aline-ai 0.6.2__py3-none-any.whl → 0.6.4__py3-none-any.whl

realign/commands/export_shares.py

@@ -24,7 +24,7 @@ from typing import Any, List, Dict, Optional, Tuple, Set, Callable
 
 from ..logging_config import setup_logger
 from ..db.base import SessionRecord, TurnRecord
-from ..llm_client import call_llm, extract_json
+from ..llm_client import extract_json, call_llm_cloud
 from ..auth import get_auth_headers, is_logged_in
 
 logger = setup_logger("realign.commands.export_shares", "export_shares.log")
@@ -2160,10 +2160,10 @@ def generate_ui_metadata_with_llm(
             user_requests.append(cleaned_request)
 
     # 如果没有 commit summary，回退到提取消息样本
+    user_messages = []
+    assistant_messages = []
     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
@@ -2176,238 +2176,166 @@ def generate_ui_metadata_with_llm(
                     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",
-            ],
-        },
-    }
-
-    # Check for user-customized prompt first
-    custom_prompt = _get_share_ui_metadata_prompt_template()
-    if custom_prompt:
-        # Use custom prompt directly
-        system_prompt = custom_prompt
-        logger.info("Using user-customized share UI metadata prompt")
-    else:
-        # 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 (with title and description provided) and generate
-personalized content for sharing and exploring that conversation.
-
-The conversation title and description are already provided. You only need to generate:
-1. Four preset questions
-2. A Slack share message
-
-Return the result strictly in JSON format:
-
-{{
-  "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)"
-  ],
-
-  "slack_message": "A friendly, casual Slack message sharing your work update with the team.
-
-                    Style Guidelines:
-                    - Tone: Casual, friendly, conversational (like chatting with teammates)
-                    - Length: 3-8 sentences depending on conversation complexity
-                    - Emoji: Use 1-3 relevant emojis to add personality (don't overdo it)
-                    - Technical depth: Mention key technologies but avoid unnecessary jargon
-                    - Clarity: Team members should understand the impact without deep technical knowledge
-
-                    Content Structure:
-                    1. Start with context - Give an overview of what you've been working on
-                    2. Highlight accomplishments - What did you achieve? (use casual language)
-                    3. Mention key details - Important technical decisions or challenges overcome
-                    4. End with status/next steps - What's the current state or what's coming next?
-
-                    Example Good Messages:
-                    - '🎉 Just wrapped up the session summary feature! Built a new command that automatically generates event summaries from recent commits - both short emotional messages and detailed reports. Had to fix a bunch of commit parsing quirks along the way, but it\\'s solid now. Next up: testing the UI expand/collapse to make sure everything feels smooth.'
-                    - '🚀 Big update, team! Been cranking on several fronts: First, got the vertical split terminal feature live - you can now split, resize, and close terminal panes as needed. Also revamped the onboarding flow and share dialog to be more compact and user-friendly. Fixed a bunch of UI quirks along the way. Next up: making sure multi-event sharing works seamlessly! 💪'
-
-                    Bad Examples to Avoid:
-                    - Too formal: 'I have successfully completed the implementation...'
-                    - Too technical: 'Implemented event summarization via LLM-driven clustering...'"
-}}
-
-Requirements:
-1. Preset questions must be based on the actual conversation content, concrete, and useful from the specified angles.
-2. The Slack message should be casual, friendly, and share your work progress like you're updating teammates.
-3. All text must be in English or Chinese, depending on the conversation language.
-4. 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 and generate content (preset_questions, slack_message).
-
-Event Title (already provided, do not regenerate):
-"{event_title}"
-
-Event Description (already provided, do not regenerate):
-"{event_description}"
-
-Conversation Details:
-- 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"}
-
-Based on the event title, description, and conversation content above, generate:
-1. Four preset questions users might ask
-2. A Slack share message
-
-Return the content in JSON format."""
-    else:
-        # 回退到使用消息样本
-        user_prompt = f"""Analyze the following conversation and generate content (preset_questions, slack_message).
-
-Event Title (already provided, do not regenerate):
-"{event_title}"
-
-Event Description (already provided, do not regenerate):
-"{event_description}"
-
-Conversation Details:
-- Number of sessions: {len(sessions)}
-- Total messages: {total_messages}
+    # Try cloud provider first if logged in
+    if provider in ("auto", "cloud") and is_logged_in():
+        logger.debug("Attempting cloud LLM for UI metadata generation")
+        # Load user custom prompt if available
+        custom_prompt = _get_share_ui_metadata_prompt_template()
 
-User message samples:
-{chr(10).join(user_messages[:10])}
-
-Assistant reply samples:
-{chr(10).join(assistant_messages[:10])}
-
-Based on the event title, description, and conversation content above, generate:
-1. Four preset questions users might ask
-2. A Slack share message
+        # Build payload
+        cloud_payload = {
+            "event_title": event_title,
+            "event_description": event_description,
+            "sessions_count": len(sessions),
+            "total_messages": total_messages,
+        }
 
-Return the content in JSON format."""
+        if commit_summaries:
+            cloud_payload["commit_summaries"] = commit_summaries
+            cloud_payload["user_requests"] = user_requests
+        else:
+            cloud_payload["user_messages"] = user_messages
+            cloud_payload["assistant_messages"] = assistant_messages
 
-    # Use unified LLM client
-    try:
-        model_name, response_text = call_llm(
-            system_prompt=system_prompt,
-            user_prompt=user_prompt,
-            provider=provider,
-            max_tokens=1000,
-            temperature=0.7,  # Higher temperature for creative UI text
-            purpose="ui_metadata",
+        model_name, result = call_llm_cloud(
+            task="ui_metadata",
+            payload=cloud_payload,
+            custom_prompt=custom_prompt,
+            preset_id=preset_id,
             silent=silent,
         )
 
-        if not response_text:
+        if result:
+            # Build ui_metadata from result
+            ui_metadata = {
+                "title": event_title,
+                "description": event_description,
+                "welcome": "",  # No longer generated by LLM
+                "preset_questions": result.get("preset_questions", []),
+                "slack_message": result.get("slack_message", ""),
+            }
+            debug_info = {
+                "system_prompt": "(cloud)",
+                "user_prompt": "(cloud)",
+                "response_text": str(result),
+                "provider": model_name or "cloud",
+            }
+            logger.info(f"Cloud LLM UI metadata generation success ({model_name})")
+            return ui_metadata, debug_info
+        else:
+            # Cloud LLM failed, return None (local fallback disabled)
+            logger.warning("Cloud LLM UI metadata failed")
             if not silent:
-                logger.warning("LLM returned empty response for UI metadata generation")
-                print(
-                    "   ⚠️  LLM returned empty response, using default UI text",
-                    file=sys.stderr,
-                )
+                print("   ⚠️  Cloud LLM UI metadata failed", file=sys.stderr)
             return None, None
 
-        # Parse JSON response
-        ui_metadata = extract_json(response_text)
-
-        # Add event title, description, and empty welcome
-        ui_metadata["title"] = event_title
-        ui_metadata["description"] = event_description
-        ui_metadata["welcome"] = ""  # No longer generated by LLM
-
-        # Return with debug info
-        debug_info = {
-            "system_prompt": system_prompt,
-            "user_prompt": user_prompt,
-            "response_text": response_text,
-            "provider": model_name or provider,
-        }
-        return ui_metadata, debug_info
-
-    except Exception as e:
-        logger.error(f"LLM UI metadata generation failed: {e}", exc_info=True)
-        if not silent:
-            print(
-                f"   ⚠️  LLM generation failed: {e}, using default UI text",
-                file=sys.stderr,
-            )
-        return None, None
-
-        # Parse JSON response
-        ui_metadata = extract_json(response_text)
-
-        # Add event title, description, and empty welcome
-        ui_metadata["title"] = event_title
-        ui_metadata["description"] = event_description
-        ui_metadata["welcome"] = ""  # No longer generated by LLM
-
-        # Return with debug info
-        debug_info = {
-            "system_prompt": system_prompt,
-            "user_prompt": user_prompt,
-            "response_text": response_text,
-            "provider": model_name or provider,
-        }
-        return ui_metadata, debug_info
-
-    except Exception as e:
-        logger.error(f"LLM UI metadata generation failed: {e}", exc_info=True)
-        print(f"   ⚠️  LLM generation failed: {e}, using default UI text", file=sys.stderr)
-        return None, None
+    # User not logged in, return None (local fallback disabled)
+    logger.warning("Not logged in, cannot use cloud LLM for UI metadata")
+    if not silent:
+        print("   ⚠️  Please login with 'aline login' to use LLM features", file=sys.stderr)
+    return None, None
+
+    # =========================================================================
+    # LOCAL LLM FALLBACK DISABLED - Code kept for reference
+    # =========================================================================
+    # # 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",
+    #         "welcome_tone": "professional and confident, highlighting accomplishments and progress",
+    #         "description_focus": "what work was done, what value was created",
+    #         "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",
+    #         "title_style": "a thought-provoking, conceptual title",
+    #         "welcome_tone": "insightful and educational",
+    #         "description_focus": "the knowledge and insights shared",
+    #         "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",
+    #         "title_style": "an analytical, personality-focused title",
+    #         "welcome_tone": "analytical yet friendly",
+    #         "description_focus": "personality traits, working styles, and communication patterns",
+    #         "question_angles": [
+    #             "overall personality traits and characteristics",
+    #             "working style and approach to problem-solving",
+    #             "communication patterns and preferences",
+    #             "strengths, growth areas, and unique qualities",
+    #         ],
+    #     },
+    # }
+    #
+    # # Check for user-customized prompt first
+    # custom_prompt = _get_share_ui_metadata_prompt_template()
+    # if custom_prompt:
+    #     system_prompt = custom_prompt
+    #     logger.info("Using user-customized share UI metadata prompt")
+    # else:
+    #     preset_config = preset_configs.get(preset_id, preset_configs["default"])
+    #     system_prompt = f"""You are a conversation interface copy generator for {preset_config["role_description"]}..."""
+    #
+    # # 构建 user prompt - 优先使用 commit summaries
+    # if commit_summaries:
+    #     user_prompt = f"""Analyze the following conversation..."""
+    # else:
+    #     user_prompt = f"""Analyze the following conversation..."""
+    #
+    # # Use unified LLM client
+    # try:
+    #     model_name, response_text = call_llm(
+    #         system_prompt=system_prompt,
+    #         user_prompt=user_prompt,
+    #         provider=provider,
+    #         max_tokens=1000,
+    #         temperature=0.7,
+    #         purpose="ui_metadata",
+    #         silent=silent,
+    #     )
+    #
+    #     if not response_text:
+    #         return None, None
+    #
+    #     ui_metadata = extract_json(response_text)
+    #     ui_metadata["title"] = event_title
+    #     ui_metadata["description"] = event_description
+    #     ui_metadata["welcome"] = ""
+    #
+    #     debug_info = {
+    #         "system_prompt": system_prompt,
+    #         "user_prompt": user_prompt,
+    #         "response_text": response_text,
+    #         "provider": model_name or provider,
+    #     }
+    #     return ui_metadata, debug_info
+    #
+    # except Exception as e:
+    #     logger.error(f"LLM UI metadata generation failed: {e}", exc_info=True)
+    #     return None, None
 
 
 def display_selection_statistics(

realign/commands/init.py

@@ -1,6 +1,8 @@
 """ReAlign init command - Initialize ReAlign tracking system."""
 
-from typing import Dict, Any, Optional, Tuple
+import shutil
+import sys
+from typing import Annotated, Any, Dict, Optional, Tuple
 from pathlib import Path
 import re
 import typer
@@ -706,9 +708,21 @@ def init_global(
 
 
 def init_command(
-    force: bool = typer.Option(
-        False, "--force", "-f", help="Overwrite global config with defaults"
-    ),
+    force: bool = typer.Option(False, "--force", "-f", help="Overwrite global config with defaults"),
+    doctor: Annotated[
+        bool,
+        typer.Option(
+            "--doctor/--no-doctor",
+            help="Run 'aline doctor' after init (best for upgrades)",
+        ),
+    ] = False,
+    install_tmux: Annotated[
+        bool,
+        typer.Option(
+            "--install-tmux/--no-install-tmux",
+            help="Auto-install tmux via Homebrew if missing (macOS only)",
+        ),
+    ] = True,
     start_watcher: Optional[bool] = typer.Option(
         None,
         "--start-watcher/--no-start-watcher",
@@ -726,6 +740,54 @@ def init_command(
         force=force,
     )
 
+    # First-time UX: tmux is required for the default dashboard experience (tmux mode).
+    # Only attempt on macOS; on other platforms, leave it to user.
+    if (
+        result.get("success")
+        and install_tmux
+        and result.get("tmux_conf")
+        and sys.platform == "darwin"
+        and shutil.which("tmux") is None
+    ):
+        console.print("\n[bold]tmux not found. Installing via Homebrew...[/bold]")
+        try:
+            from . import add as add_cmd
+
+            rc = add_cmd.add_tmux_command(install_brew=True)
+            if rc != 0:
+                result["errors"] = (result.get("errors") or []) + [
+                    "tmux install failed (required for the default tmux dashboard)",
+                    "Tip: set ALINE_TERMINAL_MODE=native to run without tmux",
+                ]
+        except Exception as e:
+            result["errors"] = (result.get("errors") or []) + [
+                f"tmux install failed: {e}",
+                "Tip: set ALINE_TERMINAL_MODE=native to run without tmux",
+            ]
+
+    if doctor and result.get("success"):
+        # Run doctor in "safe" mode: restart only if already running, and keep init fast.
+        try:
+            from . import doctor as doctor_cmd
+
+            restart_daemons = start_watcher is not False
+            doctor_exit = doctor_cmd.run_doctor(
+                restart_daemons=restart_daemons,
+                start_if_not_running=False,
+                verbose=False,
+                clear_cache=False,
+            )
+            if doctor_exit != 0:
+                result["success"] = False
+                result["errors"] = (result.get("errors") or []) + [
+                    "aline doctor failed (see output above)"
+                ]
+                result["message"] = f"{result.get('message', '').strip()} (doctor failed)".strip()
+        except Exception as e:
+            result["success"] = False
+            result["errors"] = (result.get("errors") or []) + [f"aline doctor failed: {e}"]
+            result["message"] = f"{result.get('message', '').strip()} (doctor failed)".strip()
+
     watcher_started: Optional[bool] = None
     watcher_start_exit: Optional[int] = None
     worker_started: Optional[bool] = None