indicator_analyzer.py

import ollama
import numpy as np
import pandas as pd
from lightweight_charts import Chart
from transformers import AutoModelForCausalLM, AutoTokenizer

class IndicatorAnalyzer:
    def __init__(self):
        self.model = AutoModelForCausalLM.from_pretrained("tmmdev/codellama-pattern-analysis")
        self.tokenizer = AutoTokenizer.from_pretrained("tmmdev/codellama-pattern-analysis")
        
    def analyze_indicators(self, ohlcv_data):
        indicator_prompt = f"""
        Analyze this OHLCV data and calculate optimal indicators:
        {ohlcv_data.to_json(orient='records')}
        Calculate and return:
        - Moving Averages (EMA, SMA with optimal periods)
        - Oscillators (RSI, Stochastic, MACD)
        - Volatility (Bollinger Bands, ATR)
        - Volume indicators
        - Custom combinations of indicators
        
        Return the analysis in JSON format with exact values and coordinates.
        """
        
        response = self.client.chat(
            model='codellama:latest',
            messages=[
                {
                    'role': 'system',
                    'content': 'You are a technical analysis indicator calculation model.'
                },
                {
                    'role': 'user',
                    'content': indicator_prompt
                }
            ]
        )
        
        return self.parse_indicator_analysis(response['message']['content'])
        
    def parse_indicator_analysis(self, analysis):
        try:
            # Convert string response to structured data
            if isinstance(analysis, str):
                # Extract JSON if embedded in text
                json_start = analysis.find('{')
                json_end = analysis.rfind('}') + 1
                if json_start >= 0 and json_end > 0:
                    analysis = analysis[json_start:json_end]
                    
            indicators = {
                'moving_averages': {},
                'oscillators': {},
                'volatility': {},
                'volume': {},
                'custom': {}
            }
            
            # Add any custom parsing logic here
            
            return indicators
            
        except Exception as e:
            print(f"Error parsing indicator analysis: {str(e)}")
            return {}