Crypto Trading Bot

A modular pipeline for crypto trading: data collection, feature engineering, CatBoost/LLM ensemble prediction, and live/simulated trading via Kraken. Built for research, simulation, and optional live trading.

Setup

Requirements: Python 3.12+ and a virtual environment.

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

On Debian Bookworm (Raspberry Pi), libta-lib-dev is usually unavailable. Install dependencies with pip install -r requirements-pi.txt first; if TA-Lib fails to build, compile the TA-Lib C library from source and retry.

The current low-RAM Raspberry Pi runtime profile in this repository has been tested on a Raspberry Pi 3B+ with 1 GB RAM.

Copy and edit the example config:

cp .env.example .env        # default / dry-run profile
# or
cp .env.live.example .env   # conservative live profile
# or
cp .env.small_account.example .env   # conservative small-account profile

Optional AutoResearch overlay for servers:

cp .env.live.example .env
cat .env.autoresearch.example >> .env

Optional AutoResearch overlay for real command execution (strict mode):

cp .env.live.example .env
cat .env.autoresearch.active.example >> .env

Quickstart (60 seconds)

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
python3 get_data.py
python3 data_preperation.py
python3 trading_bot.py --backtest

CI / Quality Gates

GitHub Actions pipeline in .github/workflows/ci.yml runs on push and pull requests with Python 3.12 and 3.13.

• Syntax gate: py_compile on tracked Python files
• Lint gate: ruff check with annotations in GitHub Checks
• Test gate: pytest with JUnit XML (pytest-report.xml) and artifact upload
• Explicit gate steps fail the job when lint/tests are not green

Weekly decision automation is separated in .github/workflows/weekly-scorecard.yml.

Project structure

File	Purpose
trading_bot.py	Main bot loop, backtester, risk guardrails
get_data.py	Fetch OHLCV data from Kraken
data_preperation.py	Feature engineering, training data export
predict.py	LLM ensemble inference (optional)
predict_catboost.py	CatBoost tabular inference
train_trading_model.py	Fine-tune the LLM (GPU required)
train_catboost_model.py	Train/walk-forward the CatBoost model
go_no_go_scorecard.py	Weekly Go/No-Go decision gate
analyze_trade_journal.py	Trade journal statistics
get_account_balance.py	Fetch live Kraken balance

Usage

Typed API with FastAPI

The repository now includes a typed API layer based on FastAPI and Pydantic.

Start it locally:

uvicorn fastapi_app:app --reload

Available endpoints:

• GET /health for a simple readiness check
• GET /research-signal to read and validate the current research JSON
• POST /research-signal/normalize to normalize arbitrary research payloads into the fixed feature schema
• POST /predict/catboost to run a typed single-row CatBoost inference via a strict, validated feature schema
• GET /scorecard to calculate the weekly Go/No-Go scorecard as a typed JSON response

Swagger UI is available at http://127.0.0.1:8000/docs.

1. Data collection

python3 get_data.py
python3 data_preperation.py

Outputs full_crypto_data.csv and training_data.csv.

2. Model training (optional)

Train the CatBoost model:

python3 train_catboost_model.py

Outputs a ready-to-use model in model/catboost_trading_model/.

Fine-tune the LLM (requires a GPU with ≥ 16 GB VRAM):

python3 train_trading_model.py

Outputs a LoRA adapter in model/fine_tuned_trading_model/.

Train CatBoost with AutoResearch features (optional):

python3 train_catboost_model.py --research-signal-path data/research_signal_latest.json

If omitted, the model uses neutral research defaults.

3. Dry-run / live trading

python3 trading_bot.py

• Set DRY_RUN=true (default) to simulate trades without spending real funds.
• Set SIMULATE_DATA=true to run fully offline without an exchange connection.
• Set SIMULATION_REGIME=uptrend|downtrend|sideways|crash|recovery|neutral|mixed to shape the offline market regime.
• Trades are recorded to trade_journal.csv.

Simulation helper scripts:

bash scripts/start_sim_bot.sh
bash scripts/stop_sim_bot.sh

Defaults:

• Uses .env.simulation.example
• Writes logs to logs/sim_bot.log
• Stores PID in logs/sim_bot.pid

Simulation regime comparison helper:

python3 scripts/compare_simulation_regimes.py --coins BTC,ETH,SOL,CHZ,VVV

This runs the bot's own market analysis against multiple synthetic regimes and prints the top recommendations side by side.

Pi wrappers for the comparison helper:

bash scripts/run_compare_simulation_regimes_on_pi.sh --coins BTC,ETH,SOL --regimes uptrend,crash,mixed --top 3
bash scripts/run_compare_simulation_standard_profiles_on_pi.sh

The first wrapper runs the comparison on the Raspberry Pi with the bot's service user and virtualenv. The second wrapper executes the common default profile set (uptrend,crash,mixed) for the standard coin basket. On the Pi itself, use PI_HOST=local to avoid an extra SSH hop.

4. Backtesting

python3 trading_bot.py --backtest

Prints a summary report (total return, max drawdown, number of trades).

5. Prediction (standalone)

python3 predict.py          # LLM + rule-based ensemble
python3 predict_catboost.py # CatBoost only

Configuration

All parameters are set via .env (see .env.example for the full list). Key variables:

Variable	Default	Description
EXCHANGE_NAME	kraken	Exchange (ccxt identifier)
BASE_CURRENCY	EUR	Quote currency
DRY_RUN	true	Simulate trades (no real orders)
SIMULATE_DATA	false	Use generated data instead of live API
SIMULATION_SEED	42	Deterministic seed for generated simulation data
SIMULATION_REGIME	neutral	Synthetic market regime: neutral, uptrend, downtrend, sideways, crash, recovery, mixed
EXCLUDED_COINS	USDC,USDT,EURT,DAI,TUSD,USDP,FDUSD,USDE	Always excluded symbols from live analysis
LOSSMAKER_EXCLUDED_COINS	ZEC,HYPE,TON,BTC,XRP	Additional default exclusions for symbols with persistently weak realized PnL
DYNAMIC_LOSSMAKER_EXCLUSION_ENABLED	true	Temporarily exclude symbols with repeated recent timeout-heavy losses in the journal
DYNAMIC_LOSSMAKER_WINDOW	80	Number of most recent journal rows inspected for temporary lossmaker exclusions
DYNAMIC_LOSSMAKER_MIN_SELLS	3	Minimum recent sells required before a symbol can be temporarily suppressed
DYNAMIC_LOSSMAKER_MIN_PNL_LOSS	0.003	Minimum negative summed realized PnL before the temporary suppression applies
COIN_MAX_HOLD_SECONDS	LINK:420,SUI:420,ONDO:420	Coin-specific hold-time overrides for weak follow-through symbols
TRADE_AMOUNT	10	EUR per trade
PORTFOLIO_TRADE_AMOUNT_MULTIPLIERS	50:1.10,100:1.25,200:1.50,300:1.75,500:2.20,1000:3.00	Scales the base TRADE_AMOUNT upward once the portfolio reaches each milestone
MAX_OPEN_TRADES	4	Maximum concurrent positions
PORTFOLIO_MAX_OPEN_TRADES_TIERS	100:3,200:4,500:5,1000:6	Raises the concurrent slot count in later portfolio stages without changing the early small-account regime
OHLCV_TIMEFRAME	1h	Candle interval
TICKER_BATCH_SIZE	80	Symbols per fetch_tickers request (smaller is more stable)
TICKER_FETCH_RETRIES	2	Retries per ticker batch on exchange/network errors
USE_TABULAR_MODEL	true	Enable CatBoost signal filter
TABULAR_MIN_CONFIDENCE	0.45	Minimum CatBoost confidence required for non-HOLD signal
UPTREND_ENTRY_MAX_RSI_BY_COIN	XDC:78,TRX:74	Coin-specific RSI override for uptrend entry gating
UPTREND_ENTRY_MIN_BUY_PROBA_BY_COIN	TRX:0.150	Coin-specific CatBoost buy-probability override for uptrend entry gating
UPTREND_ENTRY_MAX_SELL_PROBA_BY_COIN	ONDO:0.445	Coin-specific CatBoost sell-probability override for uptrend entry gating
UPTREND_ENTRY_MIN_PROBA_EDGE_BY_COIN	TRX:-0.13,ONDO:-0.055	Coin-specific probability-edge override for uptrend entry gating
TABULAR_RESEARCH_SIGNAL_PATH	./data/research_signal_latest.json	Latest AutoResearch JSON used as model input features
AUTO_TUNE_TABULAR_CONFIDENCE	true	Auto-adjust CatBoost threshold
USE_ML_MODEL	false	Enable LLM signal (GPU recommended)

Simulation regimes

When SIMULATE_DATA=true, the bot can emulate different market states while staying deterministic for a fixed SIMULATION_SEED.

• neutral: balanced oscillating market with moderate volume
• uptrend: rising prices with healthier trend persistence
• downtrend: falling prices with somewhat heavier defensive volume
• sideways: flat range-bound action with lower average activity
• crash: sharp downside move with distinctly higher volume
• recovery: selloff followed by rebound behavior
• mixed: assigns different regimes per coin so the same run contains both stronger and weaker assets

Quick examples:

SIMULATE_DATA=true SIMULATION_REGIME=crash python3 trading_bot.py --backtest
SIMULATE_DATA=true SIMULATION_REGIME=uptrend python3 scripts/compare_simulation_regimes.py --coins BTC,ETH,SOL
PI_HOST=local bash scripts/run_compare_simulation_standard_profiles_on_pi.sh

Default lossmaker exclusions

The runtime now merges EXCLUDED_COINS with a second default blocklist, LOSSMAKER_EXCLUDED_COINS.

• Base stablecoin-like exclusions stay in EXCLUDED_COINS
• PnL-driven exclusions currently default to ZEC,HYPE,TON,BTC,XRP
• Override LOSSMAKER_EXCLUDED_COINS in .env if you want to narrow or expand this list explicitly

Example:

LOSSMAKER_EXCLUDED_COINS=ZEC,HYPE,TON

Dynamic lossmaker throttling

The runtime can temporarily suppress fresh entries for coins that are currently producing repeated timeout-heavy losses in the recent journal window.

• The filter reads the latest DYNAMIC_LOSSMAKER_WINDOW journal rows
• A symbol is only suppressed when it has enough recent sells, negative summed PnL, weak win rate, and a high MAX-HOLD-TIME share
• Open positions are still analyzed and managed normally; the filter only blocks new candidate selection

Example override:

DYNAMIC_LOSSMAKER_EXCLUSION_ENABLED=true
DYNAMIC_LOSSMAKER_WINDOW=120
DYNAMIC_LOSSMAKER_MIN_SELLS=4
DYNAMIC_LOSSMAKER_MIN_PNL_LOSS=0.005
COIN_MAX_HOLD_SECONDS=LINK:420,SUI:420,ONDO:420

Portfolio step scaling

The runtime can grow position size and slot count in stages as the portfolio compounds.

• TRADE_AMOUNT remains the base amount for the smallest account stage
• PORTFOLIO_TRADE_AMOUNT_MULTIPLIERS scales that base amount once the portfolio crosses each milestone
• MAX_OPEN_TRADES remains the base slot count
• PORTFOLIO_MAX_OPEN_TRADES_TIERS only increases slots in later stages, so the small account stays concentrated

Example override:

TRADE_AMOUNT=10
PORTFOLIO_TRADE_AMOUNT_MULTIPLIERS=50:1.10,100:1.30,200:1.60,300:1.90,500:2.40,1000:3.20
MAX_OPEN_TRADES=2
PORTFOLIO_MAX_OPEN_TRADES_TIERS=100:3,250:4,500:5,1000:6

Coin-specific uptrend entry tuning

The uptrend entry gate can now be loosened selectively for specific coins instead of changing the global defaults for every symbol.

• UPTREND_ENTRY_MAX_RSI_BY_COIN is useful when one coin trends hotter before follow-through appears
• UPTREND_ENTRY_MIN_BUY_PROBA_BY_COIN is useful when a coin repeatedly misses entry by a small CatBoost buy-probability margin
• UPTREND_ENTRY_MAX_SELL_PROBA_BY_COIN is useful when one coin is consistently blocked by a slightly too-high sell probability

Example override:

UPTREND_ENTRY_MAX_RSI_BY_COIN=XDC:78,TRX:74
UPTREND_ENTRY_MIN_BUY_PROBA_BY_COIN=TRX:0.150
UPTREND_ENTRY_MAX_SELL_PROBA_BY_COIN=ONDO:0.445
UPTREND_ENTRY_MIN_PROBA_EDGE_BY_COIN=TRX:-0.13,ONDO:-0.055

If a coin already passes the uptrend quality gate but still misses fallback entries because the generic fallback RSI ceiling is too low, use FALLBACK_MAX_RSI_BY_COIN for a narrow override instead of lowering the global fallback limit.

FALLBACK_MAX_RSI=68
FALLBACK_MAX_RSI_BY_COIN=TRX:72

AutoResearch -> AI model features

The CatBoost model can consume a normalized AutoResearch signal vector at training and inference time.

Expected JSON file (example path: data/research_signal_latest.json):

{
  "timestamp_utc": "2026-04-18T09:00:00Z",
  "sentiment_score": 0.28,
  "confidence": 0.73,
  "risk_score": 0.31,
  "market_regime": "bull",
  "citations": [
    "https://example.com/source-1",
    "https://example.com/source-2"
  ]
}

Mapped model features:

• research_sentiment_score in [-1, 1]
• research_confidence in [0, 1]
• research_risk_score in [0, 1]
• research_regime_bull / research_regime_bear / research_regime_sideways

Implementation points:

• Shared normalization logic: research_signal.py
• Training integration: train_catboost_model.py
• Live inference integration: predict_catboost.py and trading_bot.py

AutoResearch bridge script

Use the bridge to generate/refresh the canonical signal file that the model consumes:

python3 scripts/update_autoresearch_signal.py \
  --command "<your-autoresearch-command>" \
  --source data/research_signal_raw.json \
  --output data/research_signal_latest.json

If your AutoResearch command writes directly to the output path, you can omit --source.

The bridge:

• Runs your AutoResearch command (optional)
• Validates payload age using timestamp_utc
• Normalizes values to model-safe ranges
• Writes canonical JSON used by training and inference

Fallback mode (recommended when repo/path is unknown):

python3 scripts/update_autoresearch_signal.py \
  --output data/research_signal_latest.json \
  --fallback-neutral

This writes a neutral signal (sentiment=0, confidence=0, regime=sideways) when AutoResearch is unavailable.

Weekly automation with AutoResearch

scripts/run_weekly_scorecard.sh can run the bridge before scoring.

Relevant environment variables:

• AUTORESEARCH_ENABLED=true|false (default false)
• AUTORESEARCH_REQUIRED=true|false fail closed when bridge fails (default false)
• AUTORESEARCH_CMD="..." command that executes AutoResearch
• AUTORESEARCH_SOURCE_PATH=... JSON produced by your AutoResearch run
• AUTORESEARCH_OUTPUT_PATH=... canonical model signal path (default data/research_signal_latest.json)
• AUTORESEARCH_MAX_AGE_MINUTES=180 freshness window
• AUTORESEARCH_ALLOW_STALE=true|false allow old payloads
• AUTORESEARCH_WRITE_NEUTRAL_FALLBACK=true|false write neutral signal when AutoResearch fails (default true)
• AUTORESEARCH_PRECHECK=true|false run setup validation before bridge (default true)
• AUTORESEARCH_PRECHECK_DRY_RUN=true|false run command/source and validate JSON contract (default false)
• AUTORESEARCH_PRECHECK_TIMEOUT_SEC=90 timeout for dry-run precheck execution
• In strict mode (AUTORESEARCH_REQUIRED=true), weekly automation forces dry-run precheck on.
• AUTORESEARCH_STRICT_ALLOW_MAINTENANCE_FALLBACK=true|false allow one-run fallback continuation after strict precheck failure (default false)
• AUTORESEARCH_MAINTENANCE_EXIT_CODE=4 dedicated exit code when maintenance override is active
• Pre-filled template: .env.autoresearch.example
• Strict execution template: .env.autoresearch.active.example

Manual precheck:

bash scripts/check_autoresearch_setup.sh

Manual precheck with dry-run validation:

AUTORESEARCH_ENABLED=true \
AUTORESEARCH_PRECHECK_DRY_RUN=true \
bash scripts/check_autoresearch_setup.sh

Maintenance override example (strict mode, but continue with forced fallback when precheck fails):

AUTORESEARCH_ENABLED=true \
AUTORESEARCH_REQUIRED=true \
AUTORESEARCH_STRICT_ALLOW_MAINTENANCE_FALLBACK=true \
bash scripts/run_weekly_scorecard.sh

Risk guardrails

Optional hard limits — set to 0 to disable:

Variable	Description
MAX_DAILY_LOSS_PCT	Block new BUYs once daily drawdown reaches this %
MAX_BUYS_PER_HOUR	Maximum BUY trades in the last 60 minutes
LOSS_STREAK_PAUSE_THRESHOLD	Consecutive losses before pausing entries
LOSS_STREAK_PAUSE_SECONDS	Pause duration after a loss streak

Weekly Go/No-Go scorecard

Run the automated decision gate:

bash scripts/run_weekly_scorecard.sh

Saves a timestamped report to results/scorecards/ and updates results/scorecards/latest_scorecard.txt. Also writes results/scorecards/latest_status.env (override via STATUS_FILE). Optionally writes results/scorecards/latest_status.json when STATUS_JSON_ENABLED=true. Optionally writes results/scorecards/latest_status.prom when STATUS_PROM_ENABLED=true. Exit codes: 0 = GO, 2 = HOLD, 3 = NO-GO. Missing, invalid, or empty journal inputs are treated as HOLD because they indicate insufficient usable data rather than a scorecard execution failure. When strict maintenance override is used (AUTORESEARCH_STRICT_ALLOW_MAINTENANCE_FALLBACK=true), weekly automation returns a dedicated degraded-mode code (default 4) for the run.

Machine-readable report markers are included for log parsers:

• RUN_MODE=standard|maintenance_override
• UNDERLYING_EXIT_CODE=<int> (only in maintenance override mode)
• FINAL_EXIT_CODE=<int>

latest_status.env fields:

• TIMESTAMP_UTC
• RUN_MODE
• VERDICT
• UNDERLYING_EXIT_CODE
• FINAL_EXIT_CODE
• REPORT_FILE
• LATEST_REPORT_LINK

latest_status.json fields (when enabled):

• timestamp_utc
• run_mode
• verdict
• underlying_exit_code
• final_exit_code
• report_file
• latest_report_link

latest_status.prom metrics (when enabled):

• trading_scorecard_final_exit_code
• trading_scorecard_underlying_exit_code
• trading_scorecard_timestamp_seconds
• trading_scorecard_run_mode{mode="..."}
• trading_scorecard_verdict{verdict="..."}

Monitoring extraction examples:

# Using grep (prints full key=value lines)
grep '^RUN_MODE=' results/scorecards/latest_scorecard.txt
grep '^FINAL_EXIT_CODE=' results/scorecards/latest_scorecard.txt

# Using awk (prints only values)
awk -F= '/^RUN_MODE=/{print $2}' results/scorecards/latest_scorecard.txt
awk -F= '/^FINAL_EXIT_CODE=/{print $2}' results/scorecards/latest_scorecard.txt

# Single-line status for cron/monitoring
awk -F= '/^RUN_MODE=/{m=$2} /^FINAL_EXIT_CODE=/{e=$2} END{printf "run_mode=%s exit_code=%s\n", m, e}' \
  results/scorecards/latest_scorecard.txt

# Read directly from latest_status.env
grep -E '^(RUN_MODE|VERDICT|FINAL_EXIT_CODE)=' results/scorecards/latest_status.env

# Source in shell scripts
set -a; . results/scorecards/latest_status.env; set +a
echo "mode=$RUN_MODE verdict=$VERDICT exit=$FINAL_EXIT_CODE"

# Enable and read JSON status
STATUS_JSON_ENABLED=true bash scripts/run_weekly_scorecard.sh
cat results/scorecards/latest_status.json

# Enable and read Prometheus textfile status
STATUS_PROM_ENABLED=true bash scripts/run_weekly_scorecard.sh
cat results/scorecards/latest_status.prom

Or run the scorecard manually with custom thresholds:

python3 go_no_go_scorecard.py --file trade_journal.csv \
  --lookback-days 7 --starting-capital 20 \
  --min-closed-trades 200 --min-win-rate 45 \
  --min-profit-factor 1.2 --max-drawdown-pct 10

Tests

.venv/bin/python -m pytest tests/ -v

Troubleshooting

• TA-Lib import/build issues:
  • Debian Bookworm (Raspberry Pi): libta-lib-dev is typically unavailable. Build TA-Lib C from source if pip install TA-Lib fails.
• CatBoost runtime issues on Linux CI:
  • Ensure libgomp1 is installed on the runner
• CI import errors like ModuleNotFoundError for project modules:
  • Run tests with PYTHONPATH set to the repository root
• Reproduce CI test command locally:
  • .venv/bin/python -m pytest tests/ -v --maxfail=1 --junitxml=pytest-report.xml

Public Release Safety

Before changing repository visibility from private to public:

• Rotate all exchange API keys used during development.
• Confirm .env remains untracked and ignored.
• Run a secret scan over tracked files.
• Verify local test env files are not staged.
• Use least-privilege API permissions (no withdrawals).

See SECURITY.md for the full policy and reporting guidance.

Community and Collaboration

• Code of Conduct: CODE_OF_CONDUCT.md
• Contributing Guide: CONTRIBUTING.md
• Issue templates: .github/ISSUE_TEMPLATE/
• Pull request template: .github/pull_request_template.md

Disclaimer

This project is for research and educational purposes. Use at your own risk.

Verdicts:

• GO (exit code 0)
• HOLD (exit code 2)
• NO-GO (exit code 3)
• MAINTENANCE-OVERRIDE (exit code 4 by default, configurable)

Weekly automation:

bash scripts/run_weekly_scorecard.sh

• Writes timestamped reports to results/scorecards/
• Updates results/scorecards/latest_scorecard.txt
• Updates results/scorecards/latest_status.env
• Optionally updates results/scorecards/latest_status.json when STATUS_JSON_ENABLED=true
• Optionally updates results/scorecards/latest_status.prom when STATUS_PROM_ENABLED=true
• Returns the scorecard exit code for automation
• Uses PYTHON_BIN if set, otherwise .venv/bin/python, otherwise python3

Server example:

PYTHON_BIN=./.venv/bin/python \
JOURNAL_FILE=./trade_journal.csv \
LOOKBACK_DAYS=7 \
MIN_CLOSED_TRADES=200 \
MIN_WIN_RATE=45 \
MIN_PROFIT_FACTOR=1.2 \
MAX_DRAWDOWN_PCT=10 \
bash scripts/run_weekly_scorecard.sh

Server example with AutoResearch defaults:

cp .env.live.example .env
cat .env.autoresearch.example >> .env

PYTHON_BIN=./.venv/bin/python \
AUTORESEARCH_ENABLED=true \
AUTORESEARCH_WRITE_NEUTRAL_FALLBACK=true \
bash scripts/run_weekly_scorecard.sh

Server example with real AutoResearch command (strict):

cp .env.live.example .env
cat .env.autoresearch.active.example >> .env

# then edit placeholders in .env:
# - AUTORESEARCH_REPO_PATH
# - AUTORESEARCH_CMD

PYTHON_BIN=./.venv/bin/python \
AUTORESEARCH_ENABLED=true \
AUTORESEARCH_REQUIRED=true \
bash scripts/run_weekly_scorecard.sh

AUTORESEARCH_CMD templates:

# Python repo (entry script)
AUTORESEARCH_CMD=./external/autoresearch/.venv/bin/python main.py --output ./data/research_signal_raw.json

# Python repo (module)
AUTORESEARCH_CMD=./external/autoresearch/.venv/bin/python -m autoresearch --output ./data/research_signal_raw.json

# Node repo (npm script)
AUTORESEARCH_CMD="npm run research -- --output ./data/research_signal_raw.json"

# Node repo (direct node entry)
AUTORESEARCH_CMD="node index.js --output ./data/research_signal_raw.json"

Node Exporter textfile collector (relative path setup):

# write metrics to a collector subfolder in this repo
STATUS_PROM_ENABLED=true
STATUS_PROM_FILE=./results/scorecards/textfile/trading_scorecard.prom
bash scripts/run_weekly_scorecard.sh

# run node_exporter from this repo root with relative collector directory
node_exporter --collector.textfile --collector.textfile.directory=./results/scorecards/textfile

Helper script (one command, relative paths):

bash scripts/run_node_exporter_textfile_example.sh

# optional overrides
COLLECTOR_DIR=./results/scorecards/textfile \
STATUS_PROM_FILE=./results/scorecards/textfile/trading_scorecard.prom \
NODE_EXPORTER_CMD=node_exporter \
bash scripts/run_node_exporter_textfile_example.sh

Deploy on another server (quick runbook):

# on server
git clone <your-repo-url> /opt/trading_2
cd /opt/trading_2
python3 -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt
cp .env.live.example .env

Then run bot/scorecard from that server only.

VS Code task:

• Scorecard: Weekly Go/No-Go (from .vscode/tasks.json)

🚨 Disclaimer & Risk Warning

⚠️ USE THIS TRADING BOT AT YOUR OWN RISK!

This trading bot is provided "as is" without any warranties, express or implied. The author(s) and contributors are not liable for any financial losses, damages, or other negative consequences resulting from the use of this software. Key Risks to Consider:

• Market Volatility: Cryptocurrency and financial markets are highly unpredictable. Past performance does not guarantee future results.
• Technical Failures: Bugs, connectivity issues, or exchange API changes may lead to unexpected behavior.
• Financial Loss: Trading involves risk, and you may lose part or all of your invested capital.
• No Guarantees: The bot’s strategies may not always be profitable. Always test thoroughly in a simulated environment before using real funds.

Important Notes:

• Backtest & Paper Trade First: Never deploy the bot with real money without extensive testing.
• Monitor Performance: Regularly check the bot’s activity to ensure it behaves as expected.
• Legal & Tax Compliance: Ensure compliance with local regulations regarding automated trading and taxation.

By using this bot, you acknowledge and accept full responsibility for all outcomes, including potential losses. The author(s) disclaim all liability for any damages arising from its use. Trade responsibly!