neon_sakura 0.1.4

data/bin/ai-optimize-claude-md

@@ -0,0 +1,540 @@
+#!/usr/bin/env bash
+# AI-powered CLAUDE.md optimization
+# Analyzes and suggests improvements to reduce token usage while maintaining clarity
+#
+# Usage:
+#   bin/ai-optimize-claude-md [OPTIONS] [BASE_BRANCH] [OLLAMA_URL]
+#   bin/ai-optimize-claude-md --help | -h
+#
+# Options:
+#   -m, --model MODEL  Specify the model to use (default: codestral:latest)
+#   -i, --interactive  Interactively select model from available models
+#   -o, --output FILE  Output file for optimized version (default: CLAUDE.md.optimized)
+#   -d, --diff         Show changes to CLAUDE.md between branches (no optimization)
+#   --apply            Apply changes directly to CLAUDE.md (creates backup)
+#   --help, -h         Show this help message
+#
+# Arguments:
+#   BASE_BRANCH - Branch to compare CLAUDE.md against (optional, for -d mode)
+#   OLLAMA_URL  - Ollama server URL (default: $OLLAMA_URL env var or http://localhost:11434)
+#
+# Examples:
+#   bin/ai-optimize-claude-md                    # Optimize current CLAUDE.md
+#   bin/ai-optimize-claude-md -d main            # Show CLAUDE.md changes vs main
+#   bin/ai-optimize-claude-md -i                 # Interactively select model
+#   bin/ai-optimize-claude-md -o custom.md       # Save to custom file
+#   bin/ai-optimize-claude-md --apply            # Apply changes to CLAUDE.md (with backup)
+
+# Load shared library
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+if [ -f "$SCRIPT_DIR/../lib/ai-script-base.sh" ]; then
+  # shellcheck source=../lib/ai-script-base.sh
+  source "$SCRIPT_DIR/../lib/ai-script-base.sh"
+else
+  echo "Error: Shared library not found at $SCRIPT_DIR/../lib/ai-script-base.sh"
+  exit 1
+fi
+
+# Show help
+show_help() {
+  cat << EOF
+AI CLAUDE.md Optimizer - Reduce token usage while maintaining clarity
+
+USAGE:
+  bin/ai-optimize-claude-md [OPTIONS] [BASE_BRANCH] [OLLAMA_URL]
+  bin/ai-optimize-claude-md --help | -h
+
+OPTIONS:
+  -m, --model MODEL     Specify the model to use (default: codestral:latest)
+  -i, --interactive     Interactively select model from available models
+  -o, --output FILE     Output file for optimized version (default: CLAUDE.md.optimized)
+  -d, --diff            Show CLAUDE.md changes between branches (no optimization)
+  --apply               Apply changes directly to CLAUDE.md (creates backup)
+  --help, -h            Show this help message
+
+ARGUMENTS:
+  BASE_BRANCH          Branch to compare CLAUDE.md against (required for -d mode)
+  OLLAMA_URL           Ollama server URL (default: \$OLLAMA_URL env var or http://localhost:11434)
+
+EXAMPLES:
+  # Optimize current CLAUDE.md
+  bin/ai-optimize-claude-md
+
+  # Show changes to CLAUDE.md between current branch and main
+  bin/ai-optimize-claude-md -d main
+
+  # Review CLAUDE.md changes with AI analysis
+  bin/ai-optimize-claude-md -d develop -m llama3:latest
+
+  # Interactively select model
+  bin/ai-optimize-claude-md -i
+
+  # Use specific model
+  bin/ai-optimize-claude-md -m llama3:latest
+
+  # Save to custom file
+  bin/ai-optimize-claude-md -o custom.md
+
+  # Apply changes directly (with backup)
+  bin/ai-optimize-claude-md --apply
+
+  # Use remote Ollama server
+  bin/ai-optimize-claude-md main http://server:11434
+
+WHAT IT DOES:
+  - Analyzes CLAUDE.md for redundancy and verbosity
+  - Suggests more concise wording
+  - Identifies duplicate information
+  - Recommends better organization
+  - Estimates token savings
+  - Maintains all critical information
+
+REQUIREMENTS:
+  - Ollama installed and running (https://ollama.ai)
+  - At least one model pulled (e.g., ollama pull codestral:latest)
+  - CLAUDE.md file in current directory (or in compared branch for -d mode)
+
+ENVIRONMENT VARIABLES:
+  OLLAMA_URL           Ollama server URL (overridden by command line argument)
+                       Example: export OLLAMA_URL=http://192.168.1.100:11434
+
+OUTPUT:
+  - Optimization suggestions: docs/ai-reviews/claude-md-optimization-YYYYMMDD-HHMMSS.md
+  - Optimized version: CLAUDE.md.optimized (or specified file)
+  - Backup (if --apply): CLAUDE.md.backup-YYYYMMDD-HHMMSS
+  - Diff analysis (with -d): docs/ai-reviews/claude-md-diff-YYYYMMDD-HHMMSS.md
+
+MORE INFO:
+  See docs/AI_PR_REVIEWER.md for full documentation
+EOF
+  exit 0
+}
+
+# Parse options
+INTERACTIVE_MODE=false
+DIFF_MODE=false
+MODEL_NAME="codestral:latest"
+OUTPUT_FILE="CLAUDE.md.optimized"
+APPLY_CHANGES=false
+BASE_BRANCH=""
+OLLAMA_URL=""
+
+while [[ $# -gt 0 ]]; do
+  case $1 in
+    --help|-h)
+      show_help
+      ;;
+    -i|--interactive)
+      INTERACTIVE_MODE=true
+      shift
+      ;;
+    -d|--diff)
+      DIFF_MODE=true
+      shift
+      ;;
+    -m|--model)
+      if [ -z "${2:-}" ]; then
+        log_error "Option -m/--model requires a model name"
+        exit 1
+      fi
+      MODEL_NAME="$2"
+      shift 2
+      ;;
+    -o|--output)
+      if [ -z "${2:-}" ]; then
+        log_error "Option -o/--output requires a file path"
+        exit 1
+      fi
+      OUTPUT_FILE="$2"
+      shift 2
+      ;;
+    --apply)
+      APPLY_CHANGES=true
+      shift
+      ;;
+    *)
+      # Positional arguments
+      if [ -z "$BASE_BRANCH" ]; then
+        BASE_BRANCH="$1"
+      elif [ -z "$OLLAMA_URL" ]; then
+        OLLAMA_URL="$1"
+      else
+        log_error "Unknown argument: $1"
+        echo "Run 'bin/ai-optimize-claude-md --help' for usage information"
+        exit 1
+      fi
+      shift
+      ;;
+  esac
+done
+
+# Set defaults
+OLLAMA_URL=$(get_ollama_url "$OLLAMA_URL")
+
+# Validate diff mode requirements
+if [ "$DIFF_MODE" = true ] && [ -z "$BASE_BRANCH" ]; then
+  log_error "Diff mode (-d) requires a base branch to compare against"
+  echo "Example: bin/ai-optimize-claude-md -d main"
+  exit 1
+fi
+
+# Create temp directory
+TEMP_DIR=$(create_temp_dir)
+
+if [ "$DIFF_MODE" = true ]; then
+  log_info "Starting CLAUDE.md diff analysis..."
+  log_info "Comparing against: ${CYAN}$BASE_BRANCH${NC}"
+else
+  log_info "Starting CLAUDE.md optimization analysis..."
+fi
+log_info "Ollama server: ${CYAN}$OLLAMA_URL${NC}"
+echo ""
+
+# Check if Ollama is accessible
+check_ollama_server "$OLLAMA_URL" || exit 1
+echo ""
+
+# Interactive model selection if requested
+if [ "$INTERACTIVE_MODE" = true ]; then
+  MODEL_NAME=$(select_ollama_model "$OLLAMA_URL" "$MODEL_NAME")
+  echo ""
+  log_info "Selected model: ${CYAN}$MODEL_NAME${NC}"
+  echo ""
+fi
+
+# Ensure model is available
+ensure_ollama_model "$OLLAMA_URL" "$MODEL_NAME" || exit 1
+echo ""
+
+# ============================================================================
+# Read and Analyze CLAUDE.md
+# ============================================================================
+
+log_section "📊 Analyzing CLAUDE.md"
+
+if [ "$DIFF_MODE" = true ]; then
+  # Diff mode: compare CLAUDE.md between branches
+  log_info "Getting CLAUDE.md diff between branches..."
+
+  CURRENT_BRANCH=$(get_current_branch)
+
+  # Get diff for CLAUDE.md
+  CLAUDE_DIFF=$(git diff "$BASE_BRANCH...HEAD" -- CLAUDE.md 2>/dev/null || git diff "$BASE_BRANCH..HEAD" -- CLAUDE.md 2>/dev/null || echo "")
+
+  if [ -z "$CLAUDE_DIFF" ]; then
+    log_warning "No changes to CLAUDE.md between $BASE_BRANCH and $CURRENT_BRANCH"
+    exit 0
+  fi
+
+  DIFF_LENGTH=${#CLAUDE_DIFF}
+  DIFF_LINES=$(echo "$CLAUDE_DIFF" | wc -l | tr -d ' ')
+  log_success "CLAUDE.md diff generated ($DIFF_LINES lines, $DIFF_LENGTH chars)"
+
+  # Get current version
+  CLAUDE_CONTENT=$(cat CLAUDE.md 2>/dev/null || echo "CLAUDE.md not found in current branch")
+  CLAUDE_LENGTH=${#CLAUDE_CONTENT}
+
+  echo ""
+else
+  # Optimization mode: analyze current CLAUDE.md
+  if [ ! -f "CLAUDE.md" ]; then
+    log_error "CLAUDE.md not found in current directory"
+    exit 1
+  fi
+
+  log_info "Reading CLAUDE.md..."
+  CLAUDE_CONTENT=$(cat CLAUDE.md)
+  CLAUDE_LENGTH=${#CLAUDE_CONTENT}
+  CLAUDE_LINES=$(echo "$CLAUDE_CONTENT" | wc -l | tr -d ' ')
+  log_success "CLAUDE.md loaded ($CLAUDE_LINES lines, $CLAUDE_LENGTH chars)"
+  echo ""
+fi
+
+# ============================================================================
+# Build Optimization-Focused System Prompt
+# ============================================================================
+
+log_info "Building system prompt..."
+
+if [ "$DIFF_MODE" = true ]; then
+  SYSTEM_PROMPT=$(cat <<'SYSEND'
+You are an expert technical writer reviewing changes to project documentation (CLAUDE.md).
+
+**Your Task:**
+Analyze the changes made to CLAUDE.md between two branches and provide:
+1. Summary of what changed and why
+2. Assessment of whether changes improve or degrade the documentation
+3. Recommendations for improvement
+4. Token impact analysis (did changes increase/decrease token usage?)
+
+**Focus Areas:**
+- Clarity and conciseness of new/modified content
+- Organization and structure changes
+- Addition or removal of important information
+- Impact on token usage
+- Consistency with existing style
+- Potential redundancy or verbosity
+
+**Output Format:**
+
+## Summary
+[Brief description of changes made]
+
+## Changes Analysis
+[Detailed breakdown of key changes]
+
+## Quality Assessment
+[Are changes an improvement? Any concerns?]
+
+## Token Impact
+[Did changes increase/decrease token usage? By how much?]
+
+## Recommendations
+[Specific suggestions for improvement, if any]
+
+**Guidelines:**
+- Be CONSTRUCTIVE and SPECIFIC
+- Focus on significant changes, not minor edits
+- Provide actionable feedback
+- Consider both content quality and token efficiency
+SYSEND
+)
+else
+  SYSTEM_PROMPT=$(cat <<'SYSEND'
+You are an expert technical writer and AI prompt engineer specializing in optimizing documentation for LLM token efficiency.
+
+**Your Task:**
+Analyze CLAUDE.md and provide an optimized version that:
+1. Reduces token usage by 20-40% while maintaining ALL critical information
+2. Eliminates redundancy and verbosity
+3. Uses more concise language and formatting
+4. Improves organization for better LLM comprehension
+5. Maintains all essential technical details and instructions
+
+**Optimization Strategies:**
+
+1. **Eliminate Redundancy:**
+   - Remove repeated information
+   - Consolidate similar sections
+   - Use references instead of duplication
+
+2. **Concise Language:**
+   - Replace verbose phrases with concise equivalents
+   - Use bullet points instead of paragraphs where appropriate
+   - Remove filler words and unnecessary explanations
+
+3. **Efficient Formatting:**
+   - Use tables for structured data
+   - Use code blocks only when necessary
+   - Use shorter section names
+   - Use abbreviations for commonly repeated terms (define once)
+
+4. **Better Organization:**
+   - Group related information
+   - Use clear hierarchy
+   - Front-load most important information
+   - Move less critical details to appendices
+
+5. **Smart Compression:**
+   - Combine examples where possible
+   - Use "e.g." instead of listing all examples
+   - Reference external docs for detailed info
+   - Use "See X" for cross-references
+
+**IMPORTANT:**
+- DO NOT remove critical technical information
+- DO NOT change code examples significantly
+- DO NOT remove security guidelines
+- DO maintain the same level of detail, just more concisely
+
+**Output Format:**
+
+## Analysis Summary
+- Current size: [lines, characters, estimated tokens]
+- Optimization opportunities: [brief list]
+- Estimated savings: [X% reduction]
+
+## Optimization Recommendations
+[Numbered list of specific changes to make, with examples]
+
+## Optimized CLAUDE.md
+[Full optimized version between these markers:]
+```markdown
+[OPTIMIZED_START]
+[... full optimized content ...]
+[OPTIMIZED_END]
+```
+
+## Token Savings Breakdown
+- Section 1: X tokens saved
+- Section 2: Y tokens saved
+- Total: Z tokens saved (~XX% reduction)
+
+**Guidelines:**
+- Be AGGRESSIVE with optimization but PRESERVE all critical info
+- Aim for 20-40% reduction in length
+- Maintain readability and clarity
+- Keep the same structure but make it tighter
+SYSEND
+)
+fi
+
+log_success "System prompt built"
+echo ""
+
+# ============================================================================
+# Build User Prompt
+# ============================================================================
+
+log_info "Building analysis prompt..."
+
+if [ "$DIFF_MODE" = true ]; then
+  USER_PROMPT=$(cat <<USEREND
+Please analyze the following changes to CLAUDE.md.
+
+**Branch Comparison:** $CURRENT_BRANCH vs $BASE_BRANCH
+
+**Changes to CLAUDE.md:**
+\`\`\`diff
+$CLAUDE_DIFF
+\`\`\`
+
+**Current CLAUDE.md (for context):**
+\`\`\`markdown
+${CLAUDE_CONTENT:0:5000}
+... (truncated for brevity)
+\`\`\`
+
+Please provide a thorough analysis of these changes.
+USEREND
+)
+else
+  USER_PROMPT=$(cat <<USEREND
+Please analyze and optimize the following CLAUDE.md file.
+
+Current statistics:
+- Lines: $CLAUDE_LINES
+- Characters: $CLAUDE_LENGTH
+- Estimated tokens: ~$((CLAUDE_LENGTH / 4))
+
+# Current CLAUDE.md Content
+
+\`\`\`markdown
+$CLAUDE_CONTENT
+\`\`\`
+
+Please provide an optimized version that reduces token usage by 20-40% while maintaining all critical information.
+USEREND
+)
+fi
+
+log_success "Analysis prompt built"
+echo ""
+
+# ============================================================================
+# Call Ollama API for Optimization
+# ============================================================================
+
+log_info "Calling Ollama API for optimization analysis..."
+echo -e "${CYAN}This may take a few minutes (large document)...${NC}"
+echo ""
+
+# Use more tokens for this analysis since we need a full rewrite
+# num_ctx: 16384 (large context for full document)
+# num_predict: 4000 (large output for optimized version)
+RESULT=$(call_ollama_chat "$OLLAMA_URL" "$MODEL_NAME" "$SYSTEM_PROMPT" "$USER_PROMPT" 16384 4000)
+
+# ============================================================================
+# Process Results
+# ============================================================================
+
+if [ "$DIFF_MODE" = true ]; then
+  # Diff mode: save analysis only
+  log_info "Processing diff analysis..."
+
+  ANALYSIS_FILE=$(save_review "$RESULT" "docs/ai-reviews" "claude-md-diff")
+
+  log_success "Diff analysis saved to: ${CYAN}$ANALYSIS_FILE${NC}"
+  echo ""
+
+  # Display the analysis
+  ADDITIONAL_INFO="${CYAN}Branch:${NC} $CURRENT_BRANCH vs $BASE_BRANCH
+${CYAN}Changes:${NC} $DIFF_LINES lines, $DIFF_LENGTH chars"
+
+  display_review "$RESULT" "CLAUDE.md Diff Analysis" "$MODEL_NAME" "$ADDITIONAL_INFO"
+
+  log_success "Diff analysis complete!"
+  exit 0
+fi
+
+# ============================================================================
+# Extract Optimized Version (Optimization Mode Only)
+# ============================================================================
+
+log_info "Extracting optimized version..."
+
+# Try to extract the optimized content between markers
+OPTIMIZED_CONTENT=$(echo "$RESULT" | sed -n '/\[OPTIMIZED_START\]/,/\[OPTIMIZED_END\]/p' | sed '1d;$d')
+
+if [ -z "$OPTIMIZED_CONTENT" ]; then
+  log_warning "Could not extract optimized content from markers, using full response"
+  OPTIMIZED_CONTENT="$RESULT"
+fi
+
+# Save the optimized version
+echo "$OPTIMIZED_CONTENT" > "$OUTPUT_FILE"
+OPTIMIZED_LENGTH=${#OPTIMIZED_CONTENT}
+OPTIMIZED_LINES=$(echo "$OPTIMIZED_CONTENT" | wc -l | tr -d ' ')
+log_success "Optimized version saved to: ${CYAN}$OUTPUT_FILE${NC}"
+
+# Calculate savings
+CHAR_SAVINGS=$((CLAUDE_LENGTH - OPTIMIZED_LENGTH))
+PERCENT_SAVINGS=$((CHAR_SAVINGS * 100 / CLAUDE_LENGTH))
+
+echo ""
+log_info "Optimization Results:"
+echo -e "  ${CYAN}Original:${NC}  $CLAUDE_LINES lines, $CLAUDE_LENGTH chars, ~$((CLAUDE_LENGTH / 4)) tokens"
+echo -e "  ${CYAN}Optimized:${NC} $OPTIMIZED_LINES lines, $OPTIMIZED_LENGTH chars, ~$((OPTIMIZED_LENGTH / 4)) tokens"
+echo -e "  ${GREEN}Savings:${NC}   $((CLAUDE_LINES - OPTIMIZED_LINES)) lines, $CHAR_SAVINGS chars, ~$((CHAR_SAVINGS / 4)) tokens (${PERCENT_SAVINGS}%)"
+echo ""
+
+# Display the full analysis
+ADDITIONAL_INFO="${CYAN}Original:${NC} $CLAUDE_LINES lines, $CLAUDE_LENGTH chars
+${CYAN}Optimized:${NC} $OPTIMIZED_LINES lines, $OPTIMIZED_LENGTH chars
+${CYAN}Savings:${NC} ${PERCENT_SAVINGS}% reduction"
+
+display_review "$RESULT" "CLAUDE.md Optimization Analysis" "$MODEL_NAME" "$ADDITIONAL_INFO"
+
+# Save analysis to file
+ANALYSIS_FILE=$(save_review "$RESULT" "docs/ai-reviews" "claude-md-optimization")
+log_success "Analysis saved to: ${CYAN}$ANALYSIS_FILE${NC}"
+echo ""
+
+# Apply changes if requested
+if [ "$APPLY_CHANGES" = true ]; then
+  log_info "Applying changes to CLAUDE.md..."
+
+  # Create backup
+  BACKUP_FILE="CLAUDE.md.backup-$(date +%Y%m%d-%H%M%S)"
+  cp CLAUDE.md "$BACKUP_FILE"
+  log_success "Backup created: ${CYAN}$BACKUP_FILE${NC}"
+
+  # Apply changes
+  cp "$OUTPUT_FILE" CLAUDE.md
+  log_success "Changes applied to CLAUDE.md"
+
+  echo ""
+  log_warning "Please review the changes carefully before committing!"
+  log_info "Original backed up to: $BACKUP_FILE"
+  log_info "You can restore with: cp $BACKUP_FILE CLAUDE.md"
+fi
+
+echo ""
+log_success "Optimization complete!"
+log_info "Review the optimized version at: ${CYAN}$OUTPUT_FILE${NC}"
+
+if [ "$APPLY_CHANGES" != true ]; then
+  log_info "To apply changes: bin/ai-optimize-claude-md --apply"
+fi
+
+exit 0