aline-ai 0.2.6__py3-none-any.whl → 0.3.0__py3-none-any.whl

realign/commands/export_shares.py

@@ -0,0 +1,1785 @@
+#!/usr/bin/env python3
+"""
+Export shares command - Export selected commits' chat history to JSON files.
+
+This allows users to select specific commits and extract their chat history changes
+into standalone JSON files for sharing.
+"""
+
+import json
+import os
+import subprocess
+import sys
+import secrets
+import hashlib
+import base64
+from collections import defaultdict
+from datetime import datetime
+from pathlib import Path
+from typing import List, Dict, Optional, Tuple
+
+from .review import get_unpushed_commits, UnpushedCommit
+from .hide import parse_commit_indices
+from ..logging_config import setup_logger
+
+logger = setup_logger('realign.commands.export_shares', 'export_shares.log')
+
+# Try to import cryptography
+try:
+    from cryptography.hazmat.primitives.ciphers import Cipher, algorithms, modes
+    from cryptography.hazmat.primitives.kdf.pbkdf2 import PBKDF2HMAC
+    from cryptography.hazmat.primitives import hashes
+    from cryptography.hazmat.backends import default_backend
+    CRYPTO_AVAILABLE = True
+except ImportError:
+    CRYPTO_AVAILABLE = False
+    logger.warning("cryptography package not available, interactive mode disabled")
+
+# Try to import httpx
+try:
+    import httpx
+    HTTPX_AVAILABLE = True
+except ImportError:
+    HTTPX_AVAILABLE = False
+    logger.warning("httpx package not available, interactive mode disabled")
+
+
+def get_line_timestamp(line_json: dict) -> datetime:
+    """
+    从 JSON 对象中提取时间戳。
+
+    Args:
+        line_json: JSON 对象
+
+    Returns:
+        datetime 对象，如果没有 timestamp 则返回 datetime.min
+    """
+    if 'timestamp' in line_json:
+        ts_str = line_json['timestamp']
+        # 处理 ISO 格式: "2025-12-07T17:54:42.618Z"
+        return datetime.fromisoformat(ts_str.replace('Z', '+00:00'))
+    return datetime.min  # 没有时间戳的放在最前面
+
+
+def get_session_id(line_json: dict) -> Optional[str]:
+    """
+    从 JSON 对象中提取 session ID。
+
+    Args:
+        line_json: JSON 对象
+
+    Returns:
+        session ID 字符串，如果没有则返回 None
+    """
+    return line_json.get('sessionId')
+
+
+def extract_messages_from_commit(
+    commit: UnpushedCommit,
+    repo_root: Path
+) -> Dict[str, List[Tuple[datetime, dict]]]:
+    """
+    从单个 commit 提取所有新增消息，按 session ID 分组。
+
+    Args:
+        commit: UnpushedCommit 对象
+        repo_root: shadow git 仓库路径
+
+    Returns:
+        字典: {session_id: [(timestamp, json_object), ...]}
+    """
+    logger.info(f"Extracting messages from commit {commit.hash}")
+    session_messages = defaultdict(list)
+
+    for session_file, line_ranges in commit.session_additions.items():
+        logger.debug(f"Processing session file: {session_file}")
+
+        # 获取文件内容
+        result = subprocess.run(
+            ["git", "show", f"{commit.full_hash}:{session_file}"],
+            cwd=repo_root,
+            capture_output=True,
+            text=True
+        )
+
+        if result.returncode != 0:
+            logger.warning(f"Failed to get content for {session_file}")
+            continue
+
+        lines = result.stdout.split('\n')
+
+        # 提取新增行
+        for start, end in line_ranges:
+            for line_num in range(start, end + 1):
+                if line_num <= len(lines):
+                    line = lines[line_num - 1].strip()
+
+                    if not line or "[REDACTED]" in line:
+                        continue
+
+                    try:
+                        json_obj = json.loads(line)
+
+                        # 跳过已被标记为 redacted 的内容
+                        if json_obj.get("redacted"):
+                            continue
+
+                        session_id = get_session_id(json_obj)
+                        if session_id:
+                            timestamp = get_line_timestamp(json_obj)
+                            session_messages[session_id].append((timestamp, json_obj))
+                            logger.debug(f"Extracted message from session {session_id}")
+
+                    except json.JSONDecodeError as e:
+                        logger.warning(f"Failed to parse JSON at line {line_num}: {e}")
+                        continue
+
+    logger.info(f"Extracted {sum(len(msgs) for msgs in session_messages.values())} messages from {len(session_messages)} sessions")
+    return session_messages
+
+
+def merge_messages_from_commits(
+    selected_commits: List[UnpushedCommit],
+    repo_root: Path
+) -> Dict[str, List[dict]]:
+    """
+    合并所有选中 commits 的消息，按 session ID 分组并排序。
+
+    Args:
+        selected_commits: 选中的 commits 列表
+        repo_root: shadow git 仓库路径
+
+    Returns:
+        字典: {session_id: [sorted_json_objects]}
+    """
+    logger.info(f"Merging messages from {len(selected_commits)} commits")
+    all_session_messages = defaultdict(list)
+
+    # 按时间顺序处理 commits（从旧到新）
+    for commit in reversed(selected_commits):
+        commit_messages = extract_messages_from_commit(commit, repo_root)
+
+        for session_id, messages in commit_messages.items():
+            all_session_messages[session_id].extend(messages)
+
+    # 排序和去重
+    result = {}
+    for session_id, messages_with_ts in all_session_messages.items():
+        # 按时间戳排序
+        sorted_messages = sorted(messages_with_ts, key=lambda x: x[0])
+
+        # 去重（基于 JSON 字符串）
+        seen = set()
+        unique_messages = []
+        for ts, msg in sorted_messages:
+            msg_str = json.dumps(msg, sort_keys=True, ensure_ascii=False)
+            if msg_str not in seen:
+                seen.add(msg_str)
+                unique_messages.append(msg)
+
+        result[session_id] = unique_messages
+        logger.info(f"Session {session_id}: {len(unique_messages)} unique messages")
+
+    return result
+
+
+def save_export_file(
+    session_messages: Dict[str, List[dict]],
+    output_dir: Path,
+    username: str
+) -> Path:
+    """
+    保存导出文件。
+
+    Args:
+        session_messages: {session_id: [messages]}
+        output_dir: 输出目录
+        username: 用户名
+
+    Returns:
+        导出文件路径
+    """
+    logger.info(f"Saving export file to {output_dir}")
+
+    # 创建输出目录
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    # 生成文件名: username_timestamp.json
+    timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
+    filename = f"{username}_{timestamp}.json"
+    output_path = output_dir / filename
+
+    # 构建导出数据
+    export_data = {
+        "username": username,
+        "time": datetime.now().isoformat(),
+        "sessions": [
+            {
+                "session_id": session_id,
+                "messages": messages
+            }
+            for session_id, messages in session_messages.items()
+        ]
+    }
+
+    # 写入文件
+    with open(output_path, 'w', encoding='utf-8') as f:
+        json.dump(export_data, f, indent=2, ensure_ascii=False)
+
+    logger.info(f"Export file saved to {output_path}")
+    return output_path
+
+
+def display_commits_for_selection(commits: List[UnpushedCommit]) -> None:
+    """
+    显示 commits 供用户选择。
+
+    Args:
+        commits: UnpushedCommit 列表
+    """
+    print(f"\n📋 Available commits ({len(commits)}):\n")
+
+    for commit in commits:
+        # Display format: [index] hash - message
+        print(f"  [{commit.index}] {commit.hash} - {commit.message}")
+
+        # Show user request if available
+        if commit.user_request:
+            request_preview = commit.user_request[:60]
+            if len(commit.user_request) > 60:
+                request_preview += "..."
+            print(f"      └─ {request_preview}")
+
+    print()
+
+
+def export_shares_command(
+    indices: Optional[str] = None,
+    username: Optional[str] = None,
+    repo_root: Optional[Path] = None,
+    output_dir: Optional[Path] = None
+) -> int:
+    """
+    Main entry point for export shares command.
+
+    Allows users to select commits and export their chat history to JSON files.
+
+    Args:
+        indices: Commit indices to export (e.g., "1,3,5-7"). If None, prompts user.
+        username: Username for the export. If None, uses system username.
+        repo_root: Path to user's project root (auto-detected if None)
+        output_dir: Custom output directory. If None, uses ~/.aline/{project}/share/
+
+    Returns:
+        0 on success, 1 on error
+    """
+    logger.info("======== Export shares command started ========")
+
+    # Auto-detect user project root if not provided
+    if repo_root is None:
+        try:
+            result = subprocess.run(
+                ["git", "rev-parse", "--show-toplevel"],
+                capture_output=True,
+                text=True,
+                check=True
+            )
+            repo_root = Path(result.stdout.strip())
+            logger.debug(f"Detected user project root: {repo_root}")
+        except subprocess.CalledProcessError:
+            print("Error: Not in a git repository", file=sys.stderr)
+            logger.error("Not in a git repository")
+            return 1
+
+    # Get shadow git repository path
+    from .. import get_realign_dir
+    shadow_dir = get_realign_dir(repo_root)
+    shadow_git = shadow_dir
+
+    # Verify shadow git exists
+    if not shadow_git.exists():
+        print(f"Error: ReAlign repository not found at {shadow_git}", file=sys.stderr)
+        print("Run 'aline init' first to initialize the repository.", file=sys.stderr)
+        logger.error(f"Shadow git not found at {shadow_git}")
+        return 1
+
+    # Check if it's a git repository
+    git_dir = shadow_git / '.git'
+    if not git_dir.exists():
+        print(f"Error: {shadow_git} is not a git repository", file=sys.stderr)
+        print("Run 'aline init' first to initialize the repository.", file=sys.stderr)
+        logger.error(f"No .git found in {shadow_git}")
+        return 1
+
+    logger.info(f"Using shadow git repository: {shadow_git}")
+
+    # Get username
+    if username is None:
+        username = os.environ.get('USER') or os.environ.get('USERNAME') or 'user'
+
+    logger.debug(f"Using username: {username}")
+
+    # Set output directory
+    if output_dir is None:
+        output_dir = shadow_dir / "share"
+
+    logger.debug(f"Output directory: {output_dir}")
+
+    # Get all unpushed commits from shadow git
+    try:
+        all_commits = get_unpushed_commits(shadow_git)
+    except Exception as e:
+        print(f"Error: Failed to get commits: {e}", file=sys.stderr)
+        logger.error(f"Failed to get commits: {e}", exc_info=True)
+        return 1
+
+    if not all_commits:
+        print("No unpushed commits found. Nothing to export.", file=sys.stderr)
+        logger.info("No unpushed commits found")
+        return 1
+
+    # Get commit selection
+    if indices is None:
+        # Interactive mode: display commits and prompt user
+        display_commits_for_selection(all_commits)
+
+        print("Enter commit indices to export (e.g., '1,3,5-7' or 'all'):")
+        indices_input = input("Indices: ").strip()
+
+        if not indices_input:
+            print("No commits selected. Exiting.")
+            logger.info("No commits selected by user")
+            return 0
+    else:
+        indices_input = indices
+
+    # Parse indices
+    try:
+        if indices_input.lower() == "all":
+            indices_list = [c.index for c in all_commits]
+        else:
+            indices_list = parse_commit_indices(indices_input)
+    except ValueError as e:
+        print(f"Error: Invalid indices format: {e}", file=sys.stderr)
+        logger.error(f"Invalid indices format: {e}")
+        return 1
+
+    # Validate indices
+    max_index = len(all_commits)
+    invalid_indices = [i for i in indices_list if i < 1 or i > max_index]
+    if invalid_indices:
+        print(f"Error: Invalid indices (out of range 1-{max_index}): {invalid_indices}", file=sys.stderr)
+        logger.error(f"Invalid indices: {invalid_indices}")
+        return 1
+
+    # Get selected commits
+    selected_commits = [c for c in all_commits if c.index in indices_list]
+
+    logger.info(f"Selected {len(selected_commits)} commit(s) to export")
+
+    # Merge messages from commits
+    print(f"\n🔄 Extracting chat history from {len(selected_commits)} commit(s)...")
+    try:
+        session_messages = merge_messages_from_commits(selected_commits, shadow_git)
+    except Exception as e:
+        print(f"\nError: Failed to extract messages: {e}", file=sys.stderr)
+        logger.error(f"Failed to extract messages: {e}", exc_info=True)
+        return 1
+
+    if not session_messages:
+        print("\nWarning: No chat history found in selected commits.", file=sys.stderr)
+        logger.warning("No chat history found in selected commits")
+        return 1
+
+    # Save export file
+    try:
+        output_path = save_export_file(session_messages, output_dir, username)
+    except Exception as e:
+        print(f"\nError: Failed to save export file: {e}", file=sys.stderr)
+        logger.error(f"Failed to save export file: {e}", exc_info=True)
+        return 1
+
+    # Success message
+    total_messages = sum(len(msgs) for msgs in session_messages.values())
+    print(f"\n✅ Successfully exported {len(session_messages)} session(s)")
+    print(f"📁 Export file: {output_path}")
+    print(f"📊 Total messages: {total_messages}")
+    print()
+
+    logger.info(f"======== Export shares command completed: {output_path} ========")
+    return 0
+
+
+def encrypt_conversation_data(data: dict, password: str) -> dict:
+    """
+    使用 AES-256-GCM 加密对话数据
+
+    Args:
+        data: 要加密的数据字典
+        password: 加密密码
+
+    Returns:
+        包含加密数据的字典: {encrypted_data, salt, nonce, password_hash}
+    """
+    if not CRYPTO_AVAILABLE:
+        raise RuntimeError("cryptography package not installed. Run: pip install cryptography")
+
+    # 生成盐值和随机数
+    salt = os.urandom(32)
+    nonce = os.urandom(12)
+
+    # 密钥派生
+    kdf = PBKDF2HMAC(
+        algorithm=hashes.SHA256(),
+        length=32,
+        salt=salt,
+        iterations=1000,
+        backend=default_backend()
+    )
+    key = kdf.derive(password.encode())
+
+    # 加密数据
+    cipher = Cipher(algorithms.AES(key), modes.GCM(nonce), backend=default_backend())
+    encryptor = cipher.encryptor()
+
+    json_data = json.dumps(data, ensure_ascii=False).encode('utf-8')
+    ciphertext = encryptor.update(json_data) + encryptor.finalize()
+
+    # 添加认证标签
+    ciphertext_with_tag = ciphertext + encryptor.tag
+
+    # 计算密码 hash
+    password_hash = hashlib.sha256(password.encode()).hexdigest()
+
+    return {
+        "encrypted_data": base64.b64encode(ciphertext_with_tag).decode('ascii'),
+        "salt": base64.b64encode(salt).decode('ascii'),
+        "nonce": base64.b64encode(nonce).decode('ascii'),
+        "password_hash": password_hash
+    }
+
+
+def upload_to_backend(
+    encrypted_payload: dict,
+    metadata: dict,
+    backend_url: str
+) -> dict:
+    """
+    上传加密数据到后端服务器
+
+    Args:
+        encrypted_payload: 加密后的数据
+        metadata: 元数据
+        backend_url: 后端 URL
+
+    Returns:
+        包含 share_id 和 share_url 的字典
+    """
+    if not HTTPX_AVAILABLE:
+        raise RuntimeError("httpx package not installed. Run: pip install httpx")
+
+    try:
+        response = httpx.post(
+            f"{backend_url}/api/share/create",
+            json={
+                "encrypted_payload": encrypted_payload,
+                "metadata": metadata
+            },
+            timeout=30.0
+        )
+        response.raise_for_status()
+        return response.json()
+    except httpx.HTTPError as e:
+        logger.error(f"Upload failed: {e}")
+        raise RuntimeError(f"Failed to upload to server: {e}")
+
+
+def clean_text_for_prompt(text: str) -> str:
+    """
+    清理文本中的控制字符，使其适合在 LLM prompt 中使用
+
+    Args:
+        text: 原始文本
+
+    Returns:
+        清理后的文本
+    """
+    # 替换控制字符为空格或删除
+    import re
+    # 保留换行符和制表符，删除其他控制字符
+    text = re.sub(r'[\x00-\x08\x0B\x0C\x0E-\x1F\x7F]', '', text)
+    # 将多个连续空格合并为一个
+    text = re.sub(r'\s+', ' ', text)
+    return text.strip()
+
+
+def generate_ui_metadata_with_llm(
+    conversation_data: dict,
+    selected_commits: List,
+    provider: str = "auto",
+    preset_id: str = "default"
+) -> Tuple[Optional[dict], Optional[dict]]:
+    """
+    使用 LLM 根据对话内容生成个性化的 UI 元数据
+
+    Args:
+        conversation_data: 对话数据字典 {username, time, sessions}
+        selected_commits: 选中的 UnpushedCommit 列表
+        provider: LLM provider ("auto", "claude", "openai")
+        preset_id: Prompt preset ID，用于调整生成风格
+
+    Returns:
+        Tuple[ui_metadata, debug_info]
+        - ui_metadata: UI 元数据字典，包含 title, welcome, description, preset_questions
+        - debug_info: {system_prompt, user_prompt, response_text, provider} 或 None
+        如果生成失败则都返回 None
+    """
+    logger.info(f"Generating UI metadata with LLM (preset: {preset_id})")
+
+    # 构建对话内容摘要
+    sessions = conversation_data.get("sessions", [])
+    total_messages = sum(len(s.get("messages", [])) for s in sessions)
+
+    # 提取 commit 中的 LLM summary 和 user request
+    commit_summaries = []
+    user_requests = []
+
+    for commit in selected_commits:
+        if commit.llm_summary and commit.llm_summary.strip():
+            # 清理控制字符
+            cleaned_summary = clean_text_for_prompt(commit.llm_summary)
+            commit_summaries.append(cleaned_summary)
+        if commit.user_request and commit.user_request.strip():
+            # 清理控制字符并截取前300字符
+            cleaned_request = clean_text_for_prompt(commit.user_request[:300])
+            user_requests.append(cleaned_request)
+
+    # 如果没有 commit summary，回退到提取消息样本
+    if not commit_summaries:
+        logger.warning("No commit summaries found, falling back to message samples")
+        user_messages = []
+        assistant_messages = []
+
+        for session in sessions[:5]:  # 只看前5个session
+            for msg in session.get("messages", [])[:10]:  # 每个session前10条消息
+                content = msg.get("content", "")
+                if isinstance(content, str) and content.strip():
+                    # 清理控制字符并截取
+                    cleaned_content = clean_text_for_prompt(content[:200])
+                    if msg.get("role") == "user":
+                        user_messages.append(cleaned_content)
+                    elif msg.get("role") == "assistant":
+                        assistant_messages.append(cleaned_content)
+
+    # 根据 preset_id 定制 system_prompt
+    preset_configs = {
+        "default": {
+            "role_description": "a general-purpose conversation assistant",
+            "title_style": "a neutral, descriptive summary of the topic",
+            "welcome_tone": "friendly and informative, with a brief overview of the conversation",
+            "description_focus": "what information can be found and how the assistant can help",
+            "question_angles": [
+                "high-level summary",
+                "technical or implementation details",
+                "decision-making or reasoning",
+                "results, impact, or follow-up"
+            ]
+        },
+        "work-report": {
+            "role_description": "a professional work report agent representing the user to colleagues/managers",
+            "title_style": "a professional, achievement-oriented summary (e.g., 'Progress on Project X', 'Completed Tasks for Week Y')",
+            "welcome_tone": "professional and confident, highlighting accomplishments and progress",
+            "description_focus": "what work was done, what value was created, and how the assistant represents the user's contributions",
+            "question_angles": [
+                "overall progress and achievements",
+                "technical solutions implemented",
+                "challenges overcome and decisions made",
+                "next steps and impact on project goals"
+            ]
+        },
+        "knowledge-agent": {
+            "role_description": "a knowledge-sharing agent representing the user's deep thinking as founder/architect/author",
+            "title_style": "a thought-provoking, conceptual title (e.g., 'Design Philosophy of Feature X', 'Architectural Decisions for System Y')",
+            "welcome_tone": "insightful and educational, emphasizing the thinking process and context behind decisions",
+            "description_focus": "the knowledge and insights shared, the reasoning behind decisions, and how the assistant helps others understand the user's thought process",
+            "question_angles": [
+                "core concepts and philosophy",
+                "design rationale and trade-offs",
+                "key insights and learning",
+                "practical implications and applications"
+            ]
+        },
+        "personality-analyzer": {
+            "role_description": "a personality analysis assistant that understands the user's characteristics based on conversation",
+            "title_style": "an analytical, personality-focused title (e.g., 'Minhao's Working Style Analysis', 'Communication Pattern Insights')",
+            "welcome_tone": "analytical yet friendly, introducing what aspects of personality can be explored",
+            "description_focus": "what personality traits, working styles, and communication patterns can be discovered from the conversation",
+            "question_angles": [
+                "overall personality traits and characteristics",
+                "working style and approach to problem-solving",
+                "communication patterns and preferences",
+                "strengths, growth areas, and unique qualities"
+            ]
+        }
+    }
+
+    # 获取 preset 配置，如果没有则使用默认
+    preset_config = preset_configs.get(preset_id, preset_configs["default"])
+
+    # 构建 LLM prompt
+    system_prompt = f"""You are a conversation interface copy generator for {preset_config['role_description']}.
+
+Your task is to analyze a given conversation history and generate
+personalized UI copy for a chat-based assistant that helps users
+understand and explore that conversation.
+
+Return the result strictly in JSON format:
+
+{{
+  "title": "A highly abstract and concise title (30–50 characters) that is {preset_config['title_style']}.",
+
+  "welcome": "A welcome message (60-120 characters) that is {preset_config['welcome_tone']}.
+              The message should provide context about what this conversation contains and set the right tone.",
+
+  "description": "A clear description (30-80 characters) that explains {preset_config['description_focus']}.",
+
+  "preset_questions": [
+    "Question 1: About {preset_config['question_angles'][0]} (15–30 characters)",
+    "Question 2: About {preset_config['question_angles'][1]} (15–30 characters)",
+    "Question 3: About {preset_config['question_angles'][2]} (15–30 characters)",
+    "Question 4: About {preset_config['question_angles'][3]} (15–30 characters)"
+  ]
+}}
+
+Requirements:
+1. The title must align with the preset's purpose and tone.
+2. The welcome message should match the preset's role and create appropriate expectations.
+3. The description should clearly explain the assistant's specific purpose for this preset.
+4. Preset questions must be based on the actual conversation content, concrete, and useful from the specified angles.
+5. All text must be in English or Chinese, depending on the conversation language.
+6. Output JSON only. Do not include any additional explanation or text."""
+
+    # 构建 user prompt - 优先使用 commit summaries
+    if commit_summaries:
+        user_prompt = f"""Analyze the following conversation history and generate UI copy:
+
+Number of sessions: {len(sessions)}
+Total messages: {total_messages}
+Number of commits included: {len(commit_summaries)}
+
+LLM summaries from each commit:
+{chr(10).join(f"{i+1}. {summary}" for i, summary in enumerate(commit_summaries))}
+
+User's main requests:
+{chr(10).join(f"- {req}" for req in user_requests[:10]) if user_requests else "None"}
+
+Return the UI copy in JSON format."""
+    else:
+        # 回退到使用消息样本
+        user_prompt = f"""Analyze the following conversation history and generate UI copy:
+
+Number of sessions: {len(sessions)}
+Total messages: {total_messages}
+
+User message samples:
+{chr(10).join(user_messages[:10])}
+
+Assistant reply samples:
+{chr(10).join(assistant_messages[:10])}
+
+Return the UI copy in JSON format."""
+
+    # 尝试调用 LLM
+    anthropic_key = os.getenv("ANTHROPIC_API_KEY")
+    openai_key = os.getenv("OPENAI_API_KEY")
+
+    try_claude = provider in ("auto", "claude")
+    try_openai = provider in ("auto", "openai")
+
+    # 尝试 Claude
+    if try_claude and anthropic_key:
+        print("   → Generating UI metadata with Anthropic Claude...", file=sys.stderr)
+        try:
+            import anthropic
+            client = anthropic.Anthropic(api_key=anthropic_key)
+
+            response = client.messages.create(
+                model="claude-3-5-haiku-20241022",
+                max_tokens=1000,
+                temperature=0.7,
+                system=system_prompt,
+                messages=[{"role": "user", "content": user_prompt}]
+            )
+
+            response_text = response.content[0].text.strip()
+            logger.debug(f"Claude response: {response_text}")
+
+            # 解析 JSON
+            json_str = response_text
+            if "```json" in response_text:
+                json_start = response_text.find("```json") + 7
+                json_end = response_text.find("```", json_start)
+                if json_end != -1:
+                    json_str = response_text[json_start:json_end].strip()
+            elif "```" in response_text:
+                json_start = response_text.find("```") + 3
+                json_end = response_text.find("```", json_start)
+                if json_end != -1:
+                    json_str = response_text[json_start:json_end].strip()
+
+            # 清理控制字符，避免 JSON 解析错误
+            import re
+            # 移除非法的控制字符，但保留合法的空白字符（空格、制表符、换行）
+            json_str = re.sub(r'[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]', '', json_str)
+
+            ui_metadata = json.loads(json_str)
+            print("   ✅ UI metadata generated successfully with Claude", file=sys.stderr)
+
+            # Return with debug info
+            debug_info = {
+                "system_prompt": system_prompt,
+                "user_prompt": user_prompt,
+                "response_text": response_text,
+                "provider": "claude"
+            }
+            return ui_metadata, debug_info
+
+        except ImportError:
+            if provider == "claude":
+                print("   ❌ Anthropic package not installed", file=sys.stderr)
+                return None, None
+            print("   ⚠️  Anthropic package not installed, trying OpenAI...", file=sys.stderr)
+        except Exception as e:
+            logger.error(f"Claude API error: {e}", exc_info=True)
+            if provider == "claude":
+                print(f"   ❌ Claude API error: {e}", file=sys.stderr)
+                return None, None
+            print(f"   ⚠️  Claude failed, trying OpenAI...", file=sys.stderr)
+
+    # 尝试 OpenAI
+    if try_openai and openai_key:
+        print("   → Generating UI metadata with OpenAI...", file=sys.stderr)
+        try:
+            import openai
+            client = openai.OpenAI(api_key=openai_key)
+
+            response = client.chat.completions.create(
+                model="gpt-3.5-turbo",
+                messages=[
+                    {"role": "system", "content": system_prompt},
+                    {"role": "user", "content": user_prompt}
+                ],
+                max_tokens=1000,
+                temperature=0.7
+            )
+
+            response_text = response.choices[0].message.content.strip()
+            logger.debug(f"OpenAI response: {response_text}")
+
+            # 解析 JSON
+            json_str = response_text
+            if "```json" in response_text:
+                json_start = response_text.find("```json") + 7
+                json_end = response_text.find("```", json_start)
+                if json_end != -1:
+                    json_str = response_text[json_start:json_end].strip()
+            elif "```" in response_text:
+                json_start = response_text.find("```") + 3
+                json_end = response_text.find("```", json_start)
+                if json_end != -1:
+                    json_str = response_text[json_start:json_end].strip()
+
+            # 清理控制字符，避免 JSON 解析错误
+            import re
+            # 移除非法的控制字符，但保留合法的空白字符（空格、制表符、换行）
+            json_str = re.sub(r'[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]', '', json_str)
+
+            ui_metadata = json.loads(json_str)
+            print("   ✅ UI metadata generated successfully with OpenAI", file=sys.stderr)
+
+            # Return with debug info
+            debug_info = {
+                "system_prompt": system_prompt,
+                "user_prompt": user_prompt,
+                "response_text": response_text,
+                "provider": "openai"
+            }
+            return ui_metadata, debug_info
+
+        except ImportError:
+            print("   ❌ OpenAI package not installed", file=sys.stderr)
+            return None, None
+        except Exception as e:
+            logger.error(f"OpenAI API error: {e}", exc_info=True)
+            print(f"   ❌ OpenAI API error: {e}", file=sys.stderr)
+            return None, None
+
+    # 没有可用的 API
+    logger.warning("No LLM API keys available for UI metadata generation")
+    print("   ⚠️  No LLM API keys configured, using default UI text", file=sys.stderr)
+    return None, None
+
+
+def save_export_log(
+    conversation_data: dict,
+    llm_prompts: Optional[dict],
+    llm_response: Optional[str],
+    username: str,
+    shadow_dir: Path
+) -> Path:
+    """
+    保存导出日志
+
+    Args:
+        conversation_data: 分享的对话数据
+        llm_prompts: LLM prompts字典 {system_prompt, user_prompt}
+        llm_response: LLM的原始回答
+        username: 用户名
+        shadow_dir: ReAlign目录 (~/.aline/{project}/)
+
+    Returns:
+        保存的日志文件路径
+    """
+    # 创建日志目录
+    log_dir = shadow_dir / "share" / "export_logs"
+    log_dir.mkdir(parents=True, exist_ok=True)
+
+    # 生成文件名
+    timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
+    log_file = log_dir / f"{username}_{timestamp}.log.json"
+
+    # 构建日志内容
+    log_content = {
+        "export_time": datetime.now().isoformat(),
+        "username": username,
+        "conversation_data": conversation_data,
+        "llm_generation": {
+            "prompts": llm_prompts,
+            "response": llm_response
+        } if llm_prompts else None
+    }
+
+    # 保存
+    with open(log_file, 'w', encoding='utf-8') as f:
+        json.dump(log_content, f, indent=2, ensure_ascii=False)
+
+    return log_file
+
+
+def display_selection_statistics(
+    selected_commits: List[UnpushedCommit],
+    session_messages: Dict[str, List[dict]]
+) -> None:
+    """
+    显示选择的统计信息
+
+    Args:
+        selected_commits: 选中的commits
+        session_messages: 合并后的session消息
+    """
+    try:
+        from rich.console import Console
+        from rich.panel import Panel
+    except ImportError:
+        # Fallback to plain text if rich is not available
+        print(f"\n📊 Selection Summary:")
+        print(f"  Commits: {len(selected_commits)}")
+        print(f"  Sessions: {len(session_messages)}")
+        total_messages = sum(len(msgs) for msgs in session_messages.values())
+        print(f"  Messages: {total_messages}")
+        return
+
+    console = Console()
+
+    # 计算统计信息
+    total_sessions = len(session_messages)
+    total_messages = sum(len(msgs) for msgs in session_messages.values())
+
+    # 消息角色分布
+    user_messages = 0
+    assistant_messages = 0
+    other_messages = 0
+
+    for messages in session_messages.values():
+        for msg in messages:
+            role = msg.get('role', 'unknown')
+            if role == 'user':
+                user_messages += 1
+            elif role == 'assistant':
+                assistant_messages += 1
+            else:
+                other_messages += 1
+
+    # 时间范围
+    all_timestamps = []
+    for messages in session_messages.values():
+        for msg in messages:
+            if 'timestamp' in msg:
+                ts_str = msg['timestamp']
+                try:
+                    ts = datetime.fromisoformat(ts_str.replace('Z', '+00:00'))
+                    all_timestamps.append(ts)
+                except:
+                    pass
+
+    time_info = ""
+    if all_timestamps:
+        earliest = min(all_timestamps)
+        latest = max(all_timestamps)
+        duration = latest - earliest
+
+        time_info = f"""
+[bold]Time Range:[/bold]
+  {earliest.strftime('%Y-%m-%d %H:%M')} → {latest.strftime('%Y-%m-%d %H:%M')}
+  Duration: {duration.days}d {duration.seconds//3600}h {(duration.seconds%3600)//60}m
+"""
+
+    # 构建统计文本
+    stats_text = f"""[bold cyan]Commits Selected:[/bold cyan] {len(selected_commits)}
+
+[bold green]Sessions:[/bold green] {total_sessions}
+
+[bold yellow]Messages:[/bold yellow] {total_messages}
+  ├─ User:      {user_messages}
+  ├─ Assistant: {assistant_messages}"""
+
+    if other_messages > 0:
+        stats_text += f"\n  └─ Other:     {other_messages}"
+
+    if time_info:
+        stats_text += "\n" + time_info
+
+    panel = Panel(
+        stats_text,
+        title="[bold]📊 Selection Summary[/bold]",
+        border_style="cyan",
+        padding=(1, 2)
+    )
+    console.print(panel)
+
+
+def display_session_preview(
+    session_messages: Dict[str, List[dict]],
+    max_sessions: int = 10
+) -> None:
+    """
+    显示session预览
+
+    Args:
+        session_messages: {session_id: [messages]}
+        max_sessions: 最多显示的session数量
+    """
+    try:
+        from rich.console import Console
+        from rich.table import Table
+    except ImportError:
+        # Fallback to plain text
+        print(f"\n📝 Sessions to be shared ({len(session_messages)} total):")
+        for i, (session_id, messages) in enumerate(list(session_messages.items())[:max_sessions], 1):
+            print(f"  {i}. {session_id[-25:] if len(session_id) > 25 else session_id} ({len(messages)} messages)")
+        if len(session_messages) > max_sessions:
+            print(f"  ... and {len(session_messages) - max_sessions} more sessions")
+        return
+
+    console = Console()
+
+    total_sessions = len(session_messages)
+    showing = min(total_sessions, max_sessions)
+
+    table = Table(
+        title=f"📝 Session Preview (showing {showing} of {total_sessions})"
+    )
+    table.add_column("#", style="dim", width=4)
+    table.add_column("Session ID", style="cyan", width=25)
+    table.add_column("Msgs", justify="right", width=6)
+    table.add_column("First User Message", style="dim", no_wrap=False)
+
+    for i, (session_id, messages) in enumerate(
+        list(session_messages.items())[:max_sessions], 1
+    ):
+        # 获取第一条用户消息
+        first_msg_preview = "[No user messages]"
+        for msg in messages[:10]:
+            if msg.get('role') == 'user':
+                content = msg.get('content', '')
+                if isinstance(content, str) and content.strip():
+                    first_msg_preview = content[:60]
+                    if len(content) > 60:
+                        first_msg_preview += "..."
+                    break
+
+        # 截断session ID用于显示
+        session_display = session_id[-25:] if len(session_id) > 25 else session_id
+
+        table.add_row(
+            str(i),
+            session_display,
+            str(len(messages)),
+            first_msg_preview
+        )
+
+    if total_sessions > max_sessions:
+        table.add_row(
+            "...",
+            f"[{total_sessions - max_sessions} more sessions]",
+            "...",
+            "..."
+        )
+
+    console.print(table)
+
+
+def save_preview_json(
+    conversation_data: dict,
+    username: str
+) -> Path:
+    """
+    保存预览JSON到临时位置
+
+    Args:
+        conversation_data: 完整的对话数据
+        username: 用户名(用于文件名)
+
+    Returns:
+        Path: 保存的预览文件路径
+    """
+    import tempfile
+
+    # 创建临时目录
+    temp_dir = Path(tempfile.gettempdir()) / "aline_previews"
+    temp_dir.mkdir(exist_ok=True, parents=True)
+
+    # 生成带时间戳的文件名
+    timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
+    preview_path = temp_dir / f"{username}_preview_{timestamp}.json"
+
+    # 保存并格式化
+    with open(preview_path, 'w', encoding='utf-8') as f:
+        json.dump(conversation_data, f, indent=2, ensure_ascii=False)
+
+    return preview_path
+
+
+def display_ui_metadata_preview(ui_metadata: dict) -> None:
+    """
+    显示UI metadata预览
+
+    Args:
+        ui_metadata: UI metadata字典
+    """
+    try:
+        from rich.console import Console
+        from rich.panel import Panel
+    except ImportError:
+        # Fallback to plain text
+        print("\n🎨 Generated UI Content:")
+        print(f"  Title: {ui_metadata.get('title', '[Not generated]')}")
+        print(f"  Welcome: {ui_metadata.get('welcome', '[Not generated]')}")
+        print(f"  Description: {ui_metadata.get('description', '[Not generated]')}")
+        questions = ui_metadata.get('preset_questions', [])
+        if questions:
+            print("  Preset Questions:")
+            for i, q in enumerate(questions, 1):
+                print(f"    {i}. {q}")
+        return
+
+    console = Console()
+
+    console.print("\n[bold cyan]🎨 Generated UI Content[/bold cyan]\n")
+
+    # Title
+    console.print(Panel(
+        ui_metadata.get('title', '[Not generated]'),
+        title="[bold]Title[/bold]",
+        border_style="green",
+        padding=(1, 2)
+    ))
+
+    # Welcome message
+    console.print(Panel(
+        ui_metadata.get('welcome', '[Not generated]'),
+        title="[bold]Welcome Message[/bold]",
+        border_style="blue",
+        padding=(1, 2)
+    ))
+
+    # Description
+    console.print(Panel(
+        ui_metadata.get('description', '[Not generated]'),
+        title="[bold]Description[/bold]",
+        border_style="yellow",
+        padding=(1, 2)
+    ))
+
+    # Preset questions
+    questions = ui_metadata.get('preset_questions', [])
+    if questions:
+        questions_text = "\n".join(f"{i+1}. {q}" for i, q in enumerate(questions))
+    else:
+        questions_text = "[No questions generated]"
+
+    console.print(Panel(
+        questions_text,
+        title="[bold]Preset Questions[/bold]",
+        border_style="magenta",
+        padding=(1, 2)
+    ))
+
+
+def prompt_ui_metadata_editing(ui_metadata: dict) -> dict:
+    """
+    提示用户确认或编辑UI metadata
+
+    Args:
+        ui_metadata: 生成的metadata
+
+    Returns:
+        dict: 更新后的metadata
+    """
+    try:
+        from rich.console import Console
+        from rich.prompt import Prompt, Confirm
+    except ImportError:
+        # Fallback: just return original if rich is not available
+        print("\n💡 Rich library not available for interactive editing. Using generated content.")
+        return ui_metadata
+
+    console = Console()
+
+    # 询问是否要编辑
+    console.print()
+    if not Confirm.ask(
+        "[yellow]Would you like to review and edit this content?[/yellow]",
+        default=False
+    ):
+        console.print("[green]✓ Using generated content as-is[/green]")
+        return ui_metadata
+
+    console.print("\n[dim]Press Enter to keep current value, or type new value:[/dim]\n")
+
+    edited = {}
+
+    # Title
+    console.print(f"[bold cyan]Title:[/bold cyan]")
+    console.print(f"  Current: {ui_metadata.get('title', '')}")
+    new_title = Prompt.ask("  New value", default="")
+    edited['title'] = new_title if new_title else ui_metadata.get('title', '')
+
+    # Welcome
+    console.print(f"\n[bold cyan]Welcome Message:[/bold cyan]")
+    console.print(f"  Current: {ui_metadata.get('welcome', '')}")
+    new_welcome = Prompt.ask("  New value", default="")
+    edited['welcome'] = new_welcome if new_welcome else ui_metadata.get('welcome', '')
+
+    # Description
+    console.print(f"\n[bold cyan]Description:[/bold cyan]")
+    console.print(f"  Current: {ui_metadata.get('description', '')}")
+    new_desc = Prompt.ask("  New value", default="")
+    edited['description'] = new_desc if new_desc else ui_metadata.get('description', '')
+
+    # Questions
+    questions = ui_metadata.get('preset_questions', [])
+    edited['preset_questions'] = []
+
+    console.print(f"\n[bold cyan]Preset Questions:[/bold cyan]")
+    for i, q in enumerate(questions, 1):
+        console.print(f"  Question {i}: {q}")
+        new_q = Prompt.ask(f"  New value", default="")
+        edited['preset_questions'].append(new_q if new_q else q)
+
+    console.print("\n[green]✓ Content updated[/green]")
+    return edited
+
+
+def display_share_result(
+    share_url: str,
+    password: str,
+    expiry_days: int,
+    max_views: int,
+    admin_token: Optional[str] = None
+) -> None:
+    """
+    显示分享创建结果
+
+    Args:
+        share_url: 分享URL
+        password: 加密密码
+        expiry_days: 过期天数
+        max_views: 最大浏览次数
+        admin_token: 管理员token(可选)
+    """
+    try:
+        from rich.console import Console
+        from rich.panel import Panel
+        from rich.text import Text
+    except ImportError:
+        # Fallback to plain text
+        print("\n╭────────────────────────────────────────╮")
+        print("│  ✅ Share Created Successfully!        │")
+        print("╰────────────────────────────────────────╯\n")
+        print(f"🔗 Share URL: {share_url}")
+        print(f"🔑 Password: {password}")
+        print(f"📅 Expires: {expiry_days} days")
+        print(f"👁️  Max views: {max_views}")
+        if admin_token:
+            print(f"\n📊 Admin token: {admin_token}")
+            print(f"   View stats at: {share_url}/stats?token={admin_token}")
+        print("\n💡 Share this URL and password with your team!")
+        return
+
+    console = Console()
+
+    # 成功标题
+    console.print("\n")
+    success_text = Text("✅ Share Created Successfully!", style="bold green")
+    success_panel = Panel(
+        success_text,
+        border_style="green",
+        padding=(1, 4)
+    )
+    console.print(success_panel)
+
+    # 分享详情
+    share_info = f"""[bold cyan]🔗 Share URL:[/bold cyan]
+{share_url}
+
+[bold yellow]🔑 Password:[/bold yellow]
+{password}
+
+[bold blue]📅 Expires:[/bold blue] {expiry_days} days from now
+[bold magenta]👁️  Max Views:[/bold magenta] {max_views} views"""
+
+    info_panel = Panel(
+        share_info,
+        title="[bold white]Share Details[/bold white]",
+        border_style="cyan",
+        padding=(1, 2)
+    )
+    console.print(info_panel)
+
+    # Admin token (如果有)
+    if admin_token:
+        admin_info = f"""[bold]Admin Token:[/bold]
+{admin_token}
+
+[bold]Stats URL:[/bold]
+{share_url}/stats?token={admin_token}
+
+[dim]Use this to view access statistics and manage the share[/dim]"""
+
+        admin_panel = Panel(
+            admin_info,
+            title="[bold yellow]Admin Access[/bold yellow]",
+            border_style="yellow",
+            padding=(1, 2)
+        )
+        console.print(admin_panel)
+
+    # 使用说明
+    instructions = """[bold green]How to Share:[/bold green]
+
+1. Copy the URL and password above
+2. Send them to your team (separately for security)
+3. Recipients can access the interactive chatbot
+4. They'll need the password to decrypt
+
+[dim]💡 Tip: Send URL via Slack, password via DM[/dim]"""
+
+    instructions_panel = Panel(
+        instructions,
+        title="[bold white]Next Steps[/bold white]",
+        border_style="green",
+        padding=(1, 2)
+    )
+    console.print(instructions_panel)
+    console.print()
+
+
+def export_shares_interactive_command(
+    indices: Optional[str] = None,
+    password: Optional[str] = None,
+    expiry_days: int = 7,
+    max_views: int = 100,
+    enable_preview: bool = True,
+    backend_url: Optional[str] = None,
+    repo_root: Optional[Path] = None,
+    preset: Optional[str] = None,
+    enable_mcp: bool = True
+) -> int:
+    """
+    交互式导出对话历史并生成分享链接
+
+    Args:
+        indices: Commit indices to export
+        password: 加密密码 (如果为 None 则自动生成)
+        expiry_days: 过期天数
+        max_views: 最大访问次数
+        enable_preview: 是否启用UI预览和编辑 (默认: True)
+        backend_url: 后端服务器 URL
+        repo_root: 项目根目录
+        preset: Prompt preset ID (如果为 None 则交互式选择)
+        enable_mcp: 是否启用MCP agent-to-agent通信 (默认: True)
+
+    Returns:
+        0 on success, 1 on error
+    """
+    logger.info("======== Interactive export shares command started ========")
+
+    # Check dependencies
+    if not CRYPTO_AVAILABLE:
+        print("❌ Error: cryptography package not installed", file=sys.stderr)
+        print("Install it with: pip install cryptography", file=sys.stderr)
+        return 1
+
+    if not HTTPX_AVAILABLE:
+        print("❌ Error: httpx package not installed", file=sys.stderr)
+        print("Install it with: pip install httpx", file=sys.stderr)
+        return 1
+
+    # Get backend URL
+    if backend_url is None:
+        # Try to load from config
+        from ..config import ReAlignConfig
+        if repo_root is None:
+            try:
+                result = subprocess.run(
+                    ["git", "rev-parse", "--show-toplevel"],
+                    capture_output=True,
+                    text=True,
+                    check=True
+                )
+                repo_root = Path(result.stdout.strip())
+            except subprocess.CalledProcessError:
+                repo_root = Path.cwd()
+
+        config = ReAlignConfig.load()
+        backend_url = config.share_backend_url
+
+    print("\n╭────────────────────────────────────────╮")
+    print("│  ReAlign Interactive Share Export      │")
+    print("╰────────────────────────────────────────╯\n")
+
+    # Step 1: Select commits (reuse existing logic)
+    if repo_root is None:
+        try:
+            result = subprocess.run(
+                ["git", "rev-parse", "--show-toplevel"],
+                capture_output=True,
+                text=True,
+                check=True
+            )
+            repo_root = Path(result.stdout.strip())
+        except subprocess.CalledProcessError:
+            print("Error: Not in a git repository", file=sys.stderr)
+            return 1
+
+    from .. import get_realign_dir
+    shadow_dir = get_realign_dir(repo_root)
+    shadow_git = shadow_dir
+
+    if not shadow_git.exists() or not (shadow_git / '.git').exists():
+        print(f"Error: ReAlign not initialized. Run 'aline init' first.", file=sys.stderr)
+        return 1
+
+    # Get commits
+    try:
+        all_commits = get_unpushed_commits(shadow_git)
+    except Exception as e:
+        print(f"Error: Failed to get commits: {e}", file=sys.stderr)
+        return 1
+
+    if not all_commits:
+        print("No unpushed commits found. Nothing to export.", file=sys.stderr)
+        return 1
+
+    # Display and select commits
+    if indices is None:
+        display_commits_for_selection(all_commits)
+        print("Enter commit indices to export (e.g., '1,3,5-7' or 'all'):")
+        indices = input("Indices: ").strip()
+
+        if not indices:
+            print("No commits selected. Exiting.")
+            return 0
+
+    # Parse indices
+    try:
+        if indices.lower() == "all":
+            indices_list = [c.index for c in all_commits]
+        else:
+            indices_list = parse_commit_indices(indices)
+    except ValueError as e:
+        print(f"Error: Invalid indices format: {e}", file=sys.stderr)
+        return 1
+
+    # Get selected commits
+    selected_commits = [c for c in all_commits if c.index in indices_list]
+
+    # Step 2: Extract messages
+    print(f"\n🔄 Extracting chat history from {len(selected_commits)} commit(s)...")
+    try:
+        session_messages = merge_messages_from_commits(selected_commits, shadow_git)
+    except Exception as e:
+        print(f"\nError: Failed to extract messages: {e}", file=sys.stderr)
+        return 1
+
+    if not session_messages:
+        print("\nWarning: No chat history found in selected commits.", file=sys.stderr)
+        return 1
+
+    # Build conversation data
+    username = os.environ.get('USER') or os.environ.get('USERNAME') or 'anonymous'
+    conversation_data = {
+        "username": username,
+        "time": datetime.now().isoformat(),
+        "sessions": [
+            {
+                "session_id": session_id,
+                "messages": messages
+            }
+            for session_id, messages in session_messages.items()
+        ]
+    }
+
+    total_messages = sum(len(msgs) for msgs in session_messages.values())
+    print(f"✅ Extracted {len(session_messages)} session(s) with {total_messages} messages")
+
+    # NEW: Display selection statistics
+    print()
+    display_selection_statistics(selected_commits, session_messages)
+
+    # NEW: Display session preview
+    print()
+    display_session_preview(session_messages)
+
+    # NEW: Save preview JSON
+    try:
+        preview_path = save_preview_json(conversation_data, username)
+        try:
+            from rich.console import Console
+            console = Console()
+            console.print(f"\n💾 Preview saved: [cyan]{preview_path}[/cyan]")
+            console.print("   You can inspect the full data before encryption.\n")
+        except ImportError:
+            print(f"\n💾 Preview saved: {preview_path}")
+            print("   You can inspect the full data before encryption.\n")
+    except Exception as e:
+        logger.warning(f"Failed to save preview JSON: {e}")
+        # Don't fail the export if preview save fails
+
+    # Step 3: Select prompt preset
+    from ..prompts import get_all_presets, get_preset_by_id, display_preset_menu, prompt_for_custom_instructions
+
+    selected_preset = None
+    custom_instructions = ""
+
+    if preset:
+        # Use preset specified via command line
+        selected_preset = get_preset_by_id(preset)
+        if not selected_preset:
+            print(f"\n❌ Error: Preset '{preset}' not found", file=sys.stderr)
+            print("Available presets:", file=sys.stderr)
+            all_presets = get_all_presets()
+            for p in all_presets:
+                print(f"  - {p.id}: {p.name}", file=sys.stderr)
+            return 1
+        print(f"\n✓ Using preset: {selected_preset.name} ({selected_preset.id})")
+    else:
+        # Interactive preset selection
+        all_presets = get_all_presets()
+        print(display_preset_menu(all_presets))
+
+        while True:
+            try:
+                selection = input("Enter preset number or ID [1]: ").strip()
+                if not selection:
+                    selection = "1"
+
+                # Try parsing as index first
+                try:
+                    idx = int(selection)
+                    from ..prompts import get_preset_by_index
+                    selected_preset = get_preset_by_index(idx)
+                    if not selected_preset:
+                        print(f"Invalid index. Please enter 1-{len(all_presets)}")
+                        continue
+                except ValueError:
+                    # Not a number, try as ID
+                    selected_preset = get_preset_by_id(selection)
+                    if not selected_preset:
+                        print(f"Invalid preset ID: {selection}")
+                        continue
+
+                print(f"\n✓ Selected: {selected_preset.name}")
+                break
+
+            except KeyboardInterrupt:
+                print("\n\nExport cancelled.")
+                return 0
+
+    # Collect custom instructions if allowed
+    if selected_preset.allow_custom_instructions:
+        custom_instructions = prompt_for_custom_instructions(selected_preset)
+
+    # Step 4: Generate UI metadata with LLM
+    print("\n🤖 Generating personalized UI content...")
+    from ..config import ReAlignConfig
+    config = ReAlignConfig.load()
+    ui_metadata, llm_debug_info = generate_ui_metadata_with_llm(
+        conversation_data,
+        selected_commits,
+        provider=config.llm_provider,
+        preset_id=selected_preset.id
+    )
+
+    # NEW: Display and optionally edit UI metadata
+    if ui_metadata:
+        display_ui_metadata_preview(ui_metadata)
+
+        # Only prompt for editing if enable_preview is True
+        if enable_preview:
+            try:
+                ui_metadata = prompt_ui_metadata_editing(ui_metadata)
+
+                # Show final version
+                try:
+                    from rich.console import Console
+                    console = Console()
+                    console.print("\n[bold green]✅ Final UI Content:[/bold green]")
+                except ImportError:
+                    print("\n✅ Final UI Content:")
+                display_ui_metadata_preview(ui_metadata)
+            except KeyboardInterrupt:
+                # User cancelled during editing
+                try:
+                    from rich.prompt import Confirm
+                    print("\n")
+                    if Confirm.ask("[yellow]Cancel export?[/yellow]", default=False):
+                        print("Export cancelled.")
+                        return 0
+                    else:
+                        print("Continuing with generated content...")
+                except ImportError:
+                    print("\nContinuing with generated content...")
+
+        # Add UI metadata with preset information
+        if ui_metadata is None:
+            ui_metadata = {}
+        ui_metadata["prompt_preset"] = {
+            "id": selected_preset.id,
+            "name": selected_preset.name,
+            "custom_instructions": custom_instructions
+        }
+        conversation_data["ui_metadata"] = ui_metadata
+
+        # Debug logging
+        logger.info(f"Added preset to ui_metadata: id={selected_preset.id}, custom_instructions={custom_instructions[:100] if custom_instructions else 'None'}")
+        print(f"\n🔍 Debug: Preset info added - ID: {selected_preset.id}, Custom instructions: '{custom_instructions[:50]}{'...' if len(custom_instructions) > 50 else ''}'", file=sys.stderr)
+    else:
+        # Use default UI metadata if LLM generation failed
+        try:
+            from rich.console import Console
+            console = Console()
+            console.print("[yellow]⚠️  Using default UI content[/yellow]")
+        except ImportError:
+            print("⚠️  Using default UI content")
+        # Still add preset information
+        conversation_data["ui_metadata"] = {
+            "prompt_preset": {
+                "id": selected_preset.id,
+                "name": selected_preset.name,
+                "custom_instructions": custom_instructions
+            }
+        }
+
+    # Add MCP instructions if enabled
+    if enable_mcp:
+        conversation_data["ui_metadata"]["mcp_instructions"] = {
+            "tool_name": "ask_shared_conversation",
+            "usage": "Local AI agents can install the aline MCP server and use the 'ask_shared_conversation' tool to query this conversation programmatically.",
+            "installation": {
+                "step1": "Install aline: pip install aline",
+                "step2": "Add to claude_desktop_config.json:",
+                "config": {
+                    "mcpServers": {
+                        "aline": {
+                            "command": "aline-mcp"
+                        }
+                    }
+                },
+                "step3": "Restart Claude Desktop"
+            },
+            "example_usage": "Ask your local Claude agent: 'Use the ask_shared_conversation tool to query this URL with question: ...'"
+        }
+        logger.info("MCP instructions added to ui_metadata")
+
+    # NEW: Save export log
+    try:
+        log_file = save_export_log(
+            conversation_data=conversation_data,
+            llm_prompts={
+                "system_prompt": llm_debug_info["system_prompt"],
+                "user_prompt": llm_debug_info["user_prompt"]
+            } if llm_debug_info else None,
+            llm_response=llm_debug_info["response_text"] if llm_debug_info else None,
+            username=username,
+            shadow_dir=shadow_git
+        )
+        logger.info(f"Export log saved to: {log_file}")
+        try:
+            from rich.console import Console
+            console = Console()
+            console.print(f"\n📝 Export log saved: [cyan]{log_file}[/cyan]\n")
+        except ImportError:
+            print(f"\n📝 Export log saved: {log_file}\n")
+    except Exception as e:
+        logger.warning(f"Failed to save export log: {e}")
+        # Don't fail the export if log save fails
+
+    # Step 4: Ask if user wants password protection
+    use_password = True  # Default
+    if password is None:
+        try:
+            from rich.prompt import Confirm
+            print()
+            use_password = Confirm.ask(
+                "[yellow]🔐 Would you like to protect this share with a password?[/yellow]",
+                default=False
+            )
+        except ImportError:
+            # Fallback: ask with plain input
+            print("\n🔐 Would you like to protect this share with a password? (y/N): ", end='')
+            response = input().strip().lower()
+            use_password = (response == 'y' or response == 'yes')
+
+    # Step 5: Generate password or skip encryption
+    encrypted_payload = None
+    if use_password:
+        if password is None:
+            # Generate a random password
+            password = secrets.token_urlsafe(16)
+            print(f"\n🔐 Generated password: {password}")
+            print("⚠️  Save this password - you'll need it to access the share!")
+
+        # Encrypt data
+        print("\n🔒 Encrypting conversation data...")
+        try:
+            encrypted_payload = encrypt_conversation_data(conversation_data, password)
+        except Exception as e:
+            print(f"\nError: Failed to encrypt data: {e}", file=sys.stderr)
+            logger.error(f"Encryption failed: {e}", exc_info=True)
+            return 1
+
+        print("✅ Encryption complete")
+    else:
+        print("\n📂 No password protection - data will be accessible to anyone with the link")
+        password = None
+
+    # Step 6: Upload to backend
+    print(f"\n📤 Uploading to {backend_url}...")
+
+    metadata = {
+        "username": username,
+        "expiry_days": expiry_days,
+        "max_views": max_views
+    }
+
+    try:
+        # Prepare payload based on whether encryption was used
+        if encrypted_payload:
+            payload = {"encrypted_payload": encrypted_payload, "metadata": metadata}
+        else:
+            # No encryption - send conversation data directly
+            payload = {"conversation_data": conversation_data, "metadata": metadata}
+
+        response = httpx.post(
+            f"{backend_url}/api/share/create",
+            json=payload,
+            timeout=30.0
+        )
+        response.raise_for_status()
+        result = response.json()
+    except Exception as e:
+        print(f"\n❌ Upload failed: {e}", file=sys.stderr)
+        logger.error(f"Upload failed: {e}", exc_info=True)
+        return 1
+
+    # Step 7: Display success with beautiful formatting
+    display_share_result(
+        share_url=result['share_url'],
+        password=password,
+        expiry_days=expiry_days,
+        max_views=max_views,
+        admin_token=result.get('admin_token')
+    )
+
+    # Display MCP setup instructions if enabled
+    if enable_mcp:
+        try:
+            from rich.console import Console
+            from rich.panel import Panel
+            console = Console()
+
+            mcp_instructions = f"""[bold green]🤖 MCP Access Enabled![/bold green]
+
+This share can be queried by AI agents using the Model Context Protocol.
+
+[bold]Installation:[/bold]
+
+1. Install aline (if not already installed):
+   [cyan]pip install aline[/cyan]
+
+2. Add to Claude Desktop config:
+   [dim]~/.config/Claude/claude_desktop_config.json (Linux/Mac)
+   or %APPDATA%/Claude/claude_desktop_config.json (Windows)[/dim]
+
+   {{
+     "mcpServers": {{
+       "aline": {{
+         "command": "aline-mcp"
+       }}
+     }}
+   }}
+
+3. Restart Claude Desktop
+
+[bold]Usage Example:[/bold]
+
+In Claude Desktop, say:
+[cyan]"Use the ask_shared_conversation tool to query this URL:
+{result['share_url']}
+
+Question: What were the main topics discussed?"[/cyan]
+
+{f'[yellow]Password: {password}[/yellow]' if password else '[dim](No password required)[/dim]'}
+
+[dim]💡 Tip: Agents can now directly query this conversation without human intervention![/dim]"""
+
+            mcp_panel = Panel(
+                mcp_instructions,
+                title="[bold magenta]Agent-to-Agent Communication[/bold magenta]",
+                border_style="magenta",
+                padding=(1, 2)
+            )
+            console.print()
+            console.print(mcp_panel)
+            console.print()
+
+        except ImportError:
+            # Fallback to plain text
+            print("\n" + "="*60)
+            print("🤖 MCP ACCESS ENABLED")
+            print("="*60)
+            print("\nThis share can be queried by AI agents programmatically.")
+            print("\nSetup:")
+            print("1. pip install aline")
+            print("2. Add to claude_desktop_config.json:")
+            print('   {"mcpServers": {"aline": {"command": "aline-mcp"}}}')
+            print("3. Restart Claude Desktop")
+            print(f"\nShare URL: {result['share_url']}")
+            if password:
+                print(f"Password: {password}")
+            print("\n" + "="*60 + "\n")
+
+    logger.info(f"======== Interactive export completed: {result['share_url']} ========")
+    return 0