trading-tools / agents /technical /indicator_agent.py
Deploy Bot
Deploy Trading Analysis Platform to HuggingFace Spaces
a1bf219
"""
Indicator Agent for technical indicator calculation and interpretation.
This agent computes technical indicators (RSI, MACD, Stochastic) and provides
interpretation of their values for trading decisions.
"""
import json
import logging
import time
from typing import Any, Dict, Optional
import pandas as pd
from langchain_core.messages import HumanMessage, SystemMessage
# Configure logger
logger = logging.getLogger(__name__)
from config.default_config import DEFAULT_CONFIG
from config.models import AGENT_MODELS
from config.prompt_templates import INDICATOR_AGENT_PROMPT
from graph.state.agent_state import add_agent_message, update_analysis_result
from graph.state.trading_state import TechnicalWorkflowState
from utils.charts.chart_generator import ChartGenerator
from utils.formatters.educational_content import (
generate_macd_explanation,
generate_rsi_explanation,
generate_stochastic_explanation,
)
from utils.indicators import calculate_macd, calculate_rsi, calculate_stochastic
from utils.indicators.macd import (
find_macd_crossovers,
find_macd_divergence,
interpret_macd,
)
from utils.indicators.rsi import find_rsi_divergence, interpret_rsi
from utils.indicators.stochastic import find_stochastic_crossovers, interpret_stochastic
from utils.investment_style_helpers import (
get_investment_style_from_state,
get_technical_analysis_style_context,
)
from utils.llm.provider_factory import LLMProviderFactory
class IndicatorAgent:
"""
Technical Indicator Agent.
Responsibilities:
- Calculate RSI, MACD, Stochastic Oscillator
- Interpret indicator values (overbought/oversold, bullish/bearish)
- Detect divergences and crossovers
- Provide trading signals based on indicators
"""
AGENT_NAME = "indicator_agent"
def __init__(self, config: Optional[Dict[str, Any]] = None):
"""
Initialize Indicator Agent.
Args:
config: Optional configuration override
"""
self.config = config or DEFAULT_CONFIG
# Initialize LLM - use runtime provider override if available
from config.models import DEFAULT_MODELS_BY_PROVIDER
model_config = AGENT_MODELS[self.AGENT_NAME]
runtime_provider = self.config.get("llm_provider", model_config["provider"])
# If provider is overridden but model is not, use default model for that provider
if "llm_provider" in self.config and "llm_model" not in self.config:
runtime_model = DEFAULT_MODELS_BY_PROVIDER.get(
runtime_provider, model_config["model"]
)
else:
runtime_model = self.config.get("llm_model", model_config["model"])
self.llm = LLMProviderFactory.create(
provider=runtime_provider,
model=runtime_model,
temperature=model_config["temperature"],
)
# Indicator parameters
self.indicator_params = self.config["indicator_parameters"]
# Initialize chart generator
self.chart_generator = ChartGenerator()
def run(self, state: TechnicalWorkflowState) -> TechnicalWorkflowState:
"""
Execute indicator analysis.
Args:
state: Current workflow state
Returns:
Updated state with indicator analysis
"""
start_time = time.time()
ticker = state.get("ticker", "UNKNOWN")
timeframe = state.get("timeframe", "UNKNOWN")
logger.info(
json.dumps(
{
"agent": self.AGENT_NAME,
"action": "start",
"ticker": ticker,
"timeframe": timeframe,
"timestamp": time.time(),
}
)
)
try:
# Extract market data
market_data = state["market_data"]
if not market_data.get("ohlc_data"):
raise ValueError("No OHLC data available for indicator calculation")
# Convert serialized DataFrame back to pandas DataFrame
df = self._deserialize_dataframe(market_data["ohlc_data"])
# Calculate indicators
indicators_result = self._calculate_indicators(df)
# Get investment style from state
investment_style = get_investment_style_from_state(state)
# Generate charts and educational notes (User Story 5)
chart_paths = []
educational_notes = []
try:
# Check if educational mode is enabled
config = state.get("config", {})
educational_mode = (
config.get("educational_mode", False)
if isinstance(config, dict)
else False
)
# Generate RSI chart
if "rsi" in indicators_result and "value" in indicators_result["rsi"]:
rsi_value = indicators_result["rsi"]["value"]
rsi_series_dict = indicators_result["rsi"].get("series")
if rsi_series_dict is not None:
# Convert dict back to Series
rsi_series = pd.Series(rsi_series_dict)
fig, filepath = self.chart_generator.generate_rsi_chart(
df=df,
rsi_series=rsi_series,
ticker=ticker,
timeframe=timeframe,
rsi_period=self.indicator_params["rsi_period"],
save=True,
)
if filepath:
chart_paths.append(filepath)
self.chart_generator.close_figure(fig)
# Add educational note if enabled
if educational_mode:
educational_notes.append(
f"**RSI**: {generate_rsi_explanation(rsi_value)}"
)
# Generate MACD chart
if "macd" in indicators_result:
macd_data = indicators_result["macd"]
series_dict = macd_data.get("series", {})
logger.info(
f"MACD data available: series_dict keys = {list(series_dict.keys()) if series_dict else 'None'}"
)
if (
series_dict
and "macd" in series_dict
and "signal" in series_dict
and "histogram" in series_dict
):
# Convert dicts back to Series
macd_series = pd.Series(series_dict["macd"])
signal_series = pd.Series(series_dict["signal"])
histogram_series = pd.Series(series_dict["histogram"])
logger.info(
f"Generating MACD chart: macd_len={len(macd_series)}, signal_len={len(signal_series)}, hist_len={len(histogram_series)}"
)
fig, filepath = self.chart_generator.generate_macd_chart(
df=df,
macd=macd_series,
signal=signal_series,
histogram=histogram_series,
ticker=ticker,
timeframe=timeframe,
save=True,
)
logger.info(f"MACD chart generated: filepath={filepath}")
if filepath:
chart_paths.append(filepath)
self.chart_generator.close_figure(fig)
else:
logger.warning(
f"MACD chart skipped - missing series data. series_dict keys: {list(series_dict.keys()) if series_dict else 'None'}"
)
# Add educational note if enabled
if (
educational_mode
and "macd" in macd_data
and "signal" in macd_data
and "histogram" in macd_data
):
educational_notes.append(
f"**MACD**: {generate_macd_explanation(macd_data['macd'], macd_data['signal'], macd_data['histogram'])}"
)
# Generate Stochastic chart
if "stochastic" in indicators_result:
stoch_data = indicators_result["stochastic"]
series_dict = stoch_data.get("series", {})
if series_dict and "k" in series_dict and "d" in series_dict:
# Convert dicts back to Series
k_series = pd.Series(series_dict["k"])
d_series = pd.Series(series_dict["d"])
fig, filepath = self.chart_generator.generate_stochastic_chart(
df=df,
k_series=k_series,
d_series=d_series,
ticker=ticker,
timeframe=timeframe,
save=True,
)
if filepath:
chart_paths.append(filepath)
self.chart_generator.close_figure(fig)
# Add educational note if enabled
if educational_mode and "k" in stoch_data and "d" in stoch_data:
educational_notes.append(
f"**Stochastic**: {generate_stochastic_explanation(stoch_data['k'], stoch_data['d'])}"
)
except Exception as chart_error:
logger.warning(
json.dumps(
{
"agent": self.AGENT_NAME,
"action": "chart_generation_warning",
"ticker": ticker,
"error": str(chart_error),
"timestamp": time.time(),
}
)
)
# Extract cost tracker from state
cost_tracker = state.get("_cost_tracker")
# Interpret indicators using LLM
interpretation = self._interpret_with_llm(
state["ticker"],
state["timeframe"],
indicators_result,
df,
investment_style,
cost_tracker,
)
# Append educational notes to interpretation if available
if educational_notes:
interpretation += "\n\n### 📚 Educational Notes\n\n" + "\n\n".join(
educational_notes
)
# Update state
new_state = update_analysis_result(state, "indicators", indicators_result)
new_state = add_agent_message(
new_state,
self.AGENT_NAME,
interpretation,
metadata={
"indicators": indicators_result,
"chart_paths": chart_paths,
"educational_mode": educational_mode,
},
)
execution_time = time.time() - start_time
logger.info(
json.dumps(
{
"agent": self.AGENT_NAME,
"action": "complete",
"ticker": ticker,
"timeframe": timeframe,
"execution_time": execution_time,
"indicators_calculated": list(indicators_result.keys()),
"timestamp": time.time(),
}
)
)
return new_state
except Exception as e:
execution_time = time.time() - start_time
logger.error(
json.dumps(
{
"agent": self.AGENT_NAME,
"action": "error",
"ticker": ticker,
"timeframe": timeframe,
"execution_time": execution_time,
"error": str(e),
"timestamp": time.time(),
}
)
)
# Add error message to state
error_state = add_agent_message(
state,
self.AGENT_NAME,
f"Error calculating indicators: {str(e)}",
metadata={"error": True},
)
return error_state
def _calculate_indicators(self, df: pd.DataFrame) -> Dict[str, Any]:
"""
Calculate all technical indicators.
Args:
df: OHLC DataFrame
Returns:
Dict with indicator results
"""
result = {}
# RSI
try:
rsi_series = calculate_rsi(
df,
period=self.indicator_params["rsi_period"],
)
current_rsi = float(rsi_series.iloc[-1])
rsi_interpretation = interpret_rsi(current_rsi)
# Try to find divergences, but don't fail if it doesn't work
try:
rsi_divergence = find_rsi_divergence(df, rsi_series)
except Exception:
rsi_divergence = {"bullish": [], "bearish": []}
result["rsi"] = {
"value": current_rsi,
"interpretation": rsi_interpretation,
"divergences": rsi_divergence,
"series": rsi_series.to_dict(), # For charting
}
except Exception as e:
result["rsi"] = {"error": str(e)}
# MACD
try:
logger.info(f"Calculating MACD with {len(df)} data points")
macd, signal, hist = calculate_macd(
df,
fast_period=self.indicator_params["macd_fast"],
slow_period=self.indicator_params["macd_slow"],
signal_period=self.indicator_params["macd_signal"],
)
logger.info(
f"MACD calculation succeeded: macd_len={len(macd)}, valid_values={(~pd.isna(macd)).sum()}"
)
current_macd = float(macd.iloc[-1]) if not pd.isna(macd.iloc[-1]) else None
current_signal = (
float(signal.iloc[-1]) if not pd.isna(signal.iloc[-1]) else None
)
current_hist = float(hist.iloc[-1]) if not pd.isna(hist.iloc[-1]) else None
prev_hist = (
float(hist.iloc[-2])
if len(hist) > 1 and not pd.isna(hist.iloc[-2])
else None
)
macd_interpretation = interpret_macd(
current_macd, current_signal, current_hist, prev_hist
)
# Try to find crossovers and divergences, but don't fail if it doesn't work
try:
macd_crossovers = find_macd_crossovers(macd, signal)
except Exception:
macd_crossovers = {"bullish": [], "bearish": []}
try:
macd_divergence = find_macd_divergence(df, macd)
except Exception:
macd_divergence = {"bullish": [], "bearish": []}
result["macd"] = {
"macd": current_macd,
"signal": current_signal,
"histogram": current_hist,
"interpretation": macd_interpretation,
"crossovers": macd_crossovers,
"divergences": macd_divergence,
"series": {
"macd": macd.to_dict(),
"signal": signal.to_dict(),
"histogram": hist.to_dict(),
},
}
except Exception as e:
logger.error(f"MACD calculation failed: {str(e)}")
result["macd"] = {"error": str(e)}
# Stochastic Oscillator
try:
k_series, d_series = calculate_stochastic(
df,
k_period=self.indicator_params["stoch_k_period"],
d_period=self.indicator_params["stoch_d_period"],
)
current_k = (
float(k_series.iloc[-1]) if not pd.isna(k_series.iloc[-1]) else None
)
current_d = (
float(d_series.iloc[-1]) if not pd.isna(d_series.iloc[-1]) else None
)
prev_k = (
float(k_series.iloc[-2])
if len(k_series) > 1 and not pd.isna(k_series.iloc[-2])
else None
)
prev_d = (
float(d_series.iloc[-2])
if len(d_series) > 1 and not pd.isna(d_series.iloc[-2])
else None
)
stoch_interpretation = interpret_stochastic(
current_k, current_d, prev_k, prev_d
)
stoch_crossovers = find_stochastic_crossovers(k_series, d_series)
result["stochastic"] = {
"k": current_k,
"d": current_d,
"interpretation": stoch_interpretation,
"crossovers": stoch_crossovers,
"series": {
"k": k_series.to_dict(),
"d": d_series.to_dict(),
},
}
except Exception as e:
result["stochastic"] = {"error": str(e)}
return result
def _interpret_with_llm(
self,
ticker: str,
timeframe: str,
indicators: Dict[str, Any],
df: pd.DataFrame,
investment_style: Optional[str] = None,
cost_tracker=None,
) -> str:
"""
Use LLM to interpret indicator signals holistically.
Args:
ticker: Asset ticker
timeframe: Analysis timeframe
indicators: Calculated indicators
df: OHLC DataFrame
investment_style: Investment style for context
cost_tracker: Optional cost tracker for tracking LLM costs
Returns:
LLM interpretation string
"""
# Prepare indicator summary
current_price = float(df["close"].iloc[-1])
summary_parts = [
f"Asset: {ticker}",
f"Timeframe: {timeframe}",
f"Current Price: ${current_price:.2f}",
"",
"Technical Indicators:",
]
# RSI
if "rsi" in indicators and "value" in indicators["rsi"]:
rsi = indicators["rsi"]
summary_parts.append(
f"- RSI({self.indicator_params['rsi_period']}): {rsi['value']:.2f}"
)
summary_parts.append(f" {rsi['interpretation']}")
if rsi.get("divergences", {}).get("bullish"):
summary_parts.append(
f" Bullish divergences detected at indices: {rsi['divergences']['bullish']}"
)
if rsi.get("divergences", {}).get("bearish"):
summary_parts.append(
f" Bearish divergences detected at indices: {rsi['divergences']['bearish']}"
)
# MACD
if "macd" in indicators and "macd" in indicators["macd"]:
macd = indicators["macd"]
summary_parts.append(
f"- MACD: {macd['macd']:.4f}, Signal: {macd['signal']:.4f}, Histogram: {macd['histogram']:.4f}"
)
summary_parts.append(f" {macd['interpretation']}")
if macd.get("crossovers", {}).get("bullish"):
summary_parts.append(
f" Recent bullish crossovers at indices: {macd['crossovers']['bullish'][-3:]}"
)
if macd.get("crossovers", {}).get("bearish"):
summary_parts.append(
f" Recent bearish crossovers at indices: {macd['crossovers']['bearish'][-3:]}"
)
# Stochastic
if "stochastic" in indicators and "k" in indicators["stochastic"]:
stoch = indicators["stochastic"]
summary_parts.append(
f"- Stochastic: %K={stoch['k']:.2f}, %D={stoch['d']:.2f}"
)
summary_parts.append(f" {stoch['interpretation']}")
indicator_summary = "\n".join(summary_parts)
# Get investment style context
style_context = get_technical_analysis_style_context(investment_style)
# LLM prompt with specialized indicator template
system_prompt = f"""{INDICATOR_AGENT_PROMPT}
Investment Style Context:
{style_context}
IMPORTANT: Your response MUST follow the exact structure shown in the template above, including:
- Markdown section headers (##)
- Data tables with proper markdown table syntax (| pipes)
- Bullet-pointed insights (-)
- Numbered summary points (1., 2., 3.)
- Clear conclusion with trading implication"""
user_prompt = f"""Analyze the following technical indicators for {ticker} ({timeframe} timeframe) and provide a comprehensive technical analysis following the template structure:
{indicator_summary}
Generate your response following the exact template structure with all sections, tables, bullet points, and numbered summary."""
# Call LLM with cost tracking callback
messages = [
SystemMessage(content=system_prompt),
HumanMessage(content=user_prompt),
]
# Create callback if cost tracker is available
if cost_tracker:
callback = cost_tracker.get_callback(agent_name=self.AGENT_NAME)
response = self.llm.invoke(messages, config={"callbacks": [callback]})
else:
response = self.llm.invoke(messages)
return response.content
def _deserialize_dataframe(self, data: Dict[str, Any]) -> pd.DataFrame:
"""
Convert serialized data back to DataFrame.
Args:
data: Serialized DataFrame data
Returns:
pandas DataFrame
"""
# Assuming data is stored as dict with columns
# This will be properly implemented when we serialize in the workflow
df = pd.DataFrame(data)
if "Date" in df.columns:
df["Date"] = pd.to_datetime(df["Date"])
df = df.set_index("Date")
return df