recent News

Loading…

🎓 Oxford Ecosystem

Loading…

🧬 Etcembly / AI Drug Design

Loading…

🌍 VX Pharma / Africa

Loading…

🧠 Neuroscience / Surgery

Loading…

Company Watchlist

Loading…

Deal Flow Pipeline

Loading…

Developer

Build reference · v197

System Architecture

📱 Telegram
@agent00X_bot

⟷

⬡ AGENT00
Orchestrator

⟷

🌐 Web Dashboard
Flask :8080

Sub-Agents

🕐 AGENT01
Time Management

📊 AGENT02
Trading Intelligence

₿ AGENT03
Bitcoin Intelligence

🔬 AGENT04
Life Sciences

📡 AGENT05
Market Data

🧠 Memory
Soul + KB

🗄 SQLite
agent00.db

Files: AGENT00.py · AGENT01_time/flow/state.py · AGENT02_trading.py · AGENT03_btc/pdf.py · AGENT04_lifesci.py · AGENT00_web/db/memory/system.py

External APIs

📈 yfinance
OHLCV · 15m delay

🤖 OpenRouter
LLM · gpt-4.1-mini

🌤 Open-Meteo
Weather · London

🔍 Brave Search
News · LS digest

📰 NewsAPI
Headlines

Stack

Runtime: Python 3.9 · Mac Mini (honeybees-Mini) · always-on
Web: Flask · port 8080 · single-file · vanilla JS · dark theme
Database: SQLite — agent00.db
DB tables: setup_alerts · trade_results · eod_summaries · memory · todos · agenda · h100_hours · h100_expenses · event_log · ls_watchlist · ls_deals · ls_news_cache · ls_digests · contacts · contact_tags
LLM: OpenRouter — openai/gpt-4.1-mini (configurable)
Process mgmt: launchd — com.agent00.plist
Data: yfinance (15-min delayed) · Open-Meteo weather

Files

AGENT00.py — orchestrator, Telegram, command router
AGENT01_time.py — AGENT01 Time Management · hours & expenses
AGENT01_flow.py — AGENT01 guided Telegram entry flows
AGENT01_state.py — AGENT01 conversation state persistence
AGENT02_trading.py — AGENT02 Trading Intelligence
AGENT03_btc.py — AGENT03 Bitcoin Intelligence
AGENT03_pdf.py — AGENT03 institutional PDF reports
AGENT04_lifesci.py — AGENT04 Life Sciences & Healthcare
AGENT00_memory.py — soul prompt + memory bank
AGENT00_db.py — SQLite read/write layer
AGENT00_system.py — health, backups, cost tracking
AGENT00_web.py — this dashboard
restart.sh — clean restart script
.env — API keys (never committed)

All Agents

AGENT00 — Orchestrator

Main entry point · boots all sub-agents · wires fill checker thread
Telegram bot listener · inline keyboard menu system · command router (/start /menu /brief /nav /ask /memory /todo /hours)
Dispatches to AGENT01 · AGENT02 · AGENT03 · AGENT04 based on user intent
Manages conversation state · handles errors · sends Telegram alerts

AGENT01 — H100 Advisory (Time Management)

Logs advisory hours and expenses for H100 Advisory engagement to SQLite
Step-by-step guided Telegram entry flows · people tagging · category breakdown
Bulk importer · monthly summaries · Time Management web view · PNG export
Files: AGENT01_time.py · AGENT01_flow.py · AGENT01_state.py

AGENT02 — Trading Intelligence

See Trading Engine tab for full detail
Multi-asset: BTC · GC · MNQ · ES · CL · scheduled sessions · fill checker · EOD · memory writes
File: AGENT02_trading.py

AGENT03 — Bitcoin Intelligence

H100.ST NAV calculation — BTC/SEK price, holdings (1,051.05 BTC), debt, cash, FD shares
Computes: NAV/share · mNAV (market/NAV) · EMNAV · premium/discount to NAV
Peer analysis: MSTR · Metaplanet · BSTR · Capital B · Sovereign/SWF
Institutional PDF report generator · Latest News · Daily Brief · EU/Swiss Thesis
M&A targets · Strategy analysis (MSTR model · ATM mechanics · Credit instruments)
Files: AGENT03_btc.py · AGENT03_pdf.py

AGENT04 — Life Sciences & Healthcare

Four pillars: Oxford Ecosystem · Etcembly/AI Drug Design · VX Pharma/Africa · Neuro/Surgery
Brave Search API scans per pillar daily — news cached in SQLite ls_news_cache
Deal flow pipeline: Board → Invested → Due Diligence → Meeting → Researching → Watching → Passed
Scheduled 07:30 digest → Telegram · LLM summaries injected into Ask Anything context
File: AGENT04_lifesci.py

AGENT05 — Market Data [ planned ]

Live price feeds: Gold · SPX · UST 10Y · VIX · Bitcoin · DXY
Currently served via yfinance snapshots in AGENT00_web.py
Planned: dedicated agent with caching, alerts, macro commentary

Support Modules

AGENT00_memory.py — soul prompt, key/value memory, evening digest, context injection into all LLM calls
AGENT00_db.py — all SQLite read/write operations centralised
AGENT00_system.py — health monitoring, API cost tracking, backups, error alerting

AGENT02 — Trading Intelligence

Assets: BTC · MGC (Gold) · MNQ · MES · MCL — all micro contracts
Timeframes: 4H · 1H · 15M per asset via yfinance
EMA: 9 / 21 / 50 / 200 — stack alignment, 9/21 crossover detection
VWAP: session VWAP, above/below bias, reclaim/rejection signals
RSI(14): state labelled — OVERBOUGHT / BULLISH / NEUTRAL / BEARISH / OVERSOLD
MACD: line, signal, histogram, trend state (BULLISH_MOMENTUM / FADING / BEARISH_MOMENTUM)
ATR(14): stop = 1.5×ATR · target = 3×ATR · auto-scales per volatility
Bollinger Bands: squeeze detection · expansion with EMA alignment = cleaner trend
Direction filter: LONG_ONLY / SHORT_ONLY / NO_TRADE — hard gate on signal generation
Signal quality score: 0–10 composite based on indicator alignment

Contract Assumptions

Asset	Contract	Full Name	P&L / pt	Price Feed
MNQ	Micro E-mini Nasdaq-100	Micro NQ	$0.50	NQ=F
MES	Micro E-mini S&P 500	Micro ES	$5.00	ES=F
MGC	Micro Gold Futures	Micro GC	$10.00	GC=F
MCL	Micro Crude Oil	Micro CL	$100.00	CL=F
BTC	Bitcoin spot	$10k fixed position	% return	BTC-USD

Price feeds use full contract symbols (NQ=F / ES=F / GC=F / CL=F) for OHLCV data — P&L is scaled to micro contract sizing. BTC P&L = (pts / entry) × $10,000.

ICT Methodology

FVG — Fair Value Gaps: imbalance zones left by impulsive moves
Order Blocks — last candle before impulsive move
Liquidity Sweeps — stop hunts above highs / below lows
BOS/CHoCH — Break of Structure / Change of Character
Premium/Discount — above/below 50% of HTF range
Three tiers: HTF (4H/Daily, R:R ≥2.0) · LTF (1H/15M, R:R ≥1.5) · SCALP (15M, R:R ≥1.0)
Highest prob: FVG + VWAP + EMA pullback + MACD confirm

Automation & Schedule

06:00 UK — comprehensive brief all assets
08:00 UK — London open signals
13:00 UK — pre-NY signals
21:00 UK — EOD P&L resolution + memory write
Fill checker: background thread every 2 min · Mon–Fri 07:00–22:00 UK
Checks PENDING/OPEN alerts vs live price (0.1% tolerance)
Telegram: 🟢 Entry hit · ✅ Target hit · 🛑 Stop hit

Learning Feedback Loop

Before generating signals: pulls last 30 days trade results per asset
Injects structured performance summary into LLM prompt
LLM sees: win rate by tier, by session, by direction · avg win/loss · trigger rate · R:R quality
Auto-generates instructions: avoid low-performing tiers, favour winning sessions
After EOD: writes 5 memory entries — daily P&L, per-asset learning, win/loss patterns, trigger rate
Memory entries feed into Ask Anything context
Note: scalp signals less reliable due to 15-min data delay

How Memory Works

Storage: SQLite memory table · key / value / category / timestamp
Soul prompt: Hard-coded standing context — professional background, BTC thesis, H100 context, operating principles
Dynamic entries: User-saved views, market context, people, decisions
Trading entries: Auto-written nightly — EOD P&L, per-asset 30-day stats, win/loss patterns, trigger rate
Context injection: Soul + all memory entries injected into every LLM call (Ask Anything + ICT analysis)
Evening digest: Telegram summary of all memory entries on demand

Memory Key Structure

VIEWS:* — saved market views and opinions
TRADING:EOD:YYYY-MM-DD — daily trade summary
TRADING:PERFORMANCE:BTC — 30-day stats per asset
TRADING:LOSING_PATTERNS — which tier/asset loses most
TRADING:WINNING_PATTERNS — which tier/asset wins most
TRADING:TRIGGER_RATE — entry precision metric
Categories: general · market · trading · people · decisions

Bot Config

Bot: @agent00X_bot
Chat ID: 542250707 · single user · private
Token: stored in .env — never committed

Commands & Menu

/start · /menu — inline keyboard main menu
/brief — run ICT brief for current session
/nav — H100 NAV snapshot
/ask [question] — Ask Anything via Telegram
/memory — view memory bank
/todo — view/add todos
/hours — H100 hours summary

Automated Alerts

06:00 — comprehensive ICT brief · all assets · all timeframes
08:00 — London open signals · BTC/GC/MNQ/ES/CL
13:00 — pre-NY session signals
21:00 — EOD P&L summary · trade resolution · memory write
Fill checker (every 2 min): 🟢 ENTRY HIT — asset/dir/entry/stop/target/potential P&L
Fill checker: ✅ TARGET HIT — actual P&L in $
Fill checker: 🛑 STOP HIT — actual loss in $
Deduplication via in-memory set — no duplicate alerts per trade event

Security Model

Network: Dashboard LAN-only (192.168.0.135:8080) · no public port forwarding · no authentication layer currently
Secrets: All API keys in .env · .gitignore excludes .env and agent00.db
Telegram: Single Chat ID whitelist · all messages checked against TELEGRAM_CHAT_ID before processing
GitHub: Private repo · no secrets ever committed · DB excluded
LLM: OpenRouter — no raw user data stored · prompts contain no PII beyond user-saved memory

Recommendations

✅ Add Tailscale for secure remote access (see Network tab)
⚠️ Dashboard has no login — anyone on LAN can access it
⚠️ Consider adding HTTP basic auth or token to Flask if sharing WiFi
⚠️ Restart button on dashboard is unprotected — would allow process kill from LAN
📋 TODO: add Tailscale · consider basic auth header for restart endpoint

Health Checks

launchd: auto-restarts AGENT00.py on crash · com.agent00.plist
/api/system: returns PID · DB exists + size · API cost today/MTD · last events
event_log table: every agent action logged with type + detail + timestamp
Web header: ● Online / ● Offline dot — polls /api/system every 30s
DB size monitoring: displayed in System Status panel + right rail
API cost tracking: today + MTD + all-time via OpenRouter auth endpoint
Log file: agent00.log — last 100 lines accessible in System & Logs view

Restart & Recovery

restart.sh — unload launchd → kill python → git pull → reload → confirm PID
Web dashboard restart button → POST /api/system/restart → runs restart.sh
Web dashboard runs separately — survives bot restart
SQLite is single-file — backed up by copying agent00.db
Git is the code backup — every version pushed to GitHub

Current Network Setup

Mac Mini: Ethernet → 192.168.0.135 (static DHCP recommended)
PC: Ethernet → 192.168.0.xxx
Laptop: WiFi — should reach 192.168.0.135:8080 if router allows cross-interface traffic
Router: Virgin Media Hub · admin at 192.168.0.1
⚠️ If laptop can't reach dashboard on WiFi: check AP Isolation setting in router — disable it

Remote Access — Tailscale (Recommended)

Zero-config encrypted mesh VPN · free tier covers this use case
Mac Mini gets a stable private IP e.g. 100.x.x.x
Access dashboard from anywhere: http://100.x.x.x:8080
Works through hotel WiFi, mobile hotspot, firewalls
No port forwarding · nothing exposed publicly
Install Mac Mini: brew install tailscale && tailscale up
Install laptop: download from tailscale.com · login same account

Other Options

Cloudflare Tunnel — encrypted tunnel → HTTPS URL like agent00.yourdomain.com · no open ports · more setup
Port forwarding — NOT recommended · exposes dashboard publicly with no auth
Static DHCP: reserve 192.168.0.135 for Mac Mini MAC address in Virgin Media router so IP never changes

📋 Network TODO

☐ Set static DHCP reservation for Mac Mini → 192.168.0.135 in Virgin Media router (192.168.0.1)
☐ Install Tailscale on Mac Mini: brew install tailscale && sudo tailscaled & tailscale up
☐ Install Tailscale on laptop — login same account → dashboard accessible from anywhere
☐ Test laptop WiFi access: disable AP Isolation in router if needed
☐ Consider adding basic auth to /api/system/restart endpoint

In Progress / Near Term

✅ Daily Brief + Multi-Asset Brief panels on Trading Intelligence overview
✅ Pivot Points added to Signal Monitor with confluence alerts
✅ Performance Stats tab — daily/weekly/monthly P&L tracking
✅ Signal Monitor footnotes — explanatory notes for all indicators
☐ Investigate contacts save issue — new contacts not persisting
☐ Investigate To-Do edit save — changes not persisting
☐ Signal backtester — run logic against 6-12 months OHLCV · score per setup type
☐ Signal quality filter threshold — once 4-6 weeks data · set permanent min score
☐ Telegram filter — only ping on 7+ quality score signals
✅ PC deployment fixed — git config core.autocrlf false · .gitattributes · push --force
☐ Tailscale remote access setup (see Network tab)
✅ launchd for web dashboard auto-start — both services running via launchd
☐ Static DHCP reservation — Mac Mini IP
☐ H100 live price — Nordnet auth required · Playwright headless login fix pending

Phase 2 / 3

AGENT02 — Signal backtester · study & learning module
AGENT03 — mNAV alerts · peer comparison (MSTR, Metaplanet, BSTR) · PDF reports live
AGENT04 — Deal flow CRM integration · automated pillar digests
AGENT05 — Market Data agent (dedicated) · macro alerts · price caching
H100 Advisory — invoicing module · contract tracker
Dashboard — Tailscale remote access · basic auth / login
✅ Contacts — standalone view with SQLite, tag filters, auto-migrate on startup

Dev To-Do

Loading…

📫 Telegram Alert Schedule

Managed in AGENT00.py · run_scheduler()

Loading...

Notes

Toggle and time changes are saved to the memory bank and read by the scheduler on each tick (every minute). To permanently change a schedule, edit run_scheduler() in AGENT00.py and restart.

Version History

v197 current

Loading…

System

Process · logs · quick actions

Status

Quick Actions

Recent Events (24h)

agent00.log (last 100 lines)