naas-abi 1.0.8__py3-none-any.whl → 1.0.9__py3-none-any.whl

naas_abi/apps/terminal_agent/main.py

@@ -1,553 +0,0 @@
-import re
-import sys
-import threading
-import time
-from datetime import datetime
-from pathlib import Path
-from typing import Any, Optional, Union
-
-from langchain_core.messages import ToolMessage
-from langgraph.types import Command
-from naas_abi.apps.terminal_agent.terminal_style import (
-    console,  # Add console import
-    print_image,
-    print_tool_response,
-    print_tool_usage,
-)
-from naas_abi_core import logger
-from naas_abi_core.services.agent.Agent import Agent
-
-# Global variable to track active agent for context-aware conversations
-conversation_file = None
-
-# Fixed width for consistent conversation logs (matches typical wide terminal)
-TERMINAL_WIDTH = 77  # Matches the separator length from the user's example
-
-
-def init_conversation_file():
-    """Initialize a new conversation file with timestamp"""
-    global conversation_file
-
-    # Create timestamp in format YYYYMMDDTHHMMSS
-    timestamp = datetime.now().strftime("%Y%m%dT%H%M%S")
-
-    # Create directory structure
-    conversation_dir = Path("storage/datastore/interfaces/terminal_agent")
-    conversation_dir.mkdir(parents=True, exist_ok=True)
-
-    # Create conversation file path
-    conversation_file = conversation_dir / f"{timestamp}.txt"
-
-    # Initialize file with header
-    with open(conversation_file, "w", encoding="utf-8") as f:
-        f.write(
-            f"# ABI Terminal Conversation - {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n"
-        )
-        f.write(f"# Session started at: {timestamp}\n")
-        f.write("=" * 80 + "\n\n")
-
-    logger.info(f"💾 Conversation logging to: {conversation_file}")
-    return conversation_file
-
-
-def save_to_conversation(line: str):
-    """Save exactly what appears in terminal to the conversation file"""
-    global conversation_file
-
-    if conversation_file is None:
-        return
-
-    try:
-        with open(conversation_file, "a", encoding="utf-8") as f:
-            f.write(line + "\n")
-    except Exception as e:
-        # Print error to terminal
-        print(f"⚠️ Error saving to conversation file: {e}")
-        # Try to log the error itself if possible
-        try:
-            with open(conversation_file, "a", encoding="utf-8") as f:
-                f.write(f"⚠️ LOGGING ERROR: {e}\n")
-        except Exception:
-            pass  # If we can't log the error, give up
-
-
-def get_input_with_placeholder(prompt="> ", placeholder="Send a message (/? for help)"):
-    """Get user input with a placeholder that disappears when typing starts
-
-    Supports pasting large text and multiline input.
-    Press Enter twice (empty line) to submit multiline input.
-    """
-
-    # Check if input is piped (not interactive terminal)
-    if not sys.stdin.isatty():
-        # For piped input, use simple input() without fancy terminal handling
-        print(f"\n{prompt}", end="", flush=True)
-        try:
-            return input()
-        except:  # noqa: E722
-            return "/bye"
-
-    # Show prompt with placeholder
-    print(f"\n{prompt}", end="", flush=True)
-
-    # Show placeholder in grey
-    print(f"\033[90m{placeholder}\033[0m", end="", flush=True)
-
-    # Move cursor back to start of placeholder
-    print(f"\033[{len(placeholder)}D", end="", flush=True)
-
-    lines = []
-    first_line = True
-
-    try:
-        while True:
-            try:
-                if first_line:
-                    # First line - use the placeholder display
-                    line = input()
-                    # Clear the placeholder line
-                    print(f"\033[A\033[2K{prompt}{line}", flush=True)
-                    first_line = False
-                else:
-                    # Continuation lines
-                    line = input("... ")
-
-                # If line is empty and we have previous content, submit
-                if not line.strip() and lines:
-                    break
-                # If line is not empty, add it
-                elif line.strip():
-                    lines.append(line)
-                # If first line is empty, return empty (single enter)
-                elif not lines:
-                    return ""
-
-            except EOFError:
-                # Ctrl+D pressed
-                break
-
-    except KeyboardInterrupt:
-        print()
-        raise
-
-    # Join all lines with newlines
-    result = "\n".join(lines)
-    return result
-
-
-def on_tool_response(message: Union[str, Command, dict[str, Any], ToolMessage]) -> None:
-    try:
-        message_content: str = ""
-        if isinstance(message, str):
-            message_content = message
-        elif isinstance(message, dict) and "content" in message:
-            message_content = str(message["content"])
-        elif isinstance(message, ToolMessage):
-            message_content = str(message.content)
-        else:
-            print("Unknown message type:")
-            print(type(message))
-            message_content = str(message)
-
-        print_tool_response(message_content)
-
-        # Check if the message contains a path to an image file
-        if isinstance(message_content, str):
-            # Look for image file paths in the message
-            words = message_content.split(" ")
-            for word in words:
-                if any(
-                    word.lower().endswith(ext)
-                    for ext in [".png", ".jpg", ".jpeg", ".gif"]
-                ):
-                    print_image(word)
-    except Exception as e:
-        error_msg = f"⚠️ Tool Response Error: {e}"
-        print(error_msg)
-        # Log the error to conversation file
-        save_to_conversation(error_msg)
-
-
-def on_ai_message(message: Any, agent_name: str) -> None:
-    if len(message.content) == 0:
-        return
-
-    print("\r" + " " * 15 + "\r", end="", flush=True)
-
-    from rich.markdown import Markdown
-
-    # Filter out think tags and their content
-    think_content = re.findall(r"<think>.*?</think>", message.content, flags=re.DOTALL)
-
-    # Prepare thoughts section for both display and logging
-    thoughts_for_log = ""
-    if len(think_content) > 0:
-        console.print("Thoughts:", style="grey66")
-        thoughts_for_log += "Thoughts:      \n\n"
-        for think in think_content:
-            think_text = think.replace("<think>", "").replace("</think>", "").strip()
-            console.print(think_text, style="grey66")
-            thoughts_for_log += think_text + "\n"
-        thoughts_for_log += "\n"  # Extra line after thoughts
-
-    content = re.sub(
-        r"<think>.*?</think>", "", message.content, flags=re.DOTALL
-    ).strip()
-
-    # Print agent name dynamically using the real intent_target
-
-    # Use the actual agent name from intent_target, with color coding for readability
-    if "abi" in agent_name.lower():
-        color = "bold green"
-    elif "claude" in agent_name.lower():
-        color = "bold dark_orange"  # Anthropic's orange, use rich color
-    elif "chatgpt" in agent_name.lower():
-        color = "bold bright_green"  # OpenAI's green
-    elif "deepseek" in agent_name.lower():
-        color = "bold blue"  # DeepSeek's blue
-    elif "gemini" in agent_name.lower():
-        color = "bold bright_blue"  # Google's blue
-    elif "gemma" in agent_name.lower():
-        color = "bold bright_cyan"  # Google's cyan/blue
-    elif "grok" in agent_name.lower():
-        color = "bold white"  # X/Twitter's white
-    elif "llama" in agent_name.lower():
-        color = "bold bright_blue"  # Meta's blue
-    elif "mistral" in agent_name.lower():
-        color = "bold dark_orange"  # Mistral's orange, use rich color
-    elif "perplexity" in agent_name.lower():
-        color = "bold white"  # Perplexity's white
-    elif "qwen" in agent_name.lower():
-        color = "bold bright_cyan"  # Alibaba's cyan
-    else:
-        color = "bold magenta"
-
-    # Format exactly as it appears in terminal
-    agent_message_line = f"{agent_name}: {content}"
-
-    # Display the real agent name
-    console.print(f"{agent_name}:", style=color, end=" ")
-
-    md = Markdown(content)
-    console.print(md, style="bright_white")
-    console.print("─" * console.width, style="dim")
-    print()  # Add spacing after separator
-
-    # Save exact terminal format to conversation file (with fixed width) including thoughts
-    if thoughts_for_log:
-        save_to_conversation(thoughts_for_log)
-    save_to_conversation(agent_message_line)
-    save_to_conversation("─" * TERMINAL_WIDTH)
-    save_to_conversation("")  # Empty line
-
-
-def run_agent(agent: Agent):
-    # Initialize conversation logging
-    init_conversation_file()
-
-    # All agents
-    all_agents = agent.agents + [agent]
-
-    # Show greeting when truly ready for input - instant like responses
-    greeting_line = f"{agent.name}: Hello, World!"
-
-    console.print(f"{agent.name}:", style="bold green", end=" ")
-    console.print("Hello, World!", style="bright_white")
-    console.print("─" * console.width, style="dim")
-    print()  # Add spacing after separator
-
-    # Save exact terminal output to conversation file (with fixed width)
-    save_to_conversation(greeting_line)
-    save_to_conversation("─" * TERMINAL_WIDTH)
-    save_to_conversation("")  # Empty line
-
-    # Just start chatting naturally - like the screenshot
-    while True:
-        # Get current active agent and its model
-        current_active_agent = agent.state.current_active_agent
-        if current_active_agent is None:
-            current_active_agent = agent.name
-
-        model_info = "unknown"
-
-        # Find the active agent in our agents list
-        import pydash as _
-
-        current_agent = _.find(
-            all_agents,
-            lambda a: a.name.lower() == current_active_agent.lower()
-            if a.name is not None
-            else False,
-        )
-        if current_agent:
-            if hasattr(current_agent.chat_model, "model_name"):
-                model_info = current_agent.chat_model.model_name
-            elif hasattr(current_agent.chat_model, "model"):
-                model_info = current_agent.chat_model.model
-
-        # Create clean status line showing active agent with model info
-        if current_active_agent:
-            status_line = f"Active: {current_active_agent} (model: {model_info})"
-        else:
-            status_line = "No active agent"
-
-        # Print the status line before the input prompt
-        console.print(status_line, style="dim")
-
-        # Save status line to conversation file
-        save_to_conversation(status_line)
-        save_to_conversation("")  # Empty line for spacing
-
-        user_input = get_input_with_placeholder()
-
-        # Clean the input and check for exit commands
-        clean_input = user_input.strip().lower()
-
-        # Skip empty input
-        if not user_input.strip():
-            continue
-
-        # Display user message with color coding and separator (except for commands)
-        if not clean_input.startswith("/") and clean_input not in [
-            "exit",
-            "quit",
-            "reset",
-        ]:
-            # Format exactly as it appears in terminal
-            user_message_line = f"You: {user_input.strip()}"
-
-            # Show formatted message in chat history (same format as Abi)
-            console.print("You:", style="bold cyan", end=" ")
-            console.print(user_input.strip(), style="bright_white")
-            console.print("─" * console.width, style="dim")
-            print()  # Add spacing after separator
-
-            # Save exact terminal format to conversation file (with fixed width)
-            save_to_conversation(user_message_line)
-            save_to_conversation("─" * TERMINAL_WIDTH)
-            save_to_conversation("")  # Empty line
-
-        if clean_input in ["exit", "/exit", "/bye", "quit", "/quit"]:
-            # Save session end to conversation file
-            save_to_conversation("")  # Empty line
-            save_to_conversation("# Session ended by user")
-            print(f"\n👋 See you later! Conversation saved to: {conversation_file}")
-            return
-        elif clean_input in ["reset", "/reset"]:
-            agent.reset()
-            print("🔄 Starting fresh...")
-            continue
-        elif clean_input == "/?":
-            print("\n📋 Available Commands:")
-            print("  /? - Show this help")
-            print("  /reset - Start fresh conversation")
-            print("  /bye or /exit - End conversation")
-            print("\n📝 Input Tips:")
-            print("  • Paste large text directly - it will be captured properly")
-            print(
-                "  • For multiline input: keep typing on new lines, press Enter twice to submit"
-            )
-            print(
-                "  • Single Enter on first line submits immediately (for quick messages)"
-            )
-            print("\n🤖 Available AI Agents:")
-            print("  Cloud Agents:")
-            cloud_agents = [
-                "@gemini",
-                "@claude",
-                "@mistral",
-                "@chatgpt",
-                "@perplexity",
-                "@llama",
-            ]
-            print(f"    {' '.join(cloud_agents)}")
-            print("  Local Agents (Privacy-focused):")
-            local_agents = ["@qwen", "@deepseek", "@gemma"]
-            print(f"    {' '.join(local_agents)}")
-            print("\n💡 Usage: Type @agent or 'ask agent' to switch agents")
-            print(
-                "   Example: '@qwen help me code' or 'ask deepseek solve this math problem'"
-            )
-            continue
-
-        # Matrix-style animated loading indicator
-        import threading
-        import time
-
-        # Animation control
-        loading = True
-
-        def matrix_loader():
-            i = 0
-            while loading:
-                dots_count = i % 4  # 0, 1, 2, 3, then repeat
-                if dots_count == 0:
-                    dots = "   "  # No dots, just spaces
-                else:
-                    dots = "." * dots_count + " " * (
-                        3 - dots_count
-                    )  # Pad to 3 char width
-                print(f"\r\033[92mResponding{dots}\033[0m", end="", flush=True)
-                time.sleep(0.5)
-                i += 1
-
-        # Start the animation in a separate thread
-        loader_thread = threading.Thread(target=matrix_loader)
-        loader_thread.start()
-
-        # # Update the agent's shared state with current active agent info
-        # if hasattr(agent, '_state') and hasattr(agent._state, 'set_current_active_agent'):
-        #     agent._state.set_current_active_agent(current_active_agent)
-
-        # Get the response with real streaming support
-        try:
-            # Stop the animation first
-            loading = False
-            loader_thread.join()
-            print("\r" + " " * 15 + "\r", end="", flush=True)
-
-            # Use the agent system properly
-            agent.invoke(user_input)
-
-        except Exception as e:
-            # Stop the animation if still running
-            loading = False
-            if "loader_thread" in locals():
-                loader_thread.join()
-
-            # Clear the loading line
-            print("\r" + " " * 15 + "\r", end="", flush=True)
-
-            # Display and log the error
-            error_msg = f"❌ Agent Error: {e}"
-            console.print(error_msg, style="bold red")
-            save_to_conversation(error_msg)
-
-            # Log the full traceback for debugging
-            import traceback
-
-            traceback_msg = f"Full traceback:\n{traceback.format_exc()}"
-            console.print(traceback_msg, style="dim red")
-            save_to_conversation(traceback_msg)
-            save_to_conversation("─" * TERMINAL_WIDTH)
-            save_to_conversation("")  # Empty line
-            continue  # Continue conversation instead of crashing
-
-
-def load_agent(agent_class: str) -> Agent | None:
-    from naas_abi import modules
-
-    if agent_class is None:
-        print(
-            "No agent class provided. Please set the AGENT_CLASS environment variable."
-        )
-        return
-
-    for module in modules:
-        for agent in module.agents:
-            if agent.__class__.__name__ == agent_class:
-                agent.on_tool_usage(lambda message: print_tool_usage(message))
-                agent.on_tool_response(on_tool_response)
-                agent.on_ai_message(on_ai_message)
-
-                return agent
-
-    return None
-
-
-def list_available_agents():
-    from naas_abi import modules
-
-    print("\nAvailable agents:\n")
-    agents = []
-    for module in modules:
-        for agent in module.agents:
-            agents.append(agent.__class__.__name__)
-
-    # Sort the agents alphabetically
-    agents.sort()
-
-    # Print the agents
-    for agent in agents:
-        print(f"  - {agent}")
-
-
-class ConsoleLoader:
-    def start(self, message: str):
-        # Matrix-style startup animation
-        self.loading = True
-
-        def startup_loader():
-            i = 0
-            while self.loading:
-                dots_count = i % 4  # 0, 1, 2, 3, then repeat
-                if dots_count == 0:
-                    dots = "   "  # No dots, just spaces
-                else:
-                    dots = "." * dots_count + " " * (
-                        3 - dots_count
-                    )  # Pad to 3 char width
-                print(f"\r\033[92m{message}{dots}\033[0m", end="", flush=True)
-                time.sleep(0.5)
-                i += 1
-
-        # Start the animation
-        self.loader_thread = threading.Thread(target=startup_loader)
-        self.loader_thread.daemon = True  # Make thread die when main thread exits
-        self.loader_thread.start()
-
-    def stop(self):
-        self.loading = False
-        self.loader_thread.join()
-
-        # Clear the loading line
-        print("\r" + " " * 20 + "\r", end="", flush=True)
-
-
-def generic_run_agent(agent_class: Optional[str] = None) -> None:
-    """Run an agent dynamically loaded from the src/modules directory.
-
-    This method provides a generic way to run any agent that is loaded from the modules
-    directory, eliminating the need to create individual run functions for each agent.
-    The agents are automatically discovered and loaded through the module system.
-
-    Args:
-        agent_class (str, optional): The class name of the agent to run. If None, will
-            print an error message. The class name should match exactly with the agent's
-            class name in the modules.
-
-    Example:
-        >>> generic_run_agent("AbiAgent")  # Runs the AbiAgent agent
-        >>> generic_run_agent("ContentAssistant")     # Runs the ContentAssistant agent
-
-    Note:
-        This replaces the need for individual run_*_agent() functions by dynamically
-        finding and running the requested agent from the loaded modules. The agent
-        must be properly registered in a module under src/modules for this to work.
-    """
-
-    assert agent_class is not None, "Agent class is required"
-
-    agent = load_agent(agent_class)
-
-    if agent is None:
-        print(f"Agent {agent_class} not found")
-        list_available_agents()
-
-        return
-
-    run_agent(agent)
-
-
-if __name__ == "__main__":
-    import sys
-
-    # Get the function name from command line argument
-    if len(sys.argv) > 1:
-        function_name = sys.argv[1]
-        if function_name in globals():
-            globals()[function_name](*sys.argv[2:])
-        else:
-            print(f"Function {function_name} not found")
-    else:
-        print("Please specify a function to run")