Use Pine Script indicators as data sources inside your FreqTrade strategy. You write the trading logic in Python — PyneCore handles the indicator math.

This is the most flexible integration: grab any indicator from TradingView, compile it with PyneComp, and plug it into your FreqTrade bot.

uv run run.py

Generates 500 BTC/USDT candles, runs RSI + Bollinger Bands, and prints combined signals.

1. Copy these files into your FreqTrade project:

   user_data/strategies/
   ├── strategy.py           # Your FreqTrade strategy
   ├── pynecore_bridge.py    # DataFrame ↔ PyneCore bridge
   └── scripts/
       ├── rsi.py
       └── bollinger_bands.py
   

2. Add PyneCore to your FreqTrade environment:

   pip install "pynesys-pynecore>=6.10.6"

3. Run a backtest:

   freqtrade backtesting --strategy PyneIndicatorStrategy

FreqTrade DataFrame (pandas)
        │
        ▼
  pynecore_bridge.py
  ├── dataframe_to_ohlcv()   →  Convert rows to OHLCV objects
  ├── create_syminfo()        →  Build symbol metadata
  └── run_indicator()         →  Run script, return pd.Series
        │
        ▼
  ScriptRunner.run_iter()     →  Process each bar through the Pine Script
        │
        ▼
  plot_data dict              →  {"RSI": 45.2, "Upper": 43100.5, ...}
        │
        ▼
  Back to DataFrame columns   →  df["rsi"], df["bb_upper"], ...

1. Compile a Pine Script indicator with PyneComp (or write one by hand)

2. Place the .py file in scripts/

3. Call run_indicator() in populate_indicators():

   macd_data = run_indicator(dataframe, SCRIPTS_DIR / "my_macd.py", pair=pair, timeframe=self.timeframe)
   dataframe["macd"] = macd_data.get("MACD")
   dataframe["signal"] = macd_data.get("Signal")

4. Use the new columns in populate_entry_trend() / populate_exit_trend()

FreqTrade calls populate_indicators() once per new candle (process_only_new_candles, on by default) with the full DataFrame, and this example re-runs PyneCore on all of its bars. That is the right approach: a Pine Script indicator carries state from bar to bar (moving averages, ta.rsi, var variables), so its value on the newest bar depends on every bar before it. Running only the new bars would start that state from scratch and give different values.

A full re-run is cheap. On a laptop one indicator takes a few milliseconds for 1,000 bars and a few tens of milliseconds for 5,000 bars, far below a candle's duration even on 1m.