n2b 0.5.1 → 2.0.0

data/lib/n2b/merge_cli.rb

@@ -0,0 +1,2329 @@
+require 'shellwords'
+require 'rbconfig'
+require 'optparse'
+require 'fileutils'
+require 'stringio'
+require 'cgi'
+require_relative 'base'
+require_relative 'merge_conflict_parser'
+require_relative 'jira_client'
+require_relative 'github_client'
+require_relative 'message_utils'
+
+module N2B
+  class MergeCLI < Base
+    COLOR_RED   = "\e[31m"
+    COLOR_GREEN = "\e[32m"
+    COLOR_YELLOW= "\e[33m"
+    COLOR_BLUE  = "\e[34m"
+    COLOR_GRAY  = "\e[90m"
+    COLOR_RESET = "\e[0m"
+
+    def self.run(args)
+      new(args).execute
+    end
+
+    def initialize(args)
+      @args = args
+      @options = parse_options
+      # @file_path = @args.shift # Moved to execute based on mode
+    end
+
+    def execute
+      config = get_config(reconfigure: false, advanced_flow: false)
+
+      if @options[:analyze]
+        # In analyze mode, @args might be used for custom prompt additions later,
+        # similar to how original cli.rb handled it.
+        # For now, custom_message option is the primary way.
+        handle_diff_analysis(config)
+      else
+        @file_path = @args.shift
+        if @file_path.nil?
+          show_help_and_status # Renamed from show_usage_and_unresolved
+          exit 1
+        end
+
+        unless File.exist?(@file_path)
+          puts "File not found: #{@file_path}"
+          exit 1
+        end
+
+        parser = MergeConflictParser.new(context_lines: @options[:context_lines])
+      blocks = parser.parse(@file_path)
+      if blocks.empty?
+        puts "No merge conflicts found."
+        return
+      end
+
+      lines = File.readlines(@file_path, chomp: true)
+      log_entries = []
+      aborted = false
+
+      blocks.reverse_each do |block|
+        result = resolve_block(block, config, lines.join("\n"))
+        log_entries << result.merge({
+          base_content: block.base_content,
+          incoming_content: block.incoming_content,
+          base_label: block.base_label,
+          incoming_label: block.incoming_label,
+          start_line: block.start_line,
+          end_line: block.end_line,
+          resolved_content: result[:merged_code],
+          llm_suggestion: result[:reason],
+          resolution_method: determine_resolution_method(result),
+          action: determine_action(result),
+          timestamp: Time.now.strftime('%Y-%m-%d %H:%M:%S UTC')
+        })
+        if result[:abort]
+          aborted = true
+          break
+        elsif result[:accepted]
+          replacement = result[:merged_code].to_s.split("\n")
+          lines[(block.start_line-1)...block.end_line] = replacement
+        end
+      end
+
+      unless aborted
+        File.write(@file_path, lines.join("\n") + "\n")
+
+        # Show summary
+        accepted_count = log_entries.count { |entry| entry[:accepted] }
+        skipped_count = log_entries.count { |entry| !entry[:accepted] && !entry[:abort] }
+
+        puts "\n#{COLOR_BLUE}📊 Resolution Summary:#{COLOR_RESET}"
+        puts "#{COLOR_GREEN}✅ Accepted: #{accepted_count}#{COLOR_RESET}"
+        puts "#{COLOR_YELLOW}⏭️  Skipped: #{skipped_count}#{COLOR_RESET}" if skipped_count > 0
+
+        # Only auto-mark as resolved if ALL conflicts were accepted (none skipped)
+        if accepted_count > 0 && skipped_count == 0
+          mark_file_as_resolved(@file_path)
+          puts "#{COLOR_GREEN}🎉 All conflicts resolved! File marked as resolved in VCS.#{COLOR_RESET}"
+        elsif accepted_count > 0 && skipped_count > 0
+          puts "#{COLOR_YELLOW}⚠️  Some conflicts were skipped - file NOT marked as resolved#{COLOR_RESET}"
+          puts "#{COLOR_GRAY}💡 Resolve remaining conflicts or manually mark: hg resolve --mark #{@file_path}#{COLOR_RESET}"
+        else
+          puts "#{COLOR_YELLOW}⚠️  No conflicts were accepted - file NOT marked as resolved#{COLOR_RESET}"
+        end
+      else
+        puts "\n#{COLOR_YELLOW}⚠️  Resolution aborted - no changes made#{COLOR_RESET}"
+      end
+
+      if config['merge_log_enabled'] && log_entries.any?
+        dir = '.n2b_merge_log'
+        FileUtils.mkdir_p(dir)
+        timestamp = Time.now.strftime('%Y-%m-%d-%H%M%S')
+        log_path = File.join(dir, "#{timestamp}.html")
+        html_content = generate_merge_log_html(log_entries, timestamp)
+        File.write(log_path, html_content)
+        puts "#{COLOR_GRAY}📝 Merge log saved to #{log_path}#{COLOR_RESET}"
+      end
+      end
+    end
+
+    private
+
+    def parse_options
+      options = {
+        context_lines: MergeConflictParser::DEFAULT_CONTEXT_LINES,
+        analyze: false,
+        branch: nil,
+        jira_ticket: nil,
+        github_issue: nil,
+        requirements_file: nil,
+        update_issue: nil, # nil means ask, true means update, false means no update
+        custom_message: nil
+      }
+      parser = OptionParser.new do |opts|
+        opts.banner = "Usage: n2b-diff FILE [options] OR n2b-diff --analyze [options]"
+        opts.separator ""
+        opts.separator "Merge Conflict Options:"
+        opts.on('--context N', Integer, 'Context lines for merge conflict analysis (default: 10, affects LLM context)') { |v| options[:context_lines] = v }
+
+        opts.separator ""
+        opts.separator "Diff Analysis Options:"
+        opts.on('-a', '--analyze', 'Analyze git/hg diff with AI') { options[:analyze] = true }
+        opts.on('--branch [BRANCH]', 'Branch to compare against for analysis (default: auto-detect main/master)') do |branch|
+          options[:branch] = branch || 'auto' # 'auto' will trigger detection logic
+        end
+        opts.on('-j', '--jira JIRA_ID_OR_URL', 'Jira ticket ID or URL for context or update') do |jira|
+          options[:jira_ticket] = jira
+        end
+        opts.on('--github GITHUB_ISSUE_ID_OR_URL', 'GitHub issue ID or URL for context or update (e.g., owner/repo/issues/123)') do |gh|
+          options[:github_issue] = gh
+        end
+        opts.on('-r', '--requirements FILEPATH', 'Requirements file for diff analysis context') do |file|
+          options[:requirements_file] = file
+        end
+        opts.on('--update', 'Attempt to update the linked Jira/GitHub issue with the analysis result (will ask for confirmation by default)') do
+          options[:update_issue] = true # Explicitly true
+        end
+        opts.on('--no-update', 'Do not attempt to update the linked Jira/GitHub issue') do
+          options[:update_issue] = false # Explicitly false
+        end
+        opts.on('-m', '--message MESSAGE', '--msg MESSAGE', String, 'Custom instructions for AI analysis (max 500 chars)') do |raw_msg|
+          validated_msg = N2B::MessageUtils.validate_message(raw_msg)
+          # Sanitization happens after validation (e.g. truncation)
+          # No need to log here, will log after all options parsed if message exists
+          options[:custom_message] = validated_msg # Store potentially truncated message
+        end
+
+        opts.separator ""
+        opts.separator "Common Options:"
+        opts.on_tail('-h', '--help', 'Show this help message') { puts opts; exit }
+        opts.on_tail('-v', '--version', 'Show version') { puts N2B::VERSION; exit } # Assuming VERSION is defined
+
+        opts.separator ""
+        opts.separator "Examples:"
+        opts.separator "  n2b-diff path/to/your/file_with_conflicts.rb"
+        opts.separator "  n2b-diff --analyze"
+        opts.separator "  n2b-diff --analyze --branch main"
+        opts.separator "  n2b-diff --analyze --jira PROJ-123 -m \"Focus on data validation.\""
+        opts.separator "  n2b-diff --analyze --github org/repo/issues/42 --update"
+      end
+
+      begin
+        parser.parse!(@args)
+      rescue OptionParser::InvalidOption => e
+        puts "#{COLOR_RED}Error: #{e.message}#{COLOR_RESET}"
+        puts parser.help
+        exit 1
+      end
+
+      # --- Option Validations ---
+      if options[:analyze]
+        # Options that only make sense with analyze
+      else
+        # Options that don't make sense without analyze (if any)
+        if options[:branch] || options[:jira_ticket] || options[:github_issue] || options[:requirements_file] || options[:custom_message] || !options[:update_issue].nil?
+          puts "#{COLOR_RED}Error: Options like --branch, --jira, --github, --requirements, --message, --update/--no-update are only for --analyze mode.#{COLOR_RESET}"
+          puts parser.help
+          exit 1
+        end
+      end
+
+      if options[:update_issue] == true && options[:jira_ticket].nil? && options[:github_issue].nil?
+        puts "#{COLOR_RED}Error: --update option requires --jira or --github to be specified for context.#{COLOR_RESET}"
+        puts parser.help
+        exit 1
+      end
+
+      # If --no-update is used without specifying a ticket, it's not an error, just has no effect.
+      # It's fine for --jira or --github to be specified without --update or --no-update (implies ask user)
+
+      # Perform sanitization and logging for custom_message after all options are parsed.
+      if options[:custom_message]
+        # Sanitize the already validated (potentially truncated) message
+        sanitized_message = N2B::MessageUtils.sanitize_message(options[:custom_message])
+        options[:custom_message] = sanitized_message # Store the sanitized version
+
+        # Log the final message that will be used
+        N2B::MessageUtils.log_message("Using custom analysis message: \"#{options[:custom_message]}\"", :info) unless options[:custom_message].strip.empty?
+      end
+
+      options
+    end
+
+    # --- Methods moved and adapted from original N2B::CLI for diff analysis ---
+    def handle_diff_analysis(config)
+      vcs_type = get_vcs_type
+      if vcs_type == :none
+        puts "Error: Not a git or hg repository. Diff analysis requires a VCS."
+        exit 1
+      end
+
+      requirements_filepath = @options[:requirements_file] # Adapted option name
+      # user_prompt_addition now directly uses the processed @options[:custom_message]
+      user_prompt_addition = @options[:custom_message] || ""
+
+      # Logging of custom_message is now handled in parse_options by MessageUtils.log_message
+      # So, the specific puts here can be removed or changed to debug level if needed.
+      # For now, removing:
+      # if @options[:custom_message] && !@options[:custom_message].strip.empty?
+      #   puts "DEBUG: Custom message for analysis: #{@options[:custom_message]}" # Or use MessageUtils.log_message if preferred for debug
+      # end
+
+      # Ticket / Issue Information - Standardize to use jira_ticket and github_issue
+      # The original cli.rb used @options[:jira_ticket] for both. Here we differentiate.
+      ticket_input = @options[:jira_ticket] || @options[:github_issue]
+      ticket_type = @options[:jira_ticket] ? 'jira' : (@options[:github_issue] ? 'github' : nil)
+      # ticket_update_flag from @options[:update_issue] (true, false, or nil)
+      # nil means ask, true means update, false means no update.
+      ticket_update_flag = @options[:update_issue]
+
+
+      requirements_content = nil
+
+      if ticket_input && ticket_type
+        # Determine which issue tracker based on which option was provided
+        # Default to 'jira' if somehow ticket_type is nil but ticket_input exists (should not happen)
+        tracker_service_name = ticket_type || (config['issue_tracker'] || 'jira')
+
+        case tracker_service_name
+        when 'github'
+          puts "GitHub issue specified: #{ticket_input}"
+          if config['github'] && config['github']['repo'] && config['github']['access_token']
+            begin
+              github_client = N2B::GitHubClient.new(config) # Ensure N2B::GitHubClient is available
+              puts "Fetching GitHub issue details..."
+              requirements_content = github_client.fetch_issue(ticket_input) # ticket_input here is ID/URL
+              puts "Successfully fetched GitHub issue details."
+            rescue StandardError => e
+              puts "Error fetching GitHub issue: #{e.message}"
+              puts "Proceeding with diff analysis without GitHub issue details."
+            end
+          else
+            puts "GitHub configuration is missing or incomplete in N2B settings."
+            puts "Please configure GitHub using 'n2b -c' (or main n2b config) to fetch issue details."
+            puts "Proceeding with diff analysis without GitHub issue details."
+          end
+        when 'jira'
+          puts "Jira ticket specified: #{ticket_input}"
+          if config['jira'] && config['jira']['domain'] && config['jira']['email'] && config['jira']['api_key']
+            begin
+              jira_client = N2B::JiraClient.new(config) # Ensure N2B::JiraClient is available
+              puts "Fetching Jira ticket details..."
+              requirements_content = jira_client.fetch_ticket(ticket_input)
+              puts "Successfully fetched Jira ticket details."
+            rescue N2B::JiraClient::JiraApiError => e
+              puts "Error fetching Jira ticket: #{e.message}"
+              puts "Proceeding with diff analysis without Jira ticket details."
+            rescue ArgumentError => e # Catches config errors from JiraClient init
+              puts "Jira configuration error: #{e.message}"
+              puts "Please ensure Jira is configured correctly."
+              puts "Proceeding with diff analysis without Jira ticket details."
+            rescue StandardError => e
+              puts "An unexpected error occurred while fetching Jira ticket: #{e.message}"
+              puts "Proceeding with diff analysis without Jira ticket details."
+            end
+          else
+            puts "Jira configuration is missing or incomplete in N2B settings."
+            puts "Please configure Jira using 'n2b -c' (or main n2b config) to fetch ticket details."
+            puts "Proceeding with diff analysis without Jira ticket details."
+          end
+        end
+        # Common message for ticket update status based on new flag
+        if ticket_update_flag == true
+            puts "Note: Issue update is flagged (--update)."
+        elsif ticket_update_flag == false
+            puts "Note: Issue will not be updated (--no-update)."
+        else # nil case
+            puts "Note: You will be prompted whether to update the issue after analysis."
+        end
+      end
+
+
+      if requirements_content.nil? && requirements_filepath
+        if File.exist?(requirements_filepath)
+          puts "Loading requirements from file: #{requirements_filepath}"
+          requirements_content = File.read(requirements_filepath)
+        else
+          puts "Error: Requirements file not found: #{requirements_filepath}"
+          puts "Proceeding with diff analysis without file-based requirements."
+        end
+      elsif requirements_content && requirements_filepath
+        puts "Note: Both issue details and a requirements file were provided. Using issue details for analysis context."
+      end
+
+      diff_output = execute_vcs_diff(vcs_type, @options[:branch])
+      if diff_output.nil? || diff_output.strip.empty?
+        puts "No differences found to analyze."
+        exit 0
+      end
+
+      # Pass user_prompt_addition (custom_message) to analyze_diff
+      analysis_result = analyze_diff(diff_output, config, user_prompt_addition, requirements_content)
+
+      # --- Ticket Update Logic (adapted) ---
+      if ticket_input && ticket_type && analysis_result && !analysis_result.empty?
+        tracker_service_name = ticket_type
+
+        # Determine if we should proceed with update based on ticket_update_flag
+        proceed_with_update_action = false
+        if ticket_update_flag == true # --update
+          proceed_with_update_action = true
+        elsif ticket_update_flag.nil? # Ask user
+          puts "\nWould you like to update #{tracker_service_name.capitalize} issue #{ticket_input} with this analysis? (y/n)"
+          user_choice = $stdin.gets.chomp.downcase
+          proceed_with_update_action = user_choice == 'y'
+        end # If ticket_update_flag is false, proceed_with_update_action remains false
+
+        if proceed_with_update_action
+          case tracker_service_name
+          when 'github'
+            if config['github'] && config['github']['repo'] && config['github']['access_token']
+              # Pass custom_message to formatting function
+              comment_data = format_analysis_for_github(analysis_result, @options[:custom_message])
+              begin
+                update_client = N2B::GitHubClient.new(config)
+                puts "Updating GitHub issue #{ticket_input}..."
+                if update_client.update_issue(ticket_input, comment_data) # ticket_input is ID/URL
+                  puts "GitHub issue #{ticket_input} updated successfully."
+                else
+                  puts "Failed to update GitHub issue #{ticket_input}."
+                end
+              rescue StandardError => e
+                puts "Error updating GitHub issue: #{e.message}"
+              end
+            else
+              puts "GitHub configuration is missing. Cannot update GitHub issue."
+            end
+          when 'jira'
+            if config['jira'] && config['jira']['domain'] && config['jira']['email'] && config['jira']['api_key']
+              # Pass custom_message to formatting function
+              jira_comment_data = format_analysis_for_jira(analysis_result, @options[:custom_message])
+              begin
+                update_jira_client = N2B::JiraClient.new(config)
+                puts "Updating Jira ticket #{ticket_input}..."
+                if update_jira_client.update_ticket(ticket_input, jira_comment_data)
+                  puts "Jira ticket #{ticket_input} updated successfully."
+                else
+                  puts "Failed to update Jira ticket #{ticket_input}."
+                end
+              rescue N2B::JiraClient::JiraApiError => e
+                puts "Error updating Jira ticket: #{e.message}"
+              rescue ArgumentError => e
+                puts "Jira configuration error for update: #{e.message}"
+              rescue StandardError => e
+                puts "An unexpected error occurred while updating Jira ticket: #{e.message}"
+              end
+            else
+              puts "Jira configuration is missing. Cannot update Jira ticket."
+            end
+          end
+        else
+          puts "Issue/Ticket update skipped."
+        end
+      elsif ticket_input && (analysis_result.nil? || analysis_result.empty?)
+        puts "Skipping ticket update as analysis result was empty or not generated."
+      end
+      # --- End of Ticket Update Logic ---
+      analysis_result # Return for potential further use or testing
+    end
+
+    def get_vcs_type
+      if Dir.exist?(File.join(Dir.pwd, '.git'))
+        :git
+      elsif Dir.exist?(File.join(Dir.pwd, '.hg'))
+        :hg
+      else
+        :none
+      end
+    end
+
+    def execute_vcs_diff(vcs_type, branch_option = nil)
+      # Ensure shell commands are properly escaped if branch_option comes from user input
+      # Though OptionParser should handle this for fixed options.
+      # For this method, branch_option is generally safe as it's from parsed options.
+
+      case vcs_type
+      when :git
+        target_branch_name = branch_option == 'auto' ? detect_git_default_branch : branch_option
+        if target_branch_name
+          unless validate_git_branch_exists(target_branch_name)
+            puts "Error: Git branch '#{target_branch_name}' does not exist or is not accessible."
+            # Suggest available branches if possible (might be noisy, consider a debug flag)
+            # puts "Available local branches:\n`git branch`"
+            # puts "Available remote branches for 'origin':\n`git branch -r`"
+            exit 1
+          end
+          puts "Comparing current HEAD against git branch '#{target_branch_name}'..."
+          # Use three dots for "changes on your branch since target_branch_name diverged from it"
+          # Or two dots for "changes between target_branch_name and your current HEAD"
+          # Three dots is common for PR-like diffs.
+          `git diff #{Shellwords.escape(target_branch_name)}...HEAD`
+        else
+          puts "Could not automatically detect default git branch and no branch specified. Falling back to 'git diff HEAD' (staged changes)..."
+          `git diff HEAD` # This shows staged changes. `git diff` shows unstaged.
+        end
+      when :hg
+        target_branch_name = branch_option == 'auto' ? detect_hg_default_branch : branch_option
+        if target_branch_name
+          unless validate_hg_branch_exists(target_branch_name)
+            puts "Error: Mercurial branch '#{target_branch_name}' does not exist."
+            exit 1
+          end
+          puts "Comparing current working directory against hg branch '#{target_branch_name}'..."
+          # For hg, diff against the revision of the branch tip
+          `hg diff -r #{Shellwords.escape(target_branch_name)}`
+        else
+          puts "Could not automatically detect default hg branch and no branch specified. Falling back to 'hg diff' (uncommitted changes)..."
+          `hg diff` # Shows uncommitted changes.
+        end
+      else
+        puts "Error: Unsupported VCS type." # Should be caught by get_vcs_type check earlier
+        ""
+      end
+    end
+
+    def detect_git_default_branch
+      # Method 1: Check origin/HEAD symbolic ref (most reliable for remotes)
+      result = `git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null`.strip
+      return result.split('/').last if $?.success? && !result.empty?
+
+      # Method 2: Check local 'main' or 'master' if they track a remote branch
+      ['main', 'master'].each do |branch|
+        local_tracking = `git rev-parse --abbrev-ref #{branch}@{upstream} 2>/dev/null`.strip
+        return branch if $?.success? && !local_tracking.empty?
+      end
+
+      # Method 3: Check remote show origin for HEAD branch (less direct but good fallback)
+      result = `git remote show origin 2>/dev/null | grep "HEAD branch"`.strip
+      if $?.success? && !result.empty?
+        match = result.match(/HEAD branch:\s*(\S+)/)
+        return match[1] if match && match[1] != '(unknown)'
+      end
+
+      # Method 4: Last resort, check for local main/master even if not tracking
+      ['main', 'master'].each do |branch|
+        return branch if system("git show-ref --verify --quiet refs/heads/#{branch}")
+      end
+      nil # No default branch found
+    end
+
+    def detect_hg_default_branch
+      # Mercurial typically uses 'default' as the main branch.
+      # `hg branch` shows the current branch. If it's 'default', that's a strong candidate.
+      current_branch = `hg branch 2>/dev/null`.strip
+      return current_branch if $?.success? && current_branch == 'default'
+
+      # Check if 'default' branch exists in the list of branches
+      branches_output = `hg branches 2>/dev/null`
+      if $?.success? && branches_output.lines.any? { |line| line.strip.start_with?('default ') }
+        return 'default'
+      end
+
+      # Fallback if 'default' isn't found but there's only one branch (common in simple repos)
+      if $?.success?
+          active_branches = branches_output.lines.reject { |line| line.include?('(inactive)') }
+          if active_branches.size == 1
+              return active_branches.first.split.first
+          end
+      end
+
+      # If current branch is not 'default' but exists, it might be the one being used as main.
+      return current_branch if $?.success? && !current_branch.empty?
+
+      'default' # Default assumption for Mercurial
+    end
+
+    def validate_git_branch_exists(branch)
+      # Check local branches
+      return true if system("git show-ref --verify --quiet refs/heads/#{Shellwords.escape(branch)}")
+      # Check remote branches (assuming 'origin' remote)
+      return true if system("git show-ref --verify --quiet refs/remotes/origin/#{Shellwords.escape(branch)}")
+      false
+    end
+
+    def validate_hg_branch_exists(branch)
+      # `hg branches` lists all branches. Check if the branch is in the list.
+      # `hg id -r branchname` would fail if branch doesn't exist.
+      system("hg id -r #{Shellwords.escape(branch)} > /dev/null 2>&1")
+    end
+
+    def build_diff_analysis_prompt(diff_output, user_prompt_addition = "", requirements_content = nil, config = {})
+      default_system_prompt_path = resolve_template_path('diff_system_prompt', config) # Ensure this template exists
+      default_system_prompt = File.read(default_system_prompt_path).strip rescue "Analyze this diff." # Fallback
+
+      # user_prompt_addition is @options[:custom_message], already validated and sanitized.
+      user_instructions_section = ""
+      if user_prompt_addition && !user_prompt_addition.to_s.strip.empty?
+        # No need to call .strip here again as sanitize_message in parse_options already does it.
+        user_instructions_section = "User's Custom Instructions:\n#{user_prompt_addition}\n\n---\n\n"
+      end
+
+      requirements_section = ""
+      if requirements_content && !requirements_content.to_s.strip.empty?
+        requirements_section = <<-REQUIREMENTS_BLOCK
+CRITICAL REQUIREMENTS EVALUATION:
+Based on the provided context, evaluate if the code changes meet these requirements.
+For each requirement, explicitly state whether it is:
+- ✅ IMPLEMENTED
+- ⚠️ PARTIALLY IMPLEMENTED
+- ❌ NOT IMPLEMENTED
+- 🔍 UNCLEAR
+
+--- BEGIN REQUIREMENTS ---
+#{requirements_content.strip}
+--- END REQUIREMENTS ---
+
+REQUIREMENTS_BLOCK
+      end
+
+      analysis_intro = "Analyze the following code diff. Focus on identifying potential bugs, suggesting improvements, and assessing test coverage. If requirements are provided, evaluate against them."
+
+      # Context extraction (can be intensive, consider making optional or smarter)
+      # context_sections = extract_code_context_from_diff(diff_output) # This can be slow
+      # context_info = context_sections.empty? ? "" : "\n\nRelevant Code Context:\n#{format_context_for_prompt(context_sections)}"
+      # For now, let's keep context extraction simpler or rely on LLM's ability with just the diff.
+      # If extract_code_context_from_diff is too complex or slow, it might be omitted or simplified.
+      # For this integration, assuming extract_code_context_from_diff is efficient enough.
+      context_sections = extract_code_context_from_diff(diff_output)
+      context_info = ""
+      unless context_sections.empty?
+        context_info = "\n\nCurrent Code Context (for better analysis):\n"
+        context_sections.each do |file_path, sections|
+          context_info += "\n--- Context from: #{file_path} ---\n"
+          sections.each do |section|
+            context_info += "Lines approx #{section[:start_line]}-#{section[:end_line]}:\n" # Approximate lines
+            context_info += "```\n#{section[:content]}\n```\n\n"
+          end
+        end
+      end
+
+
+      json_instruction_path = resolve_template_path('diff_json_instruction', config) # Ensure this template exists
+      json_instruction = File.read(json_instruction_path).strip rescue 'Respond in JSON format with keys: "summary", "errors", "improvements", "test_coverage", "requirements_evaluation", "ticket_implementation_summary".' # Fallback
+
+      full_prompt = [
+        user_instructions_section, # Custom instructions first
+        default_system_prompt,
+        requirements_section,
+        analysis_intro,
+        "Diff:\n```diff\n#{diff_output}\n```", # Ensure diff is marked as diff language
+        context_info,
+        json_instruction
+      ].select { |s| s && !s.strip.empty? }.join("\n\n")
+
+      full_prompt
+    end
+
+    def analyze_diff(diff_output, config, user_prompt_addition = "", requirements_content = nil)
+      # user_prompt_addition here IS the custom_message from options
+      prompt = build_diff_analysis_prompt(diff_output, user_prompt_addition, requirements_content, config)
+
+      # Use analyze_diff_with_spinner for this call
+      analysis_json_str = analyze_diff_with_spinner(config) do |llm| # Pass config to spinner wrapper
+        llm.analyze_code_diff(prompt) # This is the actual call the spinner will make
+      end
+      # Fallback if spinner itself has an issue or if LLM call within spinner fails before LlmApiError
+      analysis_json_str ||= '{"summary": "Error: Failed to get analysis from LLM.", "errors": [], "improvements": []}'
+
+
+      begin
+        json_content = extract_json_from_response(analysis_json_str)
+        analysis_result = JSON.parse(json_content)
+
+        # Output formatting (consider making this a separate method or class)
+        puts "\n#{COLOR_BLUE}📊 Code Diff Analysis:#{COLOR_RESET}"
+        puts "-------------------"
+        puts "#{COLOR_GREEN}Summary:#{COLOR_RESET}"
+        puts analysis_result['summary'] || "No summary provided."
+
+        puts "\n#{COLOR_RED}Potential Issues/Errors:#{COLOR_RESET}"
+        errors_list = [analysis_result['errors']].flatten.compact.reject(&:empty?)
+        puts errors_list.any? ? errors_list.map { |err| "- #{err}" }.join("\n") : "  No specific errors identified."
+
+        puts "\n#{COLOR_YELLOW}Suggested Improvements:#{COLOR_RESET}"
+        improvements_list = [analysis_result['improvements']].flatten.compact.reject(&:empty?)
+        puts improvements_list.any? ? improvements_list.map { |imp| "- #{imp}" }.join("\n") : "  No specific improvements suggested."
+
+        puts "\n#{COLOR_BLUE}Test Coverage Assessment:#{COLOR_RESET}"
+        test_coverage = analysis_result['test_coverage']
+        puts test_coverage && !test_coverage.to_s.strip.empty? ? test_coverage : "  No test coverage assessment provided."
+
+        if requirements_content && !requirements_content.to_s.strip.empty?
+          puts "\n#{COLOR_GREEN}Requirements Evaluation:#{COLOR_RESET}"
+          eval_text = analysis_result['requirements_evaluation']
+          puts eval_text && !eval_text.to_s.strip.empty? ? eval_text : "  No requirements evaluation provided."
+        end
+
+        # This field is often more specific for tickets
+        ticket_summary = analysis_result['ticket_implementation_summary']
+        if ticket_summary && !ticket_summary.to_s.strip.empty?
+            puts "\n#{COLOR_GREEN}Ticket Implementation Summary:#{COLOR_RESET}"
+            puts ticket_summary
+        end
+
+        puts "-------------------"
+        return analysis_result
+      rescue JSON::ParserError => e
+        puts "#{COLOR_RED}Critical Error: Failed to parse JSON response for diff analysis: #{e.message}#{COLOR_RESET}"
+        puts "Raw response was: #{analysis_json_str}"
+        # Fallback if parsing fails
+        return {"summary" => "Failed to parse analysis.", "errors" => ["JSON parsing error."], "improvements" => [], "test_coverage" => "Unknown", "requirements_evaluation" => "Unknown", "ticket_implementation_summary" => "Unknown"}
+      end
+    end
+
+    # Extracted from original N2B::CLI, might need slight path adjustments if templates are structured differently
+    # def resolve_template_path(template_key, config)
+    #   user_path = config.dig('templates', template_key) if config.is_a?(Hash)
+    #   return user_path if user_path && File.exist?(user_path)
+    #   # Assuming templates are in a subdir relative to THIS file (merge_cli.rb)
+    #   File.expand_path(File.join(__dir__, 'templates', "#{template_key}.txt"))
+    # end
+
+
+    def format_analysis_for_jira(analysis_result, custom_message = nil)
+      # Basic check, ensure analysis_result is a hash
+      return { body: "Error: Analysis result is not in the expected format." } unless analysis_result.is_a?(Hash)
+
+      summary = analysis_result['ticket_implementation_summary']&.strip || analysis_result['summary']&.strip || "No summary."
+
+      details_parts = []
+      details_parts << "*Custom Analysis Focus:*\n#{custom_message}\n" if custom_message && !custom_message.empty?
+      details_parts << "*Technical Summary:*\n#{analysis_result['summary']&.strip}\n" if analysis_result['summary']
+
+      issues = format_issues_for_adf(analysis_result['errors'])
+      improvements = format_improvements_for_adf(analysis_result['improvements'])
+
+      details_parts << "*Potential Issues:*\n" + (issues.empty? ? "None identified.\n" : issues.map { |i| "- #{i}\n" }.join)
+      details_parts << "*Suggested Improvements:*\n" + (improvements.empty? ? "None identified.\n" : improvements.map { |i| "- #{i}\n" }.join)
+      details_parts << "*Test Coverage Assessment:*\n#{analysis_result['test_coverage']&.strip || "Not assessed."}\n"
+
+      req_eval = analysis_result['requirements_evaluation']&.strip
+      details_parts << "*Requirements Evaluation:*\n#{req_eval}\n" if req_eval && !req_eval.empty?
+
+      # ADF structure for Jira
+      # This is a simplified version. Real ADF is more complex.
+      # N2B::JiraClient would need to construct the actual ADF document.
+      # Here, we are returning a hash that the client can use.
+      {
+        implementation_summary: summary, # Often used as a primary comment or field
+        technical_summary: analysis_result['summary']&.strip,
+        issues: issues, # Array of strings
+        improvements: improvements, # Array of strings
+        test_coverage: analysis_result['test_coverage']&.strip,
+        requirements_evaluation: req_eval,
+        custom_analysis_focus: custom_message # Added new key
+      }
+    end
+
+    def format_issues_for_adf(errors) # Helper for Jira/GitHub formatting
+      return [] unless errors.is_a?(Array) && errors.any?
+      errors.map(&:strip).reject(&:empty?)
+    end
+
+    def format_improvements_for_adf(improvements) # Helper for Jira/GitHub formatting
+      return [] unless improvements.is_a?(Array) && improvements.any?
+      improvements.map(&:strip).reject(&:empty?)
+    end
+
+    def format_analysis_for_github(analysis_result, custom_message = nil)
+      # Basic check
+      return "Error: Analysis result is not in the expected format." unless analysis_result.is_a?(Hash)
+
+      title = "### Code Diff Analysis Summary 🤖\n"
+      summary_section = "**Implementation/Overall Summary:**\n#{analysis_result['ticket_implementation_summary']&.strip || analysis_result['summary']&.strip || "No summary provided."}\n\n"
+
+      details_parts = []
+      details_parts << "**Custom Analysis Focus:**\n#{custom_message}\n" if custom_message && !custom_message.empty?
+      details_parts << "**Technical Detail Summary:**\n#{analysis_result['summary']&.strip}\n" if analysis_result['summary'] && analysis_result['summary'] != analysis_result['ticket_implementation_summary']
+
+
+      errors_list = [analysis_result['errors']].flatten.compact.reject(&:empty?)
+      details_parts << "**Potential Issues/Errors:**\n" + (errors_list.empty? ? "_None identified._\n" : errors_list.map { |err| "- [ ] #{err}\n" }.join) # Added checkbox
+
+      improvements_list = [analysis_result['improvements']].flatten.compact.reject(&:empty?)
+      details_parts << "**Suggested Improvements:**\n" + (improvements_list.empty? ? "_None identified._\n" : improvements_list.map { |imp| "- [ ] #{imp}\n" }.join) # Added checkbox
+
+      details_parts << "**Test Coverage Assessment:**\n#{analysis_result['test_coverage']&.strip || "_Not assessed._"}\n"
+
+      req_eval = analysis_result['requirements_evaluation']&.strip
+      details_parts << "**Requirements Evaluation:**\n#{req_eval.empty? ? "_Not applicable or not assessed._" : req_eval}\n" if req_eval
+
+      # For GitHub, typically a Markdown string is returned for the comment body.
+      # The N2B::GitHubClient would take this string.
+      # Returning a hash that the client can use to build the comment.
+      {
+        title: "Code Diff Analysis Summary 🤖",
+        implementation_summary: analysis_result['ticket_implementation_summary']&.strip || analysis_result['summary']&.strip,
+        technical_summary: analysis_result['summary']&.strip,
+        issues: errors_list, # Array of strings
+        improvements: improvements_list, # Array of strings
+        test_coverage: analysis_result['test_coverage']&.strip,
+        requirements_evaluation: req_eval,
+        custom_analysis_focus: custom_message, # Added new key
+        # Construct a body for convenience, client can override
+        body: title + summary_section + details_parts.join("\n")
+      }
+    end
+
+    def extract_json_from_response(response)
+      # This is a robust way to extract JSON that might be embedded in other text.
+      # First, try to parse as-is, in case the LLM behaves perfectly.
+      begin
+        JSON.parse(response)
+        return response # It's already valid JSON
+      rescue JSON::ParserError
+        # If not, search for the first '{' and last '}'
+      end
+
+      json_start = response.index('{')
+      json_end = response.rindex('}') # Use rindex for the last occurrence
+
+      if json_start && json_end && json_end > json_start
+        potential_json = response[json_start..json_end]
+        begin
+          JSON.parse(potential_json) # Check if this substring is valid JSON
+          return potential_json
+        rescue JSON::ParserError
+          # Fallback: if the strict extraction fails, return the original response
+          # and let the caller deal with a more comprehensive repair or error.
+          # This can happen if there are '{' or '}' in string literals within the JSON.
+          # A more sophisticated parser would be needed for those cases.
+          # For now, this is a common heuristic.
+          return response # Or perhaps an error string/nil
+        end
+      else
+        # No clear JSON structure found, return original response for error handling
+        return response
+      end
+    end
+
+    def extract_code_context_from_diff(diff_output, lines_of_context: 5)
+      # Simplified context extraction focusing on lines around changes.
+      # This is a placeholder for potentially more sophisticated context extraction.
+      # A full AST parse or more complex diff parsing could be used for richer context.
+      context_sections = {}
+      current_file = nil
+      file_lines_buffer = {} # Cache for file lines
+
+      # First pass: identify files and changed line numbers from hunk headers
+      # @@ -old_start,old_lines +new_start,new_lines
+      hunks_by_file = {}
+      diff_output.each_line do |line|
+        line.chomp!
+        if line.start_with?('diff --git')
+          # Extracts 'b_file_path' from "diff --git a/a_file_path b/b_file_path"
+          match = line.match(/diff --git a\/(.+?) b\/(.+)/)
+          current_file = match[2] if match && match[2] != '/dev/null'
+        elsif line.start_with?('+++ b/')
+          # Extracts file path from "+++ b/file_path"
+          current_file = line[6..].strip unless line.include?('/dev/null')
+        elsif line.start_with?('@@') && current_file
+          match = line.match(/@@ -\d+(?:,\d+)? \+(\d+)(?:,(\d+))? @@/)
+          next unless match
+          start_line = match[1].to_i
+          # hunk_lines = match[2]&.to_i || 1 # Number of lines in the hunk in the new file
+          # For context, we care more about the range of lines around the start_line of the change.
+          hunks_by_file[current_file] ||= []
+          hunks_by_file[current_file] << { new_start: start_line }
+        end
+      end
+
+      # Second pass: extract context for each hunk
+      hunks_by_file.each do |file_path, hunks|
+        next unless File.exist?(file_path) # Ensure file exists to read context from
+
+        file_lines_buffer[file_path] ||= File.readlines(file_path, chomp: true)
+        all_lines = file_lines_buffer[file_path]
+
+        hunks.each do |hunk|
+          context_start_line = [1, hunk[:new_start] - lines_of_context].max
+          # Estimate end of change based on typical hunk sizes or simplify
+          # For simplicity, take fixed lines after start, or up to next hunk.
+          # This is a very rough heuristic.
+          context_end_line = [all_lines.length, hunk[:new_start] + lines_of_context + 5].min # +5 for some change content
+
+          actual_start_index = context_start_line - 1
+          actual_end_index = context_end_line - 1
+
+          if actual_start_index < all_lines.length && actual_start_index <= actual_end_index
+            section_content = all_lines[actual_start_index..actual_end_index].join("\n")
+            context_sections[file_path] ||= []
+            # Avoid duplicate sections if hunks are very close
+            unless context_sections[file_path].any? { |s| s[:content] == section_content }
+              context_sections[file_path] << {
+                # Store actual line numbers for reference, not just indices
+                start_line: context_start_line,
+                end_line: context_end_line,
+                content: section_content
+              }
+            end
+          end
+        end
+      end
+      context_sections
+    end
+
+    # Spinner method specifically for diff analysis
+    def analyze_diff_with_spinner(config) # Takes config to initialize LLM
+      llm_service_name = config['llm']
+      llm = case llm_service_name # Initialize LLM based on config
+            when 'openai' then N2B::Llm::OpenAi.new(config)
+            when 'claude' then N2B::Llm::Claude.new(config)
+            when 'gemini' then N2B::Llm::Gemini.new(config)
+            when 'openrouter' then N2B::Llm::OpenRouter.new(config)
+            when 'ollama' then N2B::Llm::Ollama.new(config)
+            else raise N2B::Error, "Unsupported LLM service for analysis: #{llm_service_name}"
+            end
+
+      spinner_chars = ['⠋', '⠙', '⠹', '⠸', '⠼', '⠴', '⠦', '⠧', '⠇', '⠏']
+      spinner_thread = Thread.new do
+        i = 0
+        loop do
+          print "\r#{COLOR_BLUE}🔍 #{spinner_chars[i % spinner_chars.length]} Analyzing diff...#{COLOR_RESET}"
+          $stdout.flush
+          sleep(0.1)
+          i += 1
+        end
+      end
+
+      begin
+        # The block passed to this method contains the actual LLM call
+        result = yield(llm) if block_given?
+        spinner_thread.kill
+        spinner_thread.join # Ensure thread is fully cleaned up
+        print "\r#{' ' * 35}\r" # Clear spinner line
+        puts "#{COLOR_GREEN}✅ Diff analysis complete!#{COLOR_RESET}"
+        result
+      rescue N2B::LlmApiError => e
+        spinner_thread.kill
+        spinner_thread.join
+        print "\r#{' ' * 35}\r"
+        puts "#{COLOR_RED}LLM API Error during diff analysis: #{e.message}#{COLOR_RESET}"
+        # Provide specific model error guidance
+        if e.message.match?(/model|invalid|not found/i)
+          puts "#{COLOR_YELLOW}This might be due to an invalid or unsupported model in your config. Run 'n2b -c' to reconfigure.#{COLOR_RESET}"
+        end
+        # Return a structured error JSON
+        '{"summary": "Error: Could not analyze diff due to LLM API error.", "errors": ["#{e.message}"], "improvements": []}'
+      rescue StandardError => e # Catch other unexpected errors during the yield or LLM call
+        spinner_thread.kill
+        spinner_thread.join
+        print "\r#{' ' * 35}\r"
+        puts "#{COLOR_RED}Unexpected error during diff analysis: #{e.message}#{COLOR_RESET}"
+        '{"summary": "Error: Unexpected failure during diff analysis.", "errors": ["#{e.message}"], "improvements": []}'
+      end
+    end
+    # --- End of moved methods ---
+
+
+    def resolve_block(block, config, full_file_content)
+      comment = nil
+
+      # Display file and line information
+      puts "\n#{COLOR_BLUE}📁 File: #{@file_path}#{COLOR_RESET}"
+      puts "#{COLOR_BLUE}📍 Lines: #{block.start_line}-#{block.end_line} (#{block.base_label} ↔ #{block.incoming_label})#{COLOR_RESET}"
+      puts "#{COLOR_GRAY}💡 You can check this conflict in your editor at the specified line numbers#{COLOR_RESET}\n"
+
+      puts "#{COLOR_YELLOW}🤖 AI is analyzing the conflict...#{COLOR_RESET}"
+      suggestion = request_merge_with_spinner(block, config, comment, full_file_content)
+      puts "#{COLOR_GREEN}✅ Initial suggestion ready!#{COLOR_RESET}\n"
+
+      vcs_type = get_vcs_type_for_file_operations
+
+      # Convert conflict labels to actual VCS revision identifiers
+      base_revision = convert_label_to_revision(block.base_label, vcs_type, :base)
+      incoming_revision = convert_label_to_revision(block.incoming_label, vcs_type, :incoming)
+
+      base_content_full = get_file_content_from_vcs(base_revision, @file_path, vcs_type) || difficulté_to_load_content_placeholder("base content from #{block.base_label}")
+      incoming_content_full = get_file_content_from_vcs(incoming_revision, @file_path, vcs_type) || difficulté_to_load_content_placeholder("incoming content from #{block.incoming_label}")
+
+      generated_html_path = nil
+
+      begin
+        loop do
+          current_resolution_content_full = apply_hunk_to_full_content(full_file_content, block, suggestion['merged_code'])
+
+          # Don't delete the HTML file immediately - keep it available for user preview
+
+          generated_html_path = generate_conflict_preview_html(
+            block,
+            base_content_full,
+            incoming_content_full,
+            current_resolution_content_full,
+            block.base_label,
+            block.incoming_label,
+            @file_path,
+            suggestion
+          )
+
+          preview_link_message = ""
+          if generated_html_path && File.exist?(generated_html_path)
+            preview_link_message = "🌐 #{COLOR_BLUE}Preview: file://#{generated_html_path}#{COLOR_RESET}"
+          else
+            preview_link_message = "#{COLOR_YELLOW}⚠️  Could not generate HTML preview.#{COLOR_RESET}"
+          end
+          puts preview_link_message
+
+          print_conflict(block)
+          print_suggestion(suggestion)
+
+          prompt_message = <<~PROMPT
+            #{COLOR_YELLOW}Actions: [y] Accept, [n] Skip, [c] Comment, [e] Edit, [p] Preview, [s] Refresh, [a] Abort#{COLOR_RESET}
+            #{COLOR_GRAY}(Preview link above can be cmd/ctrl+clicked if your terminal supports it){COLOR_RESET}
+            #{COLOR_YELLOW}Your choice: #{COLOR_RESET}
+          PROMPT
+          print prompt_message
+          choice = $stdin.gets&.strip&.downcase
+
+          case choice
+          when 'y'
+            return {accepted: true, merged_code: suggestion['merged_code'], reason: suggestion['reason'], comment: comment}
+          when 'n'
+            return {accepted: false, merged_code: suggestion['merged_code'], reason: suggestion['reason'], comment: comment}
+          when 'c'
+            puts 'Enter comment (end with blank line):'
+            comment = read_multiline_input
+            puts "#{COLOR_YELLOW}🤖 AI is analyzing your comment and generating new suggestion...#{COLOR_RESET}"
+            current_file_on_disk = File.exist?(@file_path) ? File.read(@file_path) : full_file_content
+            full_file_content = current_file_on_disk
+            suggestion = request_merge_with_spinner(block, config, comment, full_file_content)
+            puts "#{COLOR_GREEN}✅ New suggestion ready!#{COLOR_RESET}\n"
+            # Loop continues, will regenerate preview
+          when 'e'
+            edit_result = handle_editor_workflow(block, config, full_file_content)
+            if edit_result[:resolved]
+              return {accepted: true, merged_code: edit_result[:merged_code], reason: edit_result[:reason], comment: comment}
+            elsif edit_result[:updated_content]
+              full_file_content = edit_result[:updated_content]
+              puts "#{COLOR_YELLOW}🤖 Content changed by editor. Re-analyzing for new suggestion...#{COLOR_RESET}"
+              suggestion = request_merge_with_spinner(block, config, comment, full_file_content)
+            end
+            # Loop continues, will regenerate preview
+          when 'p' # Open Preview in Browser
+            if generated_html_path && File.exist?(generated_html_path)
+              puts "#{COLOR_BLUE}🌐 Opening preview in browser...#{COLOR_RESET}"
+              open_html_in_browser(generated_html_path)
+            else
+              puts "#{COLOR_YELLOW}⚠️  No preview available to open.#{COLOR_RESET}"
+            end
+            # Loop continues, no changes to suggestion
+          when 's' # Refresh Preview
+            puts "#{COLOR_BLUE}🔄 Refreshing suggestion and preview...#{COLOR_RESET}"
+            suggestion = request_merge_with_spinner(block, config, comment, full_file_content)
+            # Loop continues, preview will be regenerated
+          when 'a'
+            return {abort: true, merged_code: suggestion['merged_code'], reason: suggestion['reason'], comment: comment}
+          when '', nil
+            puts "#{COLOR_RED}Please enter a valid choice.#{COLOR_RESET}"
+          else
+            puts "#{COLOR_RED}Invalid option. Please choose from the available actions.#{COLOR_RESET}"
+          end
+        end
+      ensure
+        FileUtils.rm_f(generated_html_path) if generated_html_path && File.exist?(generated_html_path)
+      end
+    end
+
+    def difficulté_to_load_content_placeholder(description)
+      # Helper to return a placeholder if VCS content fails, aiding debug in preview
+      "N2B: Could not load #{description}. Displaying this placeholder."
+    end
+
+    def request_merge(block, config, comment, full_file_content)
+      prompt = build_merge_prompt(block, comment, full_file_content)
+      json_str = call_llm_for_merge(prompt, config)
+
+      begin
+        parsed = JSON.parse(extract_json(json_str))
+
+        # Validate the response structure
+        unless parsed.is_a?(Hash) && parsed.key?('merged_code') && parsed.key?('reason')
+          raise JSON::ParserError, "Response missing required keys 'merged_code' and 'reason'"
+        end
+
+        parsed
+      rescue JSON::ParserError => e
+        # First try automatic JSON repair
+        puts "#{COLOR_YELLOW}⚠️  Invalid JSON detected, attempting automatic repair...#{COLOR_RESET}"
+        repaired_response = attempt_json_repair(json_str, config)
+
+        if repaired_response
+          puts "#{COLOR_GREEN}✅ JSON repair successful!#{COLOR_RESET}"
+          return repaired_response
+        else
+          puts "#{COLOR_RED}❌ JSON repair failed#{COLOR_RESET}"
+          handle_invalid_llm_response(json_str, e, block)
+        end
+      end
+    end
+
+    def request_merge_with_spinner(block, config, comment, full_file_content)
+      spinner_chars = ['⠋', '⠙', '⠹', '⠸', '⠼', '⠴', '⠦', '⠧', '⠇', '⠏']
+      spinner_thread = Thread.new do
+        i = 0
+        while true
+          print "\r#{COLOR_BLUE}#{spinner_chars[i % spinner_chars.length]} Processing...#{COLOR_RESET}"
+          $stdout.flush
+          sleep(0.1)
+          i += 1
+        end
+      end
+
+      begin
+        result = request_merge(block, config, comment, full_file_content)
+        spinner_thread.kill
+        print "\r#{' ' * 20}\r"  # Clear the spinner line
+        result
+      rescue => e
+        spinner_thread.kill
+        print "\r#{' ' * 20}\r"  # Clear the spinner line
+        { 'merged_code' => '', 'reason' => "Error: #{e.message}" }
+      end
+    end
+
+    def build_merge_prompt(block, comment, full_file_content)
+      config = get_config(reconfigure: false, advanced_flow: false)
+      template_path = resolve_template_path('merge_conflict_prompt', config)
+      template = File.read(template_path)
+
+      user_comment_text = comment && !comment.empty? ? "User comment: #{comment}" : ""
+
+      template.gsub('{full_file_content}', full_file_content.to_s)
+              .gsub('{start_line}', block.start_line.to_s)
+              .gsub('{end_line}', block.end_line.to_s)
+              .gsub('{context_before}', block.context_before.to_s)
+              .gsub('{base_label}', block.base_label.to_s)
+              .gsub('{base_content}', block.base_content.to_s)
+              .gsub('{incoming_content}', block.incoming_content.to_s)
+              .gsub('{incoming_label}', block.incoming_label.to_s)
+              .gsub('{context_after}', block.context_after.to_s)
+              .gsub('{user_comment}', user_comment_text)
+    end
+
+    def call_llm_for_merge(prompt, config)
+      llm_service_name = config['llm']
+      llm = case llm_service_name
+            when 'openai'
+              N2B::Llm::OpenAi.new(config)
+            when 'claude'
+              N2B::Llm::Claude.new(config)
+            when 'gemini'
+              N2B::Llm::Gemini.new(config)
+            when 'openrouter'
+              N2B::Llm::OpenRouter.new(config)
+            when 'ollama'
+              N2B::Llm::Ollama.new(config)
+            else
+              raise N2B::Error, "Unsupported LLM service: #{llm_service_name}"
+            end
+      llm.analyze_code_diff(prompt)
+    rescue N2B::LlmApiError => e
+      puts "\n#{COLOR_RED}❌ LLM API Error#{COLOR_RESET}"
+      puts "#{COLOR_YELLOW}Failed to communicate with the AI service.#{COLOR_RESET}"
+      puts "#{COLOR_GRAY}Error: #{e.message}#{COLOR_RESET}"
+
+      # Check for common error types and provide specific guidance
+      if e.message.include?('model') || e.message.include?('Model') || e.message.include?('invalid') || e.message.include?('not found')
+        puts "\n#{COLOR_BLUE}💡 This looks like a model configuration issue.#{COLOR_RESET}"
+        puts "#{COLOR_YELLOW}Run 'n2b -c' to reconfigure your model settings.#{COLOR_RESET}"
+      elsif e.message.include?('auth') || e.message.include?('unauthorized') || e.message.include?('401')
+        puts "\n#{COLOR_BLUE}💡 This looks like an authentication issue.#{COLOR_RESET}"
+        puts "#{COLOR_YELLOW}Check your API key configuration with 'n2b -c'.#{COLOR_RESET}"
+      elsif e.message.include?('timeout') || e.message.include?('network')
+        puts "\n#{COLOR_BLUE}💡 This looks like a network issue.#{COLOR_RESET}"
+        puts "#{COLOR_YELLOW}Check your internet connection and try again.#{COLOR_RESET}"
+      end
+
+      '{"merged_code":"","reason":"LLM API error: ' + e.message.gsub('"', '\\"') + '"}'
+    end
+
+    def extract_json(response)
+      JSON.parse(response)
+      response
+    rescue JSON::ParserError
+      start = response.index('{')
+      stop = response.rindex('}')
+      return response unless start && stop
+      response[start..stop]
+    end
+
+    def handle_invalid_llm_response(raw_response, error, block)
+      puts "\n#{COLOR_RED}❌ Invalid LLM Response Error#{COLOR_RESET}"
+      puts "#{COLOR_YELLOW}The AI returned an invalid response that couldn't be parsed.#{COLOR_RESET}"
+      puts "#{COLOR_YELLOW}Automatic JSON repair was attempted but failed.#{COLOR_RESET}"
+      puts "#{COLOR_GRAY}Error: #{error.message}#{COLOR_RESET}"
+
+      # Save problematic response for debugging
+      save_debug_response(raw_response, error)
+
+      # Show truncated raw response for debugging
+      truncated_response = raw_response.length > 200 ? "#{raw_response[0..200]}..." : raw_response
+      puts "\n#{COLOR_GRAY}Raw response (truncated):#{COLOR_RESET}"
+      puts "#{COLOR_GRAY}#{truncated_response}#{COLOR_RESET}"
+
+      puts "\n#{COLOR_BLUE}What would you like to do?#{COLOR_RESET}"
+      puts "#{COLOR_GREEN}[r]#{COLOR_RESET} Retry with the same prompt"
+      puts "#{COLOR_YELLOW}[c]#{COLOR_RESET} Add a comment to guide the AI better"
+      puts "#{COLOR_BLUE}[m]#{COLOR_RESET} Manually choose one side of the conflict"
+      puts "#{COLOR_RED}[s]#{COLOR_RESET} Skip this conflict"
+      puts "#{COLOR_RED}[a]#{COLOR_RESET} Abort entirely"
+
+      loop do
+        print "#{COLOR_YELLOW}Choose action [r/c/m/s/a]: #{COLOR_RESET}"
+        choice = $stdin.gets&.strip&.downcase
+
+        case choice
+        when 'r'
+          puts "#{COLOR_YELLOW}🔄 Retrying with same prompt...#{COLOR_RESET}"
+          return retry_llm_request(block)
+        when 'c'
+          puts "#{COLOR_YELLOW}💬 Please provide guidance for the AI:#{COLOR_RESET}"
+          comment = read_multiline_input
+          return retry_llm_request_with_comment(block, comment)
+        when 'm'
+          return handle_manual_choice(block)
+        when 's'
+          puts "#{COLOR_YELLOW}⏭️  Skipping this conflict#{COLOR_RESET}"
+          return { 'merged_code' => '', 'reason' => 'Skipped due to invalid LLM response' }
+        when 'a'
+          puts "#{COLOR_RED}🛑 Aborting merge resolution#{COLOR_RESET}"
+          return { 'merged_code' => '', 'reason' => 'Aborted due to invalid LLM response', 'abort' => true }
+        when '', nil
+          puts "#{COLOR_RED}Please enter a valid choice: r/c/m/s/a#{COLOR_RESET}"
+        else
+          puts "#{COLOR_RED}Invalid option. Please enter: r (retry), c (comment), m (manual), s (skip), or a (abort)#{COLOR_RESET}"
+        end
+      end
+    end
+
+    def retry_llm_request(block)
+      config = get_config(reconfigure: false, advanced_flow: false)
+      # Always re-read file content in case it was edited
+      full_file_content = File.read(@file_path)
+
+      puts "#{COLOR_YELLOW}🤖 Retrying AI analysis...#{COLOR_RESET}"
+      suggestion = request_merge_with_spinner(block, config, nil, full_file_content)
+      puts "#{COLOR_GREEN}✅ Retry successful!#{COLOR_RESET}"
+      suggestion
+    rescue => e
+      puts "#{COLOR_RED}❌ Retry failed: #{e.message}#{COLOR_RESET}"
+      { 'merged_code' => '', 'reason' => 'Retry failed due to persistent LLM error' }
+    end
+
+    def retry_llm_request_with_comment(block, comment)
+      config = get_config(reconfigure: false, advanced_flow: false)
+      # Always re-read file content in case it was edited
+      full_file_content = File.read(@file_path)
+
+      puts "#{COLOR_YELLOW}🤖 Retrying with your guidance...#{COLOR_RESET}"
+      suggestion = request_merge_with_spinner(block, config, comment, full_file_content)
+      puts "#{COLOR_GREEN}✅ Retry with comment successful!#{COLOR_RESET}"
+      suggestion
+    rescue => e
+      puts "#{COLOR_RED}❌ Retry with comment failed: #{e.message}#{COLOR_RESET}"
+      { 'merged_code' => '', 'reason' => 'Retry with comment failed due to persistent LLM error' }
+    end
+
+    def handle_manual_choice(block)
+      puts "\n#{COLOR_BLUE}Manual conflict resolution:#{COLOR_RESET}"
+      puts "#{COLOR_RED}[1] Choose HEAD version (#{block.base_label})#{COLOR_RESET}"
+      puts "#{COLOR_RED}#{block.base_content}#{COLOR_RESET}"
+      puts ""
+      puts "#{COLOR_GREEN}[2] Choose incoming version (#{block.incoming_label})#{COLOR_RESET}"
+      puts "#{COLOR_GREEN}#{block.incoming_content}#{COLOR_RESET}"
+      puts ""
+      puts "#{COLOR_YELLOW}[3] Skip this conflict#{COLOR_RESET}"
+
+      loop do
+        print "#{COLOR_YELLOW}Choose version [1/2/3]: #{COLOR_RESET}"
+        choice = $stdin.gets&.strip
+
+        case choice
+        when '1'
+          puts "#{COLOR_GREEN}✅ Selected HEAD version#{COLOR_RESET}"
+          return {
+            'merged_code' => block.base_content,
+            'reason' => "Manually selected HEAD version (#{block.base_label}) due to LLM error"
+          }
+        when '2'
+          puts "#{COLOR_GREEN}✅ Selected incoming version#{COLOR_RESET}"
+          return {
+            'merged_code' => block.incoming_content,
+            'reason' => "Manually selected incoming version (#{block.incoming_label}) due to LLM error"
+          }
+        when '3'
+          puts "#{COLOR_YELLOW}⏭️  Skipping conflict#{COLOR_RESET}"
+          return { 'merged_code' => '', 'reason' => 'Manually skipped due to LLM error' }
+        when '', nil
+          puts "#{COLOR_RED}Please enter 1, 2, or 3#{COLOR_RESET}"
+        else
+          puts "#{COLOR_RED}Invalid choice. Please enter 1 (HEAD), 2 (incoming), or 3 (skip)#{COLOR_RESET}"
+        end
+      end
+    end
+
+    def read_multiline_input
+      lines = []
+      puts "#{COLOR_GRAY}(Type your comment, then press Enter on an empty line to finish)#{COLOR_RESET}"
+      while (line = $stdin.gets)
+        line = line.chomp
+        break if line.empty?
+        lines << line
+      end
+      comment = lines.join("\n")
+      if comment.empty?
+        puts "#{COLOR_YELLOW}No comment entered.#{COLOR_RESET}"
+      else
+        puts "#{COLOR_GREEN}Comment received: #{comment.length} characters#{COLOR_RESET}"
+      end
+      comment
+    end
+
+    def print_conflict(block)
+      # Show context before conflict for better understanding
+      if block.context_before && !block.context_before.empty?
+        puts "#{COLOR_GRAY}... context before ...#{COLOR_RESET}"
+        context_lines = block.context_before.split("\n").last(3) # Show last 3 lines of context
+        context_lines.each { |line| puts "#{COLOR_GRAY}#{line}#{COLOR_RESET}" }
+      end
+
+      puts "#{COLOR_RED}<<<<<<< #{block.base_label} (lines #{block.start_line}-#{block.end_line})#{COLOR_RESET}"
+      puts "#{COLOR_RED}#{block.base_content}#{COLOR_RESET}"
+      puts "#{COLOR_YELLOW}=======#{COLOR_RESET}"
+      puts "#{COLOR_GREEN}#{block.incoming_content}#{COLOR_RESET}"
+      puts "#{COLOR_YELLOW}>>>>>>> #{block.incoming_label}#{COLOR_RESET}"
+
+      # Show context after conflict for better understanding
+      if block.context_after && !block.context_after.empty?
+        context_lines = block.context_after.split("\n").first(3) # Show first 3 lines of context
+        context_lines.each { |line| puts "#{COLOR_GRAY}#{line}#{COLOR_RESET}" }
+        puts "#{COLOR_GRAY}... context after ...#{COLOR_RESET}"
+      end
+    end
+
+    def print_suggestion(sug)
+      puts "#{COLOR_BLUE}--- Suggestion ---#{COLOR_RESET}"
+      puts "#{COLOR_BLUE}#{sug['merged_code']}#{COLOR_RESET}"
+      puts "#{COLOR_GRAY}Reason: #{sug['reason']}#{COLOR_RESET}"
+    end
+
+    def resolve_template_path(template_key, config)
+      user_path = config.dig('templates', template_key) if config.is_a?(Hash)
+      return user_path if user_path && File.exist?(user_path)
+
+      File.expand_path(File.join(__dir__, 'templates', "#{template_key}.txt"))
+    end
+
+    def mark_file_as_resolved(file_path)
+      # Detect VCS and mark file as resolved
+      if File.exist?('.hg')
+        result = execute_vcs_command_with_timeout("hg resolve --mark #{Shellwords.escape(file_path)}", 10)
+        if result[:success]
+          puts "#{COLOR_GREEN}✅ Marked #{file_path} as resolved in Mercurial#{COLOR_RESET}"
+        else
+          puts "#{COLOR_YELLOW}⚠️  Could not mark #{file_path} as resolved in Mercurial#{COLOR_RESET}"
+          puts "#{COLOR_GRAY}Error: #{result[:error]}#{COLOR_RESET}" if result[:error]
+          puts "#{COLOR_GRAY}💡 You can manually mark it with: hg resolve --mark #{file_path}#{COLOR_RESET}"
+        end
+      elsif File.exist?('.git')
+        result = execute_vcs_command_with_timeout("git add #{Shellwords.escape(file_path)}", 10)
+        if result[:success]
+          puts "#{COLOR_GREEN}✅ Added #{file_path} to Git staging area#{COLOR_RESET}"
+        else
+          puts "#{COLOR_YELLOW}⚠️  Could not add #{file_path} to Git staging area#{COLOR_RESET}"
+          puts "#{COLOR_GRAY}Error: #{result[:error]}#{COLOR_RESET}" if result[:error]
+          puts "#{COLOR_GRAY}💡 You can manually add it with: git add #{file_path}#{COLOR_RESET}"
+        end
+      else
+        puts "#{COLOR_BLUE}ℹ️  No VCS detected - file saved but not marked as resolved#{COLOR_RESET}"
+      end
+    end
+
+    def execute_vcs_command_with_timeout(command, timeout_seconds)
+      require 'open3'
+
+      begin
+        # Use Open3.popen3 with manual timeout handling to avoid thread issues
+        stdin, stdout, stderr, wait_thr = Open3.popen3(command)
+        stdin.close
+
+        # Wait for the process with timeout
+        if wait_thr.join(timeout_seconds)
+          # Process completed within timeout
+          stdout_content = stdout.read
+          stderr_content = stderr.read
+          status = wait_thr.value
+
+          stdout.close
+          stderr.close
+
+          if status.success?
+            { success: true, stdout: stdout_content, stderr: stderr_content }
+          else
+            { success: false, error: "Command failed: #{stderr_content.strip}", stdout: stdout_content, stderr: stderr_content }
+          end
+        else
+          # Process timed out, kill it
+          Process.kill('TERM', wait_thr.pid) rescue nil
+          sleep(0.1)
+          Process.kill('KILL', wait_thr.pid) rescue nil
+
+          stdout.close rescue nil
+          stderr.close rescue nil
+          wait_thr.join rescue nil
+
+          { success: false, error: "Command timed out after #{timeout_seconds} seconds" }
+        end
+      rescue => e
+        { success: false, error: "Unexpected error: #{e.message}" }
+      end
+    end
+
+    def show_help_and_status
+      # This method will be invoked by OptionParser for -h/--help,
+      # or when n2b-diff is run without arguments in non-analyze mode.
+      # The OptionParser instance itself will print most of the help text.
+      # We just add any extra status info here.
+
+      puts "" # Extra newline for spacing after OptionParser's output if it called this.
+      # If running not due to -h, but due to missing file_path in merge mode:
+      if !@options[:analyze] && @file_path.nil? && !@args.include?('-h') && !@args.include?('--help')
+         puts "#{COLOR_RED}Error: No file path provided for merge conflict resolution."
+         puts "Run with -h or --help for detailed usage."
+         puts ""
+      end
+
+
+      # Show unresolved conflicts if in a VCS repository and not in analyze mode
+      # (or if specifically requested, but for now, tied to non-analyze mode)
+      if !@options[:analyze] && (File.exist?('.hg') || File.exist?('.git'))
+        puts "#{COLOR_BLUE}📋 Unresolved Conflicts Status:#{COLOR_RESET}"
+        if File.exist?('.hg')
+          puts "#{COLOR_GRAY}Checking Mercurial...#{COLOR_RESET}"
+          result = execute_vcs_command_with_timeout("hg resolve --list", 5)
+          if result[:success]
+            unresolved_files = result[:stdout].lines.select { |line| line.start_with?('U ') }
+            if unresolved_files.any?
+              unresolved_files.each { |line| puts "  #{COLOR_RED}❌ #{line.strip.sub(/^U /, '')} (Mercurial)#{COLOR_RESET}" }
+              puts "\n#{COLOR_YELLOW}💡 Use: n2b-diff <filename> to resolve listed conflicts.#{COLOR_RESET}"
+            else
+              puts "  #{COLOR_GREEN}✅ No unresolved Mercurial conflicts.#{COLOR_RESET}"
+            end
+          else
+            puts "  #{COLOR_YELLOW}⚠️  Could not check Mercurial status: #{result[:error]}#{COLOR_RESET}"
+          end
+        end
+
+        if File.exist?('.git')
+          puts "#{COLOR_GRAY}Checking Git...#{COLOR_RESET}"
+          # For Git, `git status --porcelain` is better as `git diff --name-only --diff-filter=U` only shows unmerged paths.
+          # We want to show files with conflict markers.
+          # `git status --porcelain=v1` shows "UU" for unmerged files.
+          result = execute_vcs_command_with_timeout("git status --porcelain=v1", 5)
+          if result[:success]
+            unresolved_files = result[:stdout].lines.select{|line| line.start_with?('UU ')}.map{|line| line.sub(/^UU /, '').strip}
+            if unresolved_files.any?
+              unresolved_files.each { |file| puts "  #{COLOR_RED}❌ #{file} (Git)#{COLOR_RESET}" }
+              puts "\n#{COLOR_YELLOW}💡 Use: n2b-diff <filename> to resolve listed conflicts.#{COLOR_RESET}"
+            else
+              puts "  #{COLOR_GREEN}✅ No unresolved Git conflicts.#{COLOR_RESET}"
+            end
+          else
+            puts "  #{COLOR_YELLOW}⚠️  Could not check Git status: #{result[:error]}#{COLOR_RESET}"
+          end
+        end
+      elsif @options[:analyze]
+        # This part might not be reached if -h is used, as OptionParser exits.
+        # But if called for other reasons in analyze mode:
+        puts "#{COLOR_BLUE}ℹ️  Running in analysis mode. VCS conflict status check is skipped.#{COLOR_RESET}"
+      end
+      # Ensure exit if we got here due to an operational error (like no file in merge mode)
+      # but not if it was just -h (OptionParser handles exit for -h)
+      if !@options[:analyze] && @file_path.nil? && !@args.include?('-h') && !@args.include?('--help')
+        exit 1
+      end
+    end
+
+    def attempt_json_repair(malformed_response, config)
+      repair_prompt = build_json_repair_prompt(malformed_response)
+
+      begin
+        puts "#{COLOR_BLUE}🔧 Asking AI to fix the JSON...#{COLOR_RESET}"
+        repaired_json_str = call_llm_for_merge(repair_prompt, config)
+
+        # Try to parse the repaired response
+        parsed = JSON.parse(extract_json(repaired_json_str))
+
+        # Validate the repaired response structure
+        if parsed.is_a?(Hash) && parsed.key?('merged_code') && parsed.key?('reason')
+          return parsed
+        else
+          puts "#{COLOR_YELLOW}⚠️  Repaired JSON missing required keys#{COLOR_RESET}"
+          return nil
+        end
+      rescue JSON::ParserError => e
+        puts "#{COLOR_YELLOW}⚠️  JSON repair attempt also returned invalid JSON#{COLOR_RESET}"
+        return nil
+      rescue => e
+        puts "#{COLOR_YELLOW}⚠️  JSON repair failed: #{e.message}#{COLOR_RESET}"
+        return nil
+      end
+    end
+
+    def build_json_repair_prompt(malformed_response)
+      <<~PROMPT
+        The following response was supposed to be valid JSON with keys "merged_code" and "reason", but it has formatting issues. Please fix it and return ONLY the corrected JSON:
+
+        Original response:
+        #{malformed_response}
+
+        Requirements:
+        - Must be valid JSON
+        - Must have "merged_code" key with the code content
+        - Must have "reason" key with explanation
+        - Return ONLY the JSON, no other text
+
+        Fixed JSON:
+      PROMPT
+    end
+
+    def handle_editor_workflow(block, config, full_file_content)
+      editor_command = config.dig('editor', 'command')
+      editor_type = config.dig('editor', 'type')
+      editor_configured = config.dig('editor', 'configured')
+
+      lines = File.readlines(@file_path, chomp: true)
+      current_block_content_with_markers = lines[(block.start_line - 1)...block.end_line].join("\n")
+
+      if editor_configured && editor_command && editor_type == 'diff_tool'
+        require 'tmpdir'
+        Dir.mktmpdir("n2b_diff_") do |tmpdir|
+          base_file_path = File.join(tmpdir, "base_#{File.basename(@file_path)}")
+          remote_file_path = File.join(tmpdir, "remote_#{File.basename(@file_path)}")
+          merged_file_path = File.join(tmpdir, "merged_#{File.basename(@file_path)}")
+
+          File.write(base_file_path, block.base_content)
+          File.write(remote_file_path, block.incoming_content)
+
+          # Initial content for the merged file: LLM suggestion or current block with markers
+          initial_merged_content = block.suggestion&.dig('merged_code')
+          if initial_merged_content.nil? || initial_merged_content.strip.empty?
+            initial_merged_content = current_block_content_with_markers
+          end
+          File.write(merged_file_path, initial_merged_content)
+
+          # Common pattern: tool base merged remote. Some tools might vary.
+          # Example: meld uses local remote base --output output_file
+          # For now, using a common sequence. Needs documentation for custom tools.
+          # We assume the tool edits the `merged_file_path` (second argument) in place or uses it as output.
+          full_diff_command = "#{editor_command} #{Shellwords.escape(base_file_path)} #{Shellwords.escape(merged_file_path)} #{Shellwords.escape(remote_file_path)}"
+          puts "#{COLOR_BLUE}🔧 Launching diff tool: #{editor_command}...#{COLOR_RESET}"
+          puts "#{COLOR_GRAY}   Base:   #{base_file_path}#{COLOR_RESET}"
+          puts "#{COLOR_GRAY}   Remote: #{remote_file_path}#{COLOR_RESET}"
+          puts "#{COLOR_GRAY}   Merged: #{merged_file_path} (edit this one)#{COLOR_RESET}"
+
+          system(full_diff_command)
+
+          puts "#{COLOR_BLUE}📁 Diff tool closed.#{COLOR_RESET}"
+          merged_code_from_editor = File.read(merged_file_path)
+
+          # Check if the merged content is different from the initial content with markers
+          # to avoid considering unchanged initial conflict markers as a resolution.
+          if merged_code_from_editor.strip == current_block_content_with_markers.strip && merged_code_from_editor.include?('<<<<<<<')
+             puts "#{COLOR_YELLOW}⚠️  It seems the conflict markers are still present. Did you resolve the conflict?#{COLOR_RESET}"
+          end
+
+          print "#{COLOR_YELLOW}Did you resolve this conflict using the diff tool? [y/n]: #{COLOR_RESET}"
+          response = $stdin.gets&.strip&.downcase
+          if response == 'y'
+            puts "#{COLOR_GREEN}✅ Conflict marked as resolved by user via diff tool#{COLOR_RESET}"
+            return {
+              resolved: true,
+              merged_code: merged_code_from_editor,
+              reason: "User resolved conflict with diff tool: #{editor_command}"
+            }
+          else
+            puts "#{COLOR_BLUE}🔄 Conflict not marked as resolved. Continuing with AI assistance...#{COLOR_RESET}"
+            # If user says 'n', we don't use the content from the diff tool as a resolution.
+            # We might need to re-fetch LLM suggestion or just go back to menu.
+            # For now, return resolved: false. The updated_content is not from the main file.
+            return { resolved: false, updated_content: full_file_content } # original full_file_content
+          end
+        end # Tempdir is automatically removed
+      else
+        # Fallback to text editor or if editor is 'text_editor'
+        editor_to_use = editor_command || detect_system_editor # Use configured or system editor
+
+        original_file_content_for_block_check = File.read(@file_path) # Before text editor opens it
+
+        puts "#{COLOR_BLUE}🔧 Opening #{@file_path} in editor (#{editor_to_use})...#{COLOR_RESET}"
+        open_file_in_editor(@file_path, editor_to_use) # Pass specific editor
+        puts "#{COLOR_BLUE}📁 Editor closed. Checking for changes...#{COLOR_RESET}"
+
+        current_file_content_after_edit = File.read(@file_path)
+
+        if file_changed?(original_file_content_for_block_check, current_file_content_after_edit)
+          puts "#{COLOR_YELLOW}📝 File has been modified.#{COLOR_RESET}"
+          print "#{COLOR_YELLOW}Did you resolve this conflict yourself in the editor? [y/n]: #{COLOR_RESET}"
+          response = $stdin.gets&.strip&.downcase
+
+          if response == 'y'
+            puts "#{COLOR_GREEN}✅ Conflict marked as resolved by user in text editor#{COLOR_RESET}"
+            # Extract the changed block content
+            # Re-read lines as they might have changed in number
+            edited_lines = File.readlines(@file_path, chomp: true)
+            # Heuristic: if lines were added/removed, block boundaries might shift.
+            # For simplicity, we'll use original block's line numbers to extract,
+            # but this might be inaccurate if user adds/removes many lines *outside* the conflict block.
+            # A more robust way would be to re-parse or use markers if they exist.
+            # For now, assume user edits primarily *within* the original start/end lines.
+            # The number of lines in the resolved code could be different.
+            # We need to ask the user to ensure the markers are gone.
+
+            # Let's get the content of the lines that corresponded to the original block.
+            # This isn't perfect if the user adds/deletes lines *within* the block,
+            # changing its length. The LLM's suggestion is for a block of a certain size.
+            # For user resolution, they define the new block.
+            # We need to get the content from start_line to (potentially new) end_line.
+            # This is tricky. The simplest is to take the whole file, but that's not what merge tools do.
+            # The contract is that the user removed the conflict markers.
+
+            # We will return the content of the file from the original start line
+            # to an end line that reflects the number of lines in the manually merged code.
+            # This is still tricky. Let's assume the user edited the block and the surrounding lines are stable.
+            # The `resolve_block` method replaces `lines[(block.start_line-1)...block.end_line]`
+            # So, the returned `merged_code` should be what replaces that segment.
+
+            # Simplest approach: user confirms resolution, we assume the relevant part of the file is the resolution.
+            # We need to extract the content of the resolved block from current_file_content_after_edit
+            # based on block.start_line and the *new* end_line of the resolved conflict.
+            # This is hard without re-parsing.
+            # A practical approach: The user resolved it. The file is now correct *at those lines*.
+            # The `resolve_block` method will write the *entire* `lines` array back to the file.
+            # If the user resolved it, the `lines` array (after their edit) IS the resolution for that part.
+            # So, we need to give `resolve_block` the lines from the file that correspond to the original block markers.
+            # This means the `merged_code` should be the content of the file from `block.start_line`
+            # up to where the `block.end_line` *would* be after their edits.
+
+            # Let's refine: the user has edited the file. The section of the file
+            # that previously contained the conflict markers (block.start_line to block.end_line)
+            # now contains their resolution. We need to extract this segment.
+            # The number of lines might have changed.
+            # The `resolve_block` function will replace `lines[original_start_idx..original_end_idx]` with the new content.
+            # So we must provide the exact lines that should go into that slice.
+
+            # We need to ask the user to confirm the new end line if it changed, or trust they know.
+            # The simplest is to return the segment from the file from original start_line to original end_line,
+            # assuming the user's changes fit there. This is too naive.
+
+            # If the user says 'y', the file is considered resolved in that region.
+            # The `resolve_block` will then write the `lines` array (which is `current_file_content_after_edit.split("\n")`)
+            # back to the file. The key is that `resolve_block` *already has* the full `lines` from the modified file
+            # when it reconstructs the file if result[:accepted] is true.
+            # So, the `merged_code` we return here is more for logging/consistency.
+            # The critical part is that `lines` in `resolve_block` needs to be updated if the file was changed by the editor.
+
+            # The `resolve_block` method reads `lines = File.readlines(@file_path, chomp: true)` at the beginning.
+            # If we edit the file here, `lines` in `resolve_block` becomes stale.
+            # This means `handle_editor_workflow` must return the *new* full file content if it changed.
+            # And the `merged_code` for the log should be the segment of the new file.
+
+            # Let's re-read the file and extract the relevant segment for the log.
+            # The actual application of changes happens because `resolve_block` will use the modified `lines` array.
+            # We need to estimate the new end_line. This is complex.
+            # For now, let's just say "user resolved". The actual diff applied will be based on the whole file change.
+            # The `merged_code` for logging can be a placeholder or the new content of the block.
+            # Let's assume the user ensures the markers are gone.
+            # The content of lines from block.start_line to block.end_line in the *new file* is their resolution.
+            # The number of lines in this resolution can be different from the original block.
+            # This is fine, as the `replacement` in `resolve_block` handles this.
+
+            # Simplification: if user says 'y', the code that will be used is the content
+            # of the file from block.start_line to some new end_line.
+            # The crucial part is that `resolve_block` needs to operate on the *modified* file content.
+            # So, we should pass `current_file_content_after_edit` back up.
+            # And for logging, extract the lines from `block.start_line` to `block.end_line` from this new content.
+            # This assumes the user's resolution fits within the original line numbers, which is not always true.
+
+            # The most robust is to re-parse the file for conflict markers. If none are found in this region, it's resolved.
+            # The "merged_code" would be the lines from the edited file that replaced the original conflict.
+
+            # Let's assume the user has resolved the conflict markers from `block.start_line` to `block.end_line`.
+            # The content of these lines in `current_file_content_after_edit` is the resolution.
+            # The number of lines of this resolution might be different.
+            # The `resolve_block` needs to replace `lines[(block.start_line-1)...block.end_line]`
+            # The `merged_code` should be this new segment.
+
+            # For now, let `merged_code` be a conceptual value. The `resolve_block` loop needs to use the new file content.
+            # The key is `updated_content` for the main loop, and `merged_code` for logging.
+
+            # The `resolve_block` method needs to use the content of the file *after* the edit.
+            # The current structure of `resolve_block` re-reads `lines` only if `request_merge_with_spinner` is called again.
+            # This needs adjustment.
+
+            # For now, if user says 'y':
+            # 1. The `merged_code` will be what's in the file from `block.start_line` to `block.end_line` (original numbering).
+            #    This is imperfect for logging if lines were added/removed.
+            # 2. The `resolve_block` loop must use `current_file_content_after_edit` for its `lines` variable.
+            #    This is the most important part for correctness.
+
+            # Let's return the segment from the modified file for `merged_code`.
+            # This is still tricky because `block.end_line` is from the original parse.
+            # If user deleted lines, `block.end_line` might be out of bounds for `edited_lines`.
+            # If user added lines, we wouldn't capture all of it.
+
+            # Simplest for now: the user resolved it. The merged_code for logging can be a placeholder.
+            # The main thing is that `resolve_block` now operates on `current_file_content_after_edit`.
+            # The subtask asks for `merged_code` to be `<content_from_editor_or_file>`.
+            # This means the content of the resolved block.
+
+            # Let's try to extract the content from the edited file using original line numbers as a guide.
+            # This is a known limitation. A better way would be for user to indicate new block end.
+            resolved_segment = edited_lines[(block.start_line - 1)..[block.end_line - 1]].join("\n") rescue "User resolved - content not easily extracted due to line changes"
+            if edited_lines.slice((block.start_line-1)...(block.end_line)).join("\n").include?("<<<<<<<")
+                puts "#{COLOR_YELLOW}⚠️  Conflict markers seem to still be present in the edited file. Please ensure they are removed for proper resolution.#{COLOR_RESET}"
+            end
+
+
+            return {
+              resolved: true,
+              merged_code: resolved_segment, # Content from the file for the resolved block
+              reason: "User resolved conflict in text editor: #{editor_to_use}",
+              updated_content: current_file_content_after_edit # Pass back the full content
+            }
+          else
+            puts "#{COLOR_BLUE}🔄 Conflict not marked as resolved. Continuing with AI assistance...#{COLOR_RESET}"
+            return {
+              resolved: false,
+              updated_content: current_file_content_after_edit # Pass back the full content
+            }
+          end
+        else
+          puts "#{COLOR_GRAY}📋 No changes detected in the file. Continuing...#{COLOR_RESET}"
+          return { resolved: false, updated_content: nil } # No changes, so original full_file_content is still valid
+        end
+      end
+    end
+
+    def detect_system_editor
+      # This is the ultimate fallback if no configuration is set.
+      # The ENV['EDITOR'] || ENV['VISUAL'] check should be done by the caller if preferred before this.
+      case RbConfig::CONFIG['host_os']
+      when /darwin|mac os/
+        'open' # Typically non-blocking, might need different handling or user awareness.
+      when /linux/
+        # Prefer common user-friendly editors if available, then vi as fallback.
+        # This simple version just picks one. `command_exists?` could be used here.
+        ENV['EDITOR'] || ENV['VISUAL'] || 'nano' # or 'vi'
+      when /mswin|mingw/
+        ENV['EDITOR'] || ENV['VISUAL'] || 'notepad'
+      else
+        ENV['EDITOR'] || ENV['VISUAL'] || 'vi' # vi is a common default on Unix-like systems
+      end
+    end
+
+    def open_file_in_editor(file_path, editor_command = nil)
+      # If no specific editor_command is passed, try configured editor, then system fallbacks.
+      # This method is now simplified as the decision of *which* editor (configured vs fallback)
+      # is made in handle_editor_workflow. This method just executes it.
+      # However, the original call from resolve_block (before this change) did not pass editor_command.
+      # So, if editor_command is nil, we should still try to get it from config or fallback.
+
+      effective_editor = editor_command # Use passed command if available
+
+      if effective_editor.nil?
+        config = get_config(reconfigure: false, advanced_flow: false) # Ensure config is loaded
+        effective_editor = config.dig('editor', 'command') if config.dig('editor', 'configured')
+        effective_editor ||= detect_system_editor # Ultimate fallback
+      end
+
+      puts "#{COLOR_GRAY}Attempting to open with: #{effective_editor} #{Shellwords.escape(file_path)}#{COLOR_RESET}"
+      begin
+        if RbConfig::CONFIG['host_os'] =~ /darwin|mac os/ && effective_editor == 'open'
+          # 'open' is non-blocking. This is fine.
+          result = system("open #{Shellwords.escape(file_path)}")
+          unless result
+            # `system` returns true if command found and exited with 0, false otherwise for `open`.
+            # It returns nil if command execution fails.
+            puts "#{COLOR_YELLOW}⚠️  'open' command might have failed or file opened in background. Please check manually.#{COLOR_RESET}"
+          end
+          # For 'open', we don't wait. User needs to manually come back.
+          # Consider adding a prompt "Press Enter after closing the editor..." for non-blocking editors.
+          # For now, keeping it simple.
+        else
+          # For most terminal editors, system() will block until the editor is closed.
+          system("#{effective_editor} #{Shellwords.escape(file_path)}")
+        end
+      rescue StandardError => e
+        puts "#{COLOR_RED}❌ Failed to open editor '#{effective_editor}': #{e.message}#{COLOR_RESET}"
+        puts "#{COLOR_YELLOW}💡 Please ensure your configured editor is correct or set your EDITOR environment variable.#{COLOR_RESET}"
+        puts "#{COLOR_BLUE}You may need to open #{file_path} manually in your preferred editor to make changes.#{COLOR_RESET}"
+        print "#{COLOR_YELLOW}Press Enter to continue after manually editing (if you choose to do so)...#{COLOR_RESET}"
+        $stdin.gets # Pause to allow manual editing
+      end
+    end
+
+    def file_changed?(original_content, current_content)
+      original_content != current_content
+    end
+
+    def save_debug_response(raw_response, error)
+      begin
+        debug_dir = '.n2b_debug'
+        FileUtils.mkdir_p(debug_dir)
+        timestamp = Time.now.strftime('%Y-%m-%d-%H%M%S')
+        debug_file = File.join(debug_dir, "invalid_response_#{timestamp}.txt")
+
+        debug_content = <<~DEBUG
+          N2B Debug: Invalid LLM Response
+          ===============================
+          Timestamp: #{Time.now}
+          File: #{@file_path}
+          Error: #{error.message}
+          Error Class: #{error.class}
+
+          Raw LLM Response:
+          -----------------
+          #{raw_response}
+
+          End of Response
+          ===============
+        DEBUG
+
+        File.write(debug_file, debug_content)
+        puts "#{COLOR_GRAY}🐛 Debug info saved to #{debug_file}#{COLOR_RESET}"
+      rescue => e
+        puts "#{COLOR_GRAY}⚠️  Could not save debug info: #{e.message}#{COLOR_RESET}"
+      end
+    end
+
+    def generate_merge_log_html(log_entries, timestamp)
+      git_info = extract_git_info
+
+      html = <<~HTML
+        <!DOCTYPE html>
+        <html lang="en">
+        <head>
+          <meta charset="UTF-8">
+          <meta name="viewport" content="width=device-width, initial-scale=1.0">
+          <title>N2B Merge Log - #{@file_path} - #{timestamp}</title>
+          <style>
+            body {
+              font-family: 'SF Mono', 'Monaco', 'Inconsolata', 'Fira Code', monospace;
+              margin: 0;
+              padding: 20px;
+              background-color: #f8f9fa;
+              color: #333;
+              line-height: 1.6;
+            }
+            .header {
+              background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+              color: white;
+              padding: 20px;
+              border-radius: 8px;
+              margin-bottom: 20px;
+              box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
+            }
+            .header h1 {
+              margin: 0 0 10px 0;
+              font-size: 24px;
+            }
+            .header .meta {
+              opacity: 0.9;
+              font-size: 14px;
+            }
+            .conflict-container {
+              background: white;
+              border-radius: 8px;
+              margin-bottom: 20px;
+              box-shadow: 0 2px 4px rgba(0, 0, 0, 0.1);
+              overflow: hidden;
+            }
+            .conflict-header {
+              background-color: #e9ecef;
+              padding: 15px;
+              border-bottom: 1px solid #dee2e6;
+              font-weight: bold;
+              color: #495057;
+            }
+            .conflict-table {
+              width: 100%;
+              border-collapse: collapse;
+            }
+            .conflict-table th {
+              background-color: #f8f9fa;
+              padding: 12px;
+              text-align: left;
+              font-weight: 600;
+              border-bottom: 2px solid #dee2e6;
+              color: #495057;
+            }
+            .conflict-table td {
+              padding: 12px;
+              border-bottom: 1px solid #e9ecef;
+              vertical-align: top;
+            }
+            .code-block {
+              background-color: #f8f9fa;
+              border: 1px solid #e9ecef;
+              border-radius: 4px;
+              padding: 10px;
+              font-family: 'SF Mono', 'Monaco', 'Inconsolata', 'Fira Code', monospace;
+              font-size: 13px;
+              white-space: pre-wrap;
+              overflow-x: auto;
+              max-height: 300px;
+              overflow-y: auto;
+            }
+            .base-code { border-left: 4px solid #dc3545; }
+            .incoming-code { border-left: 4px solid #007bff; }
+            .resolution-code { border-left: 4px solid #28a745; }
+            .method-badge {
+              display: inline-block;
+              padding: 4px 8px;
+              border-radius: 12px;
+              font-size: 12px;
+              font-weight: 500;
+              text-transform: uppercase;
+            }
+            .method-llm { background-color: #e3f2fd; color: #1976d2; }
+            .method-manual { background-color: #fff3e0; color: #f57c00; }
+            .method-skip { background-color: #fce4ec; color: #c2185b; }
+            .method-abort { background-color: #ffebee; color: #d32f2f; }
+            .footer {
+              text-align: center;
+              margin-top: 30px;
+              padding: 20px;
+              color: #6c757d;
+              font-size: 14px;
+            }
+            .stats {
+              display: flex;
+              gap: 20px;
+              margin-top: 10px;
+            }
+            .stat-item {
+              background: rgba(255, 255, 255, 0.2);
+              padding: 8px 12px;
+              border-radius: 4px;
+            }
+          </style>
+        </head>
+        <body>
+          <div class="header">
+            <h1>🔀 N2B Merge Resolution Log</h1>
+            <div class="meta">
+              <strong>File:</strong> #{@file_path}<br>
+              <strong>Timestamp:</strong> #{Time.now.strftime('%Y-%m-%d %H:%M:%S UTC')}<br>
+              <strong>Branch:</strong> #{git_info[:branch]}<br>
+              <strong>Total Conflicts:</strong> #{log_entries.length}
+            </div>
+            <div class="stats">
+              <div class="stat-item">
+                <strong>Resolved:</strong> #{log_entries.count { |e| e[:action] == 'accepted' }}
+              </div>
+              <div class="stat-item">
+                <strong>Skipped:</strong> #{log_entries.count { |e| e[:action] == 'skipped' }}
+              </div>
+              <div class="stat-item">
+                <strong>Aborted:</strong> #{log_entries.count { |e| e[:action] == 'aborted' }}
+              </div>
+            </div>
+          </div>
+      HTML
+
+      log_entries.each_with_index do |entry, index|
+        conflict_number = index + 1
+        method_class = case entry[:resolution_method]
+                      when /llm|ai|suggested/i then 'method-llm'
+                      when /manual|user|edit/i then 'method-manual'
+                      when /skip/i then 'method-skip'
+                      when /abort/i then 'method-abort'
+                      else 'method-llm'
+                      end
+
+        html += <<~CONFLICT_HTML
+          <div class="conflict-container">
+            <div class="conflict-header">
+              Conflict ##{conflict_number} - Lines #{entry[:start_line]}-#{entry[:end_line]}
+              <span class="method-badge #{method_class}">#{entry[:resolution_method]}</span>
+            </div>
+            <table class="conflict-table">
+              <thead>
+                <tr>
+                  <th style="width: 25%">Base Branch Code</th>
+                  <th style="width: 25%">Incoming Branch Code</th>
+                  <th style="width: 25%">Final Resolution</th>
+                  <th style="width: 25%">Resolution Details</th>
+                </tr>
+              </thead>
+              <tbody>
+                <tr>
+                  <td>
+                    <div class="code-block base-code">#{escape_html(entry[:base_content] || 'N/A')}</div>
+                  </td>
+                  <td>
+                    <div class="code-block incoming-code">#{escape_html(entry[:incoming_content] || 'N/A')}</div>
+                  </td>
+                  <td>
+                    <div class="code-block resolution-code">#{escape_html(entry[:resolved_content] || 'N/A')}</div>
+                  </td>
+                  <td>
+                    <div class="code-block">
+                      <strong>Method:</strong> #{escape_html(entry[:resolution_method])}<br>
+                      <strong>Action:</strong> #{escape_html(entry[:action])}<br>
+                      <strong>Time:</strong> #{entry[:timestamp]}<br><br>
+                      #{entry[:llm_suggestion] ? "<strong>LLM Analysis:</strong><br>#{escape_html(entry[:llm_suggestion])}" : ''}
+                    </div>
+                  </td>
+                </tr>
+              </tbody>
+            </table>
+          </div>
+        CONFLICT_HTML
+      end
+
+      html += <<~FOOTER_HTML
+          <div class="footer">
+            Generated by N2B v#{N2B::VERSION} - AI-Powered Merge Conflict Resolution<br>
+            <small>This log contains the complete history of merge conflict resolutions for audit and review purposes.</small>
+          </div>
+        </body>
+        </html>
+      FOOTER_HTML
+
+      html
+    end
+
+    def escape_html(text)
+      return '' if text.nil?
+      text.to_s
+          .gsub('&', '&amp;')
+          .gsub('<', '&lt;')
+          .gsub('>', '&gt;')
+          .gsub('"', '&quot;')
+          .gsub("'", '&#39;')
+    end
+
+    def extract_git_info
+      branch = `git rev-parse --abbrev-ref HEAD 2>/dev/null`.strip rescue 'unknown'
+      {
+        branch: branch.empty? ? 'unknown' : branch
+      }
+    end
+
+    def determine_resolution_method(result)
+      return 'Aborted' if result[:abort]
+      return 'Manual Edit' if result[:reason]&.include?('manually resolved')
+      return 'Manual Choice' if result[:reason]&.include?('Manually selected')
+      return 'Skipped' if !result[:accepted] && !result[:abort]
+      return 'LLM Suggestion' if result[:accepted] && result[:reason]
+      'Unknown'
+    end
+
+    def determine_action(result)
+      return 'aborted' if result[:abort]
+      return 'accepted' if result[:accepted]
+      return 'skipped'
+    end
+
+    private # Ensure all subsequent methods are private unless specified
+
+    def get_vcs_type_for_file_operations
+      # Leverages existing get_vcs_type but more for file content retrieval context
+      get_vcs_type
+    end
+
+    def convert_label_to_revision(label, vcs_type, side)
+      # Convert conflict marker labels to actual VCS revision identifiers
+      case vcs_type
+      when :git
+        case label.downcase
+        when /head|working.*copy|current/
+          'HEAD'
+        when /merge.*rev|incoming|branch/
+          'MERGE_HEAD'
+        else
+          # If it's already a valid Git revision, use it as-is
+          label
+        end
+      when :hg
+        case label.downcase
+        when /working.*copy|current/
+          '.'  # Current working directory parent
+        when /merge.*rev|incoming|branch/
+          'p2()'  # Second parent (incoming branch)
+        else
+          # If it's already a valid Mercurial revision, use it as-is
+          label
+        end
+      else
+        label
+      end
+    end
+
+    def get_file_content_from_vcs(label, file_path, vcs_type)
+      # Note: file_path is the path in the working directory.
+      # VCS commands often need path relative to repo root if not run from root.
+      # Assuming execution from repo root or that file_path is appropriately relative.
+      # Pathname can help make it relative if needed:
+      # relative_file_path = Pathname.new(File.absolute_path(file_path)).relative_path_from(Pathname.new(Dir.pwd)).to_s
+
+      # A simpler approach for `git show` is that it usually works with paths from repo root.
+      # If @file_path is already relative to repo root or absolute, it might just work.
+      # For robustness, ensuring it's relative to repo root is better.
+      # However, current `execute_vcs_diff` uses `Dir.pwd`, implying commands are run from current dir.
+      # Let's assume file_path as given is suitable for now.
+
+      command = case vcs_type
+                when :git
+                  # For git show <commit-ish>:<path>, path is usually from repo root.
+                  # If @file_path is not from repo root, this might need adjustment.
+                  # Let's assume @file_path is correctly specified for this context.
+                  "git show #{Shellwords.escape(label)}:#{Shellwords.escape(file_path)}"
+                when :hg
+                  "hg cat -r #{Shellwords.escape(label)} #{Shellwords.escape(file_path)}"
+                else
+                  nil
+                end
+
+      return nil unless command
+
+      begin
+        # Timeout might be needed for very large files or slow VCS.
+        content = `#{command}` # Using backticks captures stdout
+        # Check $? for command success.
+        # `git show` returns 0 on success, non-zero otherwise (e.g. 128 if path not found).
+        # `hg cat` also returns 0 on success.
+        return content if $?.success?
+
+        # If command failed, log a warning but don't necessarily halt everything.
+        # The preview will just show empty for that panel.
+        puts "#{COLOR_YELLOW}Warning: VCS command '#{command}' failed or returned no content. Exit status: #{$?.exitstatus}.#{COLOR_RESET}"
+        nil
+      rescue StandardError => e
+        puts "#{COLOR_YELLOW}Warning: Could not fetch content for '#{file_path}' from VCS label '#{label}': #{e.message}#{COLOR_RESET}"
+        nil
+      end
+    end
+
+    def apply_hunk_to_full_content(original_full_content_with_markers, conflict_block_details, resolved_hunk_text)
+      return original_full_content_with_markers if resolved_hunk_text.nil?
+
+      lines = original_full_content_with_markers.lines.map(&:chomp)
+      # Convert 1-based line numbers from block_details to 0-based array indices
+      start_idx = conflict_block_details.start_line - 1
+      end_idx = conflict_block_details.end_line - 1
+
+      # Basic validation of indices
+      if start_idx < 0 || start_idx >= lines.length || end_idx < start_idx || end_idx >= lines.length
+        # This case should ideally not happen if block_details are correct
+        # Return original content or handle error appropriately
+        # For preview, it's safer to show the original content with markers if hunk application is problematic
+        return original_full_content_with_markers
+      end
+
+      hunk_lines = resolved_hunk_text.lines.map(&:chomp)
+
+      new_content_lines = []
+      new_content_lines.concat(lines[0...start_idx]) if start_idx > 0 # Lines before the conflict block
+      new_content_lines.concat(hunk_lines) # The resolved hunk
+      new_content_lines.concat(lines[(end_idx + 1)..-1]) if (end_idx + 1) < lines.length # Lines after the conflict block
+
+      new_content_lines.join("\n")
+    end
+
+    # --- HTML Preview Generation ---
+
+    # private (already established)
+
+    def get_language_class(file_path)
+      ext = File.extname(file_path).downcase
+      case ext
+      when '.rb' then 'ruby'
+      when '.js' then 'javascript'
+      when '.py' then 'python'
+      when '.java' then 'java'
+      when '.c', '.h', '.cpp', '.hpp' then 'cpp'
+      when '.cs' then 'csharp'
+      when '.go' then 'go'
+      when '.php' then 'php'
+      when '.ts' then 'typescript'
+      when '.swift' then 'swift'
+      when '.kt', '.kts' then 'kotlin'
+      when '.rs' then 'rust'
+      when '.scala' then 'scala'
+      when '.pl' then 'perl'
+      when '.pm' then 'perl'
+      when '.sh' then 'bash'
+      when '.html', '.htm', '.xhtml', '.xml' then 'xml' # Highlight.js uses 'xml' for HTML
+      when '.css' then 'css'
+      when '.json' then 'json'
+      when '.yml', '.yaml' then 'yaml'
+      when '.md', '.markdown' then 'markdown'
+      else
+        '' # Let Highlight.js auto-detect or default to plain text
+      end
+    end
+
+    def find_sub_content_lines(full_content, sub_content)
+      return nil if sub_content.nil? || sub_content.empty?
+      full_lines = full_content.lines.map(&:chomp)
+      sub_lines = sub_content.lines.map(&:chomp)
+
+      return nil if sub_lines.empty?
+
+      full_lines.each_with_index do |_, index|
+        match = true
+        sub_lines.each_with_index do |sub_line_content, sub_index|
+          unless full_lines[index + sub_index] == sub_line_content
+            match = false
+            break
+          end
+        end
+        return { start: index + 1, end: index + sub_lines.length } if match # 1-based line numbers
+      end
+      nil
+    end
+
+    def generate_conflict_preview_html(block_details, base_content_full, incoming_content_full, current_resolution_content_full, base_branch_name, incoming_branch_name, file_path, llm_suggestion = nil)
+      require 'cgi' # For CGI.escapeHTML
+      require 'fileutils' # For FileUtils.mkdir_p
+      require 'shellwords' # For Shellwords.escape, already used elsewhere but good to have contextually
+      require 'rbconfig' # For RbConfig::CONFIG
+
+      lang_class = get_language_class(file_path)
+
+      # Find line numbers for highlighting
+      # These are line numbers within their respective full_content strings (1-based)
+      base_highlight_lines = find_sub_content_lines(base_content_full, block_details.base_content)
+      incoming_highlight_lines = find_sub_content_lines(incoming_content_full, block_details.incoming_content)
+
+      # For resolution, highlight the conflict area within the full resolution content
+      # The current_resolution_content_full already contains the resolved content
+      # We'll highlight the same line range as the original conflict
+      resolution_highlight_lines = { start: block_details.start_line, end: block_details.end_line }
+
+      html_content = StringIO.new
+      html_content << "<!DOCTYPE html>\n<html lang=\"en\">\n<head>\n"
+      html_content << "  <meta charset=\"UTF-8\">\n"
+      html_content << "  <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n"
+      html_content << "  <title>Conflict Preview: #{CGI.escapeHTML(File.basename(file_path))}</title>\n"
+      html_content << "  <link rel=\"stylesheet\" href=\"https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/styles/default.min.css\">\n"
+      html_content << "  <script src=\"https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/highlight.min.js\"></script>\n"
+      # Optional: specific languages if needed, e.g.,
+      # html_content << "  <script src=\"https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/languages/#{lang_class}.min.js\"></script>\n" if lang_class && !lang_class.empty?
+
+      html_content << "  <style>\n"
+      html_content << "    body { font-family: sans-serif; margin: 0; display: flex; flex-direction: column; height: 100vh; }\n"
+      html_content << "    .header { padding: 10px; background-color: #f0f0f0; border-bottom: 1px solid #ccc; text-align: center; }\n"
+      html_content << "    .header h2 { margin: 0; }\n"
+      html_content << "    .llm-message { padding: 10px; background-color: #e8f4fd; border-bottom: 1px solid #ccc; margin: 0; }\n"
+      html_content << "    .llm-message h3 { margin: 0 0 5px 0; color: #1976d2; font-size: 0.9em; }\n"
+      html_content << "    .llm-message p { margin: 0; color: #424242; font-size: 0.85em; line-height: 1.3; }\n"
+      html_content << "    .columns-container { display: flex; flex: 1; overflow: hidden; }\n"
+      html_content << "    .column { flex: 1; padding: 0; border-left: 1px solid #ccc; overflow-y: auto; display: flex; flex-direction: column; }\n"
+      html_content << "    .column:first-child { border-left: none; }\n"
+      html_content << "    .column h3 { background-color: #e0e0e0; padding: 8px 10px; margin: 0; border-bottom: 1px solid #ccc; text-align: center; font-size: 1em; }\n"
+      html_content << "    .code-container { flex: 1; overflow-y: auto; position: relative; }\n"
+      html_content << "    pre { margin: 0; padding: 0; height: 100%; }\n"
+      html_content << "    code { display: block; padding: 10px 10px 10px 60px; font-family: 'SF Mono', Monaco, Inconsolata, 'Fira Code', monospace; font-size: 0.85em; line-height: 1.4em; }\n"
+      html_content << "    .line { display: block; position: relative; }\n"
+      html_content << "    .line-number { position: absolute; left: 0; width: 50px; padding-right: 10px; text-align: right; color: #999; user-select: none; font-size: 0.8em; background-color: #f8f8f8; border-right: 1px solid #e0e0e0; }\n"
+      html_content << "    .conflict-lines-base { background-color: #ffebee; border-left: 3px solid #f44336; }\n"
+      html_content << "    .conflict-lines-incoming { background-color: #e3f2fd; border-left: 3px solid #2196f3; }\n"
+      html_content << "    .conflict-lines-resolution { background-color: #e8f5e9; border-left: 3px solid #4caf50; }\n"
+      html_content << "    @media (max-width: 768px) { .columns-container { flex-direction: column; } .column { border-left: none; border-top: 1px solid #ccc;} }\n"
+      html_content << "  </style>\n</head>\n<body>\n"
+
+      html_content << "  <div class=\"header\"><h2>Conflict Preview: #{CGI.escapeHTML(file_path)}</h2></div>\n"
+
+      # Add LLM message section if available
+      if llm_suggestion && llm_suggestion['reason']
+        html_content << "  <div class=\"llm-message\">\n"
+        html_content << "    <h3>🤖 AI Analysis & Suggestion</h3>\n"
+        html_content << "    <p>#{CGI.escapeHTML(llm_suggestion['reason'])}</p>\n"
+        html_content << "  </div>\n"
+      end
+
+      html_content << "  <div class=\"columns-container\">\n"
+
+      # Helper to generate HTML for one column
+      generate_column_html = lambda do |title, full_code, highlight_info, highlight_class_suffix|
+        html_content << "    <div class=\"column\">\n"
+        html_content << "      <h3>#{CGI.escapeHTML(title)}</h3>\n"
+        html_content << "      <div class=\"code-container\">\n" # Wrapper for scrolling
+        html_content << "        <pre><code class=\"#{lang_class}\">\n"
+
+        full_code.lines.each_with_index do |line_text, index|
+          line_number = index + 1
+          line_class = "line"
+          if highlight_info && line_number >= highlight_info[:start] && line_number <= highlight_info[:end]
+            line_class += " conflict-lines conflict-lines-#{highlight_class_suffix}"
+          end
+          html_content << "<span class=\"#{line_class}\"><span class=\"line-number\">#{line_number}</span>#{CGI.escapeHTML(line_text.chomp)}</span>\n"
+        end
+
+        html_content << "        </code></pre>\n"
+        html_content << "      </div>\n" # end .code-container
+        html_content << "    </div>\n" # end .column
+      end
+
+      generate_column_html.call("Base (#{base_branch_name})", base_content_full, base_highlight_lines, "base")
+      generate_column_html.call("Incoming (#{incoming_branch_name})", incoming_content_full, incoming_highlight_lines, "incoming")
+      generate_column_html.call("Current Resolution", current_resolution_content_full, resolution_highlight_lines, "resolution")
+
+      html_content << "  </div>\n" # end .columns-container
+      html_content << "  <script>hljs.highlightAll();</script>\n"
+      html_content << "</body>\n</html>"
+
+      # Save to file with unique but persistent naming
+      log_dir = '.n2b_merge_log'
+      FileUtils.mkdir_p(log_dir)
+
+      # Create a more stable filename based on file path and conflict location
+      file_basename = File.basename(file_path, '.*')
+      conflict_id = "#{block_details.start_line}_#{block_details.end_line}"
+      preview_filename = "conflict_#{file_basename}_lines_#{conflict_id}.html"
+      full_preview_path = File.join(log_dir, preview_filename)
+
+      File.write(full_preview_path, html_content.string)
+
+      return File.absolute_path(full_preview_path)
+    end
+
+    def open_html_in_browser(html_file_path)
+      absolute_path = File.absolute_path(html_file_path)
+      # Ensure correct file URL format for different OSes, especially Windows
+      file_url = if RbConfig::CONFIG['host_os'] =~ /mswin|mingw|cygwin/
+                   "file:///#{absolute_path.gsub("\\", "/")}" # Windows needs forward slashes
+                 else
+                   "file://#{absolute_path}"
+                 end
+
+      command = nil
+      os = RbConfig::CONFIG['host_os']
+
+      case os
+      when /darwin|mac os/
+        command = "open #{Shellwords.escape(file_url)}"
+      when /linux/
+        # Check for WSL environment, as xdg-open might not work as expected directly
+        # or might open browser inside WSL, not on Windows host.
+        # Powershell.exe can be used to open it on Windows host from WSL.
+        if ENV['WSL_DISTRO_NAME'] || (ENV['IS_WSL'] == 'true') || File.exist?('/proc/sys/fs/binfmt_misc/WSLInterop')
+          # Using powershell.exe to open the URL on the Windows host
+          # Ensure the file_url is accessible from Windows (e.g. via /mnt/c/...)
+          # This assumes the path is already a Windows-accessible path if running in WSL context
+          # or that the user has set up their environment for this.
+          # For file URLs, it's often better to translate to Windows path format.
+          windows_path = absolute_path.gsub(%r{^/mnt/([a-z])}, '\1:') # Basic /mnt/c -> C:
+          command = "powershell.exe -c \"Start-Process '#{windows_path}'\""
+          puts "#{COLOR_YELLOW}Detected WSL, attempting to open in Windows browser: #{command}#{COLOR_RESET}"
+        else
+          command = "xdg-open #{Shellwords.escape(file_url)}"
+        end
+      when /mswin|mingw|cygwin/ # Windows
+        # `start` command with an empty title "" for paths with spaces
+        command = "start \"\" \"#{file_url.gsub("file:///", "")}\"" # `start` takes path directly
+      else
+        puts "#{COLOR_YELLOW}Unsupported OS: #{os}. Cannot open browser automatically.#{COLOR_RESET}"
+        return false
+      end
+
+      if command
+        puts "#{COLOR_BLUE}Attempting to open preview in browser: #{command}#{COLOR_RESET}"
+        begin
+          success = system(command)
+          unless success
+            # system() returns false if command executes with non-zero status, nil if command fails to execute
+            puts "#{COLOR_RED}Failed to execute command to open browser. Exit status: #{$?.exitstatus if $?}.#{COLOR_RESET}"
+            raise "Command execution failed" # Will be caught by rescue
+          end
+          puts "#{COLOR_GREEN}Preview should now be open in your browser.#{COLOR_RESET}"
+          return true
+        rescue => e
+          puts "#{COLOR_RED}Failed to automatically open the HTML preview: #{e.message}#{COLOR_RESET}"
+        end
+      end
+
+      puts "#{COLOR_YELLOW}Please open it manually: #{file_url}#{COLOR_RESET}"
+      false
+    end
+
+  end
+end