services/simple_llm.py

"""
Simple LLM service that works without API keys using basic text processing
"""
import re
from typing import List, Dict

class SimpleLLM:
    """A basic LLM that provides helpful responses about code repositories without requiring API keys"""
    
    def __init__(self):
        self.code_keywords = {
            'python': ['def ', 'class ', 'import ', 'from ', '__init__', 'self.'],
            'javascript': ['function', 'const ', 'let ', 'var ', 'async', 'await'],
            'java': ['public class', 'private ', 'public ', 'static', 'void'],
            'cpp': ['#include', 'int main', 'class ', 'namespace'],
            'go': ['func ', 'package ', 'import', 'type'],
            'rust': ['fn ', 'struct ', 'impl ', 'use ', 'mod '],
        }
    
    def analyze_code_context(self, context: str) -> Dict:
        """Analyze the code context to extract useful information"""
        lines = context.split('\n')
        
        # Detect programming languages
        languages = set()
        for lang, keywords in self.code_keywords.items():
            for keyword in keywords:
                if keyword in context:
                    languages.add(lang)
        
        # Extract file paths
        file_paths = []
        for line in lines:
            if 'path' in line.lower() and ('/' in line or '\\' in line):
                file_paths.append(line.strip())
        
        # Count functions and classes
        functions = len(re.findall(r'\bdef\s+\w+|function\s+\w+|func\s+\w+', context))
        classes = len(re.findall(r'\bclass\s+\w+', context))
        
        # Extract imports/dependencies
        imports = re.findall(r'import\s+[\w.]+|from\s+[\w.]+\s+import|#include\s*<[\w.]+>', context)
        
        return {
            'languages': list(languages),
            'file_paths': file_paths[:5],  # Limit to 5 paths
            'functions_count': functions,
            'classes_count': classes,
            'imports': imports[:10],  # Limit to 10 imports
            'total_lines': len(lines)
        }
    
    def generate_response(self, question: str, context: str) -> str:
        """Generate a helpful response based on the question and code context"""
        question_lower = question.lower()
        analysis = self.analyze_code_context(context)
        
        # Question type detection and response generation
        if any(word in question_lower for word in ['what', 'about', 'project', 'repository']):
            return self._describe_project(analysis, context)
        
        elif any(word in question_lower for word in ['structure', 'organized', 'architecture']):
            return self._describe_structure(analysis, context)
        
        elif any(word in question_lower for word in ['function', 'method', 'class']):
            return self._describe_functions_classes(analysis, context)
        
        elif any(word in question_lower for word in ['dependency', 'dependencies', 'import', 'library']):
            return self._describe_dependencies(analysis, context)
        
        elif any(word in question_lower for word in ['test', 'testing']):
            return self._describe_tests(analysis, context)
        
        elif any(word in question_lower for word in ['error', 'exception', 'handling']):
            return self._describe_error_handling(analysis, context)
        
        else:
            return self._general_response(analysis, context)
    
    def _describe_project(self, analysis: Dict, context: str) -> str:
        languages = ", ".join(analysis['languages']) if analysis['languages'] else "multiple languages"
        
        response = f"This project appears to be written in {languages}. "
        
        if analysis['classes_count'] > 0:
            response += f"It contains {analysis['classes_count']} classes "
        if analysis['functions_count'] > 0:
            response += f"and {analysis['functions_count']} functions. "
        
        if 'api' in context.lower() or 'endpoint' in context.lower():
            response += "It appears to be an API or web service. "
        
        if 'test' in context.lower():
            response += "The project includes test files. "
        
        return response.strip()
    
    def _describe_structure(self, analysis: Dict, context: str) -> str:
        response = "The code is organized with the following structure:\n\n"
        
        if analysis['file_paths']:
            response += "**Key files/directories:**\n"
            for path in analysis['file_paths']:
                response += f"- {path}\n"
            response += "\n"
        
        if analysis['languages']:
            response += f"**Languages used:** {', '.join(analysis['languages'])}\n\n"
        
        if analysis['classes_count'] > 0:
            response += f"**Classes found:** {analysis['classes_count']}\n"
        if analysis['functions_count'] > 0:
            response += f"**Functions found:** {analysis['functions_count']}\n"
        
        return response
    
    def _describe_functions_classes(self, analysis: Dict, context: str) -> str:
        response = ""
        
        if analysis['classes_count'] > 0:
            response += f"Found {analysis['classes_count']} classes in the codebase. "
        
        if analysis['functions_count'] > 0:
            response += f"Found {analysis['functions_count']} functions/methods. "
        
        # Extract some actual function/class names from context
        class_names = re.findall(r'class\s+(\w+)', context)
        function_names = re.findall(r'def\s+(\w+)|function\s+(\w+)', context)
        
        if class_names:
            response += f"\n\n**Some classes:** {', '.join(class_names[:5])}"
        
        if function_names:
            func_list = [name for group in function_names for name in group if name]
            response += f"\n\n**Some functions:** {', '.join(func_list[:5])}"
        
        return response if response else "No classes or functions clearly identified in the provided context."
    
    def _describe_dependencies(self, analysis: Dict, context: str) -> str:
        if analysis['imports']:
            response = "**Dependencies and imports found:**\n\n"
            for imp in analysis['imports']:
                response += f"- {imp}\n"
            return response
        else:
            return "No clear dependencies or imports identified in the provided context."
    
    def _describe_tests(self, analysis: Dict, context: str) -> str:
        if 'test' in context.lower():
            return "Test files appear to be present in this codebase. Look for files with 'test' in their names or directories."
        else:
            return "No obvious test files identified in the provided context."
    
    def _describe_error_handling(self, analysis: Dict, context: str) -> str:
        error_patterns = ['try:', 'except:', 'catch', 'throw', 'error', 'exception']
        found_patterns = [pattern for pattern in error_patterns if pattern in context.lower()]
        
        if found_patterns:
            return f"Error handling appears to be implemented using: {', '.join(found_patterns)}"
        else:
            return "No obvious error handling patterns identified in the provided context."
    
    def _general_response(self, analysis: Dict, context: str) -> str:
        response = "Based on the code context provided:\n\n"
        
        if analysis['languages']:
            response += f"- **Languages:** {', '.join(analysis['languages'])}\n"
        
        if analysis['total_lines'] > 0:
            response += f"- **Code size:** {analysis['total_lines']} lines analyzed\n"
        
        if analysis['functions_count'] > 0:
            response += f"- **Functions:** {analysis['functions_count']} found\n"
        
        if analysis['classes_count'] > 0:
            response += f"- **Classes:** {analysis['classes_count']} found\n"
        
        response += "\nFor more specific information, please ask about particular aspects of the code."
        
        return response

class NoAPILLM:
    """Wrapper to use SimpleLLM with the same interface as other LLMs"""
    
    def __init__(self):
        self.simple_llm = SimpleLLM()
    
    def __call__(self, prompt: str) -> str:
        # Extract context and question from the prompt
        if "Context from repository:" in prompt and "Question:" in prompt:
            parts = prompt.split("Question:")
            if len(parts) == 2:
                context_part = parts[0].replace("Context from repository:", "").strip()
                question_part = parts[1].replace("Answer:", "").strip()
                
                return self.simple_llm.generate_response(question_part, context_part)
        
        # Fallback for other prompt formats
        return "I can help analyze code repositories. Please provide specific questions about the codebase structure, functions, dependencies, or other aspects of the code."