CodeGraph CLI MCP Server

/// Optimized prompts for Qwen2.5-Coder-14B-128K in CodeGraph MCP context /// /// These prompts are specifically designed to leverage Qwen2.5-Coder's strengths: /// - 128K context window for comprehensive codebase analysis /// - SOTA code understanding and pattern recognition /// - Structured output for MCP-calling LLMs use serde_json::Value; /// Build optimized prompt for comprehensive semantic analysis pub fn build_semantic_analysis_prompt(query: &str, codebase_context: &str) -> String { format!( "<|im_start|>system\n\ You are Qwen2.5-Coder, a state-of-the-art AI specialized in comprehensive code analysis. \ You are providing intelligence for powerful LLMs (Claude, GPT-4) via the CodeGraph MCP protocol.\n\n\ Your role:\n\ - Analyze code semantically, not just syntactically\n\ - Understand architectural patterns and relationships\n\ - Detect team conventions and coding standards\n\ - Provide actionable intelligence for code generation\n\ - Use your 128K context window for complete understanding\n\n\ Output Structure (always follow this format):\n\ 1. SEMANTIC_UNDERSTANDING: Core functionality and purpose\n\ 2. ARCHITECTURAL_CONTEXT: How components fit together\n\ 3. PATTERNS_DETECTED: Consistent patterns and conventions\n\ 4. QUALITY_ASSESSMENT: Code quality and best practices\n\ 5. GENERATION_GUIDANCE: Specific guidance for generating similar code\n\ 6. TEAM_INTELLIGENCE: Team conventions and standards\n\n\ Be comprehensive, structured, and actionable.\n\ <|im_end|>\n\ <|im_start|>user\n\ COMPREHENSIVE CODEBASE ANALYSIS REQUEST\n\n\ ANALYSIS QUERY: {}\n\n\ CODEBASE CONTEXT:\n{}\n\n\ Using your 128K context window and code expertise, provide comprehensive analysis following the structured format above.\n\ <|im_end|>\n\ <|im_start|>assistant\n\ I'll analyze this codebase comprehensively using my 128K context window.\n\n", query, codebase_context ) } /// Build optimized prompt for enhanced search analysis pub fn build_enhanced_search_prompt(query: &str, search_results: &str) -> String { format!( "<|im_start|>system\n\ You are Qwen2.5-Coder providing enhanced search analysis for MCP-calling LLMs.\n\n\ Your task:\n\ - Analyze search results for semantic relevance\n\ - Explain why each result matches the query\n\ - Identify patterns and architectural context\n\ - Provide generation guidance for similar code\n\n\ Output Structure:\n\ 1. SEARCH_ANALYSIS: Why results are semantically relevant\n\ 2. ARCHITECTURAL_INSIGHTS: How results fit in system architecture\n\ 3. USAGE_PATTERNS: How this functionality is typically used\n\ 4. GENERATION_GUIDANCE: How to generate similar high-quality code\n\ <|im_end|>\n\ <|im_start|>user\n\ ENHANCED SEARCH ANALYSIS REQUEST\n\n\ SEARCH QUERY: {}\n\n\ SEARCH RESULTS:\n{}\n\n\ Analyze these search results following the structured format above.\n\ <|im_end|>\n\ <|im_start|>assistant\n\ I'll analyze these search results for semantic understanding and generation guidance.\n\n", query, search_results ) } /// Build optimized prompt for pattern detection pub fn build_pattern_analysis_prompt(code_samples: &[String], focus_area: &str) -> String { let samples_text = code_samples.join("\n---CODE_SAMPLE---\n"); format!( "<|im_start|>system\n\ You are Qwen2.5-Coder analyzing code patterns for consistency and quality.\n\n\ Your expertise:\n\ - Detect established patterns and conventions\n\ - Assess code quality and adherence to best practices\n\ - Identify team-specific standards and preferences\n\ - Provide templates for generating consistent code\n\n\ Output Structure:\n\ 1. IDENTIFIED_PATTERNS: Consistent patterns across code samples\n\ 2. QUALITY_METRICS: Code quality assessment and scores\n\ 3. TEAM_CONVENTIONS: Team-specific conventions detected\n\ 4. GENERATION_TEMPLATES: Reusable templates for similar code\n\ 5. IMPROVEMENT_OPPORTUNITIES: Areas for enhancement\n\ <|im_end|>\n\ <|im_start|>user\n\ PATTERN ANALYSIS REQUEST\n\n\ FOCUS AREA: {}\n\n\ CODE SAMPLES:\n{}\n\n\ Analyze these code samples for patterns and conventions following the structured format.\n\ <|im_end|>\n\ <|im_start|>assistant\n\ I'll analyze these code samples for patterns, conventions, and quality.\n\n", focus_area, samples_text ) } /// Build optimized prompt for impact analysis pub fn build_impact_analysis_prompt( target_function: &str, file_path: &str, dependencies: &str, change_type: &str, ) -> String { format!( "<|im_start|>system\n\ You are Qwen2.5-Coder providing critical impact analysis for code changes.\n\n\ Your expertise:\n\ - Assess risk levels for proposed changes\n\ - Identify affected components and dependencies\n\ - Recommend safe implementation strategies\n\ - Provide testing and validation requirements\n\n\ Output Structure:\n\ 1. RISK_ASSESSMENT: Overall risk level (LOW/MEDIUM/HIGH) with reasoning\n\ 2. AFFECTED_COMPONENTS: Specific components that will be impacted\n\ 3. DEPENDENCY_ANALYSIS: Critical dependencies and cascade effects\n\ 4. TESTING_STRATEGY: Required tests and validation approaches\n\ 5. IMPLEMENTATION_PLAN: Step-by-step safe implementation\n\ 6. ROLLBACK_STRATEGY: How to safely undo changes if needed\n\n\ Prioritize safety and thoroughness.\n\ <|im_end|>\n\ <|im_start|>user\n\ IMPACT ANALYSIS REQUEST\n\n\ TARGET FUNCTION: {} in {}\n\ CHANGE TYPE: {}\n\n\ DEPENDENCIES AND USAGE:\n{}\n\n\ Analyze the impact following the structured format above.\n\ <|im_end|>\n\ <|im_start|>assistant\n\ I'll provide comprehensive impact analysis for this proposed change.\n\n", target_function, file_path, change_type, dependencies ) } /// Extract generation guidance from Qwen analysis with improved parsing pub fn extract_enhanced_generation_guidance(analysis: &str) -> Value { let mut guidance = serde_json::json!({}); // Look for GENERATION_GUIDANCE section if let Some(start) = analysis.find("GENERATION_GUIDANCE:") { let guidance_section = &analysis[start..]; let end_pos = find_section_end(guidance_section); let content = guidance_section[..end_pos].trim(); guidance["detailed_guidance"] = serde_json::json!(content); } // Extract specific guidance elements guidance["patterns_to_follow"] = extract_patterns_from_analysis(analysis); guidance["required_imports"] = extract_imports_from_analysis(analysis); guidance["error_handling"] = extract_error_handling_from_analysis(analysis); guidance["testing_approach"] = extract_testing_from_analysis(analysis); guidance } /// Extract quality assessment with enhanced parsing pub fn extract_enhanced_quality_assessment(analysis: &str) -> Value { let mut assessment = serde_json::json!({}); // Look for QUALITY_ASSESSMENT section if let Some(start) = analysis.find("QUALITY_ASSESSMENT:") { let quality_section = &analysis[start..]; let end_pos = find_section_end(quality_section); let content = quality_section[..end_pos].trim(); assessment["detailed_assessment"] = serde_json::json!(content); } // Extract quality indicators assessment["code_quality_score"] = extract_quality_score(analysis); assessment["best_practices"] = extract_best_practices(analysis); assessment["anti_patterns"] = extract_anti_patterns(analysis); assessment["improvement_suggestions"] = extract_improvements(analysis); assessment } /// Extract structured insights with better section parsing pub fn extract_enhanced_structured_insights(analysis: &str) -> Value { let mut insights = serde_json::json!({}); // Parse numbered sections more reliably for i in 1..=6 { let section_name = match i { 1 => "semantic_understanding", 2 => "architectural_context", 3 => "patterns_detected", 4 => "quality_assessment", 5 => "generation_guidance", 6 => "team_intelligence", _ => continue, }; if let Some(content) = extract_numbered_section(analysis, i) { insights[section_name] = serde_json::json!(content); } } insights } // Helper functions for enhanced parsing fn find_section_end(text: &str) -> usize { // Find the end of a section by looking for next numbered section or end of text for i in 2..=7 { if let Some(pos) = text.find(&format!("{}.", i)) { return pos; } } text.len() } fn extract_numbered_section(text: &str, section_num: i32) -> Option<String> { let section_marker = format!("{}.", section_num); if let Some(start) = text.find(&section_marker) { let section_text = &text[start..]; let end_pos = find_section_end(section_text); Some(section_text[..end_pos].trim().to_string()) } else { None } } fn extract_patterns_from_analysis(analysis: &str) -> Value { let patterns = if analysis.contains("pattern") || analysis.contains("convention") { "Patterns detected in analysis" } else { "No specific patterns mentioned" }; serde_json::json!(patterns) } fn extract_imports_from_analysis(analysis: &str) -> Value { let imports = if analysis.contains("import") || analysis.contains("require") || analysis.contains("use") { "Import requirements mentioned in analysis" } else { "No specific import requirements" }; serde_json::json!(imports) } fn extract_error_handling_from_analysis(analysis: &str) -> Value { let error_handling = if analysis.contains("error") || analysis.contains("exception") || analysis.contains("try") { "Error handling patterns identified" } else { "No specific error handling guidance" }; serde_json::json!(error_handling) } fn extract_testing_from_analysis(analysis: &str) -> Value { let testing = if analysis.contains("test") || analysis.contains("spec") || analysis.contains("assert") { "Testing approach mentioned in analysis" } else { "No specific testing guidance" }; serde_json::json!(testing) } fn extract_quality_score(analysis: &str) -> Value { // Look for quality indicators in the text if analysis.contains("high quality") || analysis.contains("excellent") { serde_json::json!(0.9) } else if analysis.contains("good quality") || analysis.contains("solid") { serde_json::json!(0.8) } else if analysis.contains("average") || analysis.contains("acceptable") { serde_json::json!(0.7) } else { serde_json::json!(0.75) // Default } } fn extract_best_practices(analysis: &str) -> Value { let practices = if analysis.contains("best practice") || analysis.contains("follows standards") { "Best practices identified in analysis" } else { "Best practices analysis included" }; serde_json::json!(practices) } fn extract_anti_patterns(analysis: &str) -> Value { let anti_patterns = if analysis.contains("anti-pattern") || analysis.contains("avoid") || analysis.contains("problematic") { "Anti-patterns or issues identified" } else { "No significant anti-patterns detected" }; serde_json::json!(anti_patterns) } fn extract_improvements(analysis: &str) -> Value { let improvements = if analysis.contains("improve") || analysis.contains("optimize") || analysis.contains("enhance") { "Improvement opportunities identified" } else { "Code appears well-implemented" }; serde_json::json!(improvements) }