refactor: enhance issue analysis prompt and add context extraction for user inquiries

Author: phodal

packages/context-worker/src/index.ts

@@ -28,6 +28,12 @@ export {
   SymbolKind
 } from './code-context/base/SymbolExtractor';
 
+export {
+  InstantiationService,
+  providerContainer
+} from './base/common/instantiation/instantiationService';
+export * from './base/common/instantiation/instantiation';
+
 // Language utilities
 export { inferLanguage } from './base/common/languages/languages';
 

packages/remote-agent/package.json

@@ -72,6 +72,7 @@
     "cheerio": "^1.0.0",
     "dotenv": "^16.5.0",
     "express": "^5.1.0",
+    "jsdom": "^26.1.0",
     "turndown": "^7.2.0"
   },
   "devDependencies": {
@@ -83,6 +84,7 @@
     "@rollup/plugin-typescript": "^11.1.6",
     "@types/express": "^5.0.1",
     "@types/jest": "^29.5.11",
+    "@types/jsdom": "^21.1.7",
     "@types/node": "^20.17.46",
     "@typescript-eslint/eslint-plugin": "^7.1.0",
     "@typescript-eslint/parser": "^7.1.0",

packages/remote-agent/quick-test-agent.js

@@ -17,12 +17,12 @@ const issuesToTest = [
     description: 'Analyze GitHub issue #105 in unit-mesh/autodev-workbench'
   },
   // Add more issues here, for example:
-  {
-    owner: 'unit-mesh',
-    repo: 'autodev-workbench',
-    issueNumber: 104,
-    description: 'Analyze GitHub issue #104 in unit-mesh/autodev-workbench'
-  }
+  // {
+  //   owner: 'unit-mesh',
+  //   repo: 'autodev-workbench',
+  //   issueNumber: 104,
+  //   description: 'Analyze GitHub issue #104 in unit-mesh/autodev-workbench'
+  // }
 ];
 
 async function quickTest (issueConfig) {
@@ -62,7 +62,7 @@ async function quickTest (issueConfig) {
     console.log(`🔧 Tools: ${agent.getAvailableTools().join(', ')}`)
 
     // Simple test
-    console.log('\n🧪 Running test for current issue...')
+    console.log('🧪 Running test for current issue...')
     const response = await agent.start(
       issueConfig.description,
       {

packages/remote-agent/src/agent/final-report-generator.ts

@@ -25,91 +25,124 @@ export class FinalReportGenerator {
       toolResultsCount: allToolResults.length
     });
 
-    const resultsByRound = this.groupResultsByRound(allToolResults);
+    // const resultsByRound = this.groupResultsByRound(allToolResults);
     const successfulResults = allToolResults.filter(r => r.success);
     const failedResults = allToolResults.filter(r => !r.success);
 
     // const executionSummary = this.buildExecutionSummary(resultsByRound, totalRounds);
     const toolResultsSummary = this.buildToolResultsSummary(successfulResults);
+    const issueContext = this.extractIssueContext(userInput, toolResultsSummary);
 
-    const comprehensivePrompt = `You are an expert GitHub issue analyst tasked with providing a comprehensive PROPOSED ACTION PLAN based on the analysis results. The plan should detail what the user should do next to address their issue.
+    const comprehensivePrompt = `You are an expert GitHub issue analyst and software architect. Your task is to provide a comprehensive response that DIRECTLY ADDRESSES the user's specific issue/question while also providing a detailed development action plan.
 
-## 📝 Original User Request
+## 📝 User's Specific Issue/Question
 ${userInput}
 
-## 🔍 Analysis Data Available
+## 🔍 Analysis Results from Code Investigation
 ${toolResultsSummary}
 
 ${failedResults.length > 0 ? `## ⚠️ Analysis Limitations
-Some tools failed to execute:
+Some analysis tools encountered issues:
 ${failedResults.map(r => `- ${r.functionCall.name}: ${r.error}`).join('\n')}
 ` : ''}
 
 ## ✅ Required Response Format
 
-Generate a **proposed action plan** with these sections:
-
-### 1. Executive Summary
-A 2-3 sentence summary of what was found and what needs to be done to address the user's request.
-
-### 2. Key Findings from Analysis
-List 3-5 concrete, evidence-based findings from the analysis that inform the action plan:
-- Each finding should be clearly stated
-- Reference specific evidence found (e.g., "Analysis of \`file.js\` reveals...")
-- Explain how each finding impacts the proposed solution
-
-### 3. Proposed Action Plan
-Present a step-by-step plan of action:
-- Numbered steps in recommended sequence
-- Each step should be concrete and actionable
-- Include code snippets, file paths, or configuration changes when applicable
-- Indicate expected outcomes for each step
-
-### 4. Technical Implementation Details
-When relevant, provide:
-- Specific code changes recommended
-- Dependencies to add/remove
-- Configuration updates
-- Integration points with existing codebase
-
-### 5. Considerations & Alternative Approaches
-Briefly note:
-- Potential challenges or risks in the proposed approach
-- Alternative approaches if the main plan encounters obstacles
-- Additional information that might be needed
-- Future improvements to consider after implementation
+Your response must include these sections in order:
+
+### 1. Direct Answer to User's Question
+- **Directly address the user's specific question/issue first**
+- Provide a clear, concise answer based on the analysis findings
+- If it's a "how to" question, give the specific steps
+- If it's a "why" question, explain the root cause
+- If it's a "what" question, provide the specific information requested
+
+### 2. Visual Architecture/Flow Diagram
+Create a Mermaid diagram that illustrates:
+- Current system architecture (if analyzing existing code)
+- Proposed solution flow (if implementing new features)
+- Data flow or process flow relevant to the user's issue
+- Component relationships and dependencies
+
+Use appropriate Mermaid diagram types:
+- \`graph TD\` for flowcharts and process flows
+- \`sequenceDiagram\` for interaction flows
+- \`classDiagram\` for class relationships
+- \`gitgraph\` for development workflows
+
+Example format:
+\`\`\`mermaid
+graph TD
+    A["Current State"] --> B["Identified Issue"]
+    B --> C["Proposed Solution"]
+    C --> D["Expected Outcome"]
+\`\`\`
+
+### 3. Evidence-Based Findings
+List 3-5 key findings from the code analysis that support your answer:
+- Reference specific files, functions, or code patterns found
+- Explain how each finding relates to the user's question
+- Include relevant code snippets when helpful
+
+### 4. Step-by-Step Development Plan
+Provide a detailed, actionable plan:
+1. **Immediate Actions** - What to do first
+2. **Core Implementation** - Main development tasks
+3. **Integration Steps** - How to connect with existing code
+4. **Testing Strategy** - How to verify the solution works
+5. **Deployment Considerations** - Production readiness steps
+
+For each step, include:
+- Specific files to modify/create
+- Code examples or templates
+- Expected outcomes
+- Potential challenges
+
+### 5. Implementation Timeline & Priorities
+- **High Priority** (Must do first)
+- **Medium Priority** (Important but can wait)
+- **Low Priority** (Nice to have)
+- **Future Enhancements** (Post-implementation improvements)
+
+### 6. Risk Assessment & Alternatives
+- Potential risks in the proposed approach
+- Mitigation strategies
+- Alternative solutions if the main approach fails
+- Rollback plan if needed
 
 ## 🎯 Response Guidelines
 
-- **Be specific and actionable** - focus on concrete next steps
-- **Reference evidence** discovered during analysis
-- **Present a logical sequence** of actions
-- **Consider the project context** revealed by the analysis
-- **Prioritize practical solutions** that can be implemented immediately
-- **Keep it concise** - aim for clarity and usefulness
+**CRITICAL**: Start by directly answering the user's question before diving into technical details.
 
-Remember: The response should provide a clear, actionable plan that the user can follow to address their issue.`;
+- **Be user-centric** - Address their specific concern first
+- **Use visual aids** - Include relevant Mermaid diagrams
+- **Provide evidence** - Reference actual code findings
+- **Be actionable** - Give concrete next steps
+- **Consider context** - Respect the existing codebase architecture
+- **Think holistically** - Balance immediate needs with long-term maintainability
+
+Remember: The user asked a specific question - answer it clearly first, then provide the comprehensive development plan to implement the solution.`;
 
     try {
       const messages: CoreMessage[] = [
         {
           role: "system",
-          content: "You are an expert software architect and GitHub issue analyst. Your role is to provide a clear, actionable plan based on thorough code analysis. Focus on specific next steps the user should take rather than describing the analysis process. Present a logical sequence of actions that addresses the user's request in the context of their project architecture."
+          content: "You are an expert software architect and GitHub issue analyst. Your primary goal is to directly answer the user's specific question while providing comprehensive technical guidance. Always start with a direct answer to their question, then provide detailed implementation guidance with visual diagrams. Use Mermaid diagrams to illustrate architecture, flows, and relationships. Be specific, actionable, and evidence-based in your recommendations."
         },
         { role: "user", content: comprehensivePrompt }
       ];
 
       this.logger.log('Sending request to LLM', {
         messages,
         temperature: 0.1,
-        maxTokens: 3000
+        maxTokens: 4000
       });
 
       const { text } = await generateText({
         model: this.llmConfig.openai(this.llmConfig.fullModel),
         messages,
         temperature: 0.1,
-        maxTokens: 3000
+        maxTokens: 4000
       });
 
       this.logger.log('Received response from LLM', {
@@ -128,6 +161,28 @@ Remember: The response should provide a clear, actionable plan that the user can
     }
   }
 
+  private extractIssueContext(userInput: string, toolResultsSummary: string): string {
+    // Extract key context from user input to better understand the issue type
+    const issueKeywords = {
+      implementation: ['how to implement', 'how do I', 'how can I', 'implement', 'create', 'build'],
+      debugging: ['error', 'bug', 'issue', 'problem', 'not working', 'fails', 'broken'],
+      architecture: ['architecture', 'design', 'structure', 'organize', 'best practice'],
+      integration: ['integrate', 'connect', 'combine', 'merge', 'link'],
+      optimization: ['optimize', 'improve', 'performance', 'faster', 'better']
+    };
+
+    const lowerInput = userInput.toLowerCase();
+    const detectedTypes: string[] = [];
+
+    Object.entries(issueKeywords).forEach(([type, keywords]) => {
+      if (keywords.some(keyword => lowerInput.includes(keyword))) {
+        detectedTypes.push(type);
+      }
+    });
+
+    return detectedTypes.length > 0 ? detectedTypes.join(', ') : 'general inquiry';
+  }
+
   private groupResultsByRound(results: ToolResult[]): Map<number, ToolResult[]> {
     const grouped = new Map<number, ToolResult[]>();