An advanced web application that analyzes long PR messages using AI with a streamlined concurrent processing workflow.

1. Token Measurement: Precisely count tokens in the input text using gpt-tokenizer
2. Configurable Chunking: Split text into chunks based on user-defined max tokens per chunk
3. Concurrent Analysis: Send all chunks simultaneously to separate LLM calls for maximum speed
4. Summary Generation: Create an executive summary from all chunk analyses
5. Complete Report: Return summary + all individual chunk responses in markdown format

The AI now provides opinionated technical analysis with a structured format:

1. Specific Issues, Concerns, or Problems Mentioned: Complete extraction of all issues
2. Why Issues Matter & Agreement with Solutions: Technical reasoning for why each issue is important and whether the proposed solutions are adequate
3. Code Changes with Suggestions: Original code vs. corrected code with AI's additional improvements
4. What's Missing & Deviation Recommendations: Critical analysis of what's missing and why developers should consider alternative approaches

This format provides not just summarization, but actual technical insights and code improvements.

1. Input Processing: Measure total tokens in the PR conversation text
2. Smart Chunking: Split text into chunks based on user-specified token limits
3. Concurrent Analysis: Process all chunks simultaneously with specialized prompts
4. Summary Generation: Create comprehensive summary from all chunk analyses
5. Report Assembly: Combine summary with detailed chunk analyses

• Concurrent Processing: All chunks analyzed simultaneously (no sequential processing)
• Configurable Limits: User controls chunk size and response detail level
• Precise Token Counting: Accurate measurement ensures optimal chunk sizes
• No Content Loss: Every piece of information is preserved and analyzed
• Fast Results: Concurrent processing dramatically reduces analysis time

• backend/index.ts - Streamlined Hono server with concurrent chunk processing
• frontend/index.html - Main HTML interface
• frontend/index.tsx - React frontend with token configuration and results display
• shared/types.ts - Simplified TypeScript types for the new workflow

1. Copy the entire conversation from your GitHub PR page
2. Paste the text into the textarea
3. Configure token limits:
   • Max Tokens Per Chunk: How much content each AI call receives (100-4000)
   • Max Tokens Per Response: How detailed each analysis can be (100-2000)
4. Click "Analyze PR Messages"
5. Get comprehensive analysis with:
   • Token count and chunking statistics
   • Executive summary of the entire PR
   • Detailed analysis for each chunk
   • Full markdown report for download

• Backend: Hono + Val Town OpenAI (GPT-4o-mini)
• Frontend: React + TailwindCSS with markdown rendering
• Storage: None (stateless analysis)
• Tokenization: gpt-tokenizer for precise token counting
• Processing: Concurrent LLM calls for maximum speed

• ✅ Precise token measurement and configurable chunking
• ✅ Concurrent processing of all chunks simultaneously
• ✅ Executive summary generation from all analyses
• ✅ Complete transparency with original content visibility
• ✅ Configurable response detail levels
• ✅ Full markdown report export
• ✅ No content filtering - every detail preserved
• ✅ Fast results through concurrent processing
• ✅ User-controlled token limits for optimal performance
• ✅ Error handling and validation

• Speed: Concurrent processing eliminates sequential bottlenecks
• Flexibility: User controls chunk size and response detail
• Completeness: No information is filtered out or lost
• Transparency: See exactly how text was chunked and analyzed
• Efficiency: Optimal token usage based on user preferences
• Export: Complete markdown report for documentation