my_agent/chat_cli.py

from __future__ import annotations

import json
from pathlib import Path
import sys
import threading
import time
from typing import Any

if __package__ in (None, ""):
    sys.path.insert(0, str(Path(__file__).resolve().parent.parent))

from core_agent.agent import CoreAgent
from core_agent.config import apply_compat_env_aliases, build_core_agent_config, load_core_agent_env
from core_agent.providers.openai_compatible import OpenAICompatibleProvider


class TerminalRenderer:
    def __init__(self, heartbeat_seconds: float = 3.0) -> None:
        self.heartbeat_seconds = heartbeat_seconds
        self._last_activity = time.monotonic()
        self._stop = threading.Event()
        self._thread: threading.Thread | None = None
        self._printed_thinking = False
        self._printed_answer = False
        self._saw_text = False

    def start_turn(self) -> None:
        self._last_activity = time.monotonic()
        self._stop.clear()
        self._printed_thinking = False
        self._printed_answer = False
        self._saw_text = False
        print("Assistant: ", end="", flush=True)
        self._thread = threading.Thread(target=self._heartbeat_loop, daemon=True)
        self._thread.start()

    def stop_turn(self) -> None:
        self._stop.set()
        if self._thread is not None:
            self._thread.join(timeout=0.2)
            self._thread = None
        print()

    def on_reasoning(self, text: str) -> None:
        self._touch()
        if not text:
            return
        if not self._printed_thinking:
            print("\n[thinking] ", end="", flush=True)
            self._printed_thinking = True
        print(text, end="", flush=True)

    def on_content(self, text: str) -> None:
        self._touch()
        if not text:
            return
        if not self._printed_answer:
            prefix = "\n[answer] " if self._printed_thinking else ""
            print(prefix, end="", flush=True)
            self._printed_answer = True
        print(text, end="", flush=True)
        self._saw_text = True

    def on_tool_call(self, tool_name: str, tool_args: dict[str, Any]) -> None:
        self._touch()
        rendered_args = _compact_json(tool_args, limit=200)
        print(f"\n[tool] {tool_name} {rendered_args}", flush=True)

    def on_tool_result(self, tool_result: str) -> None:
        self._touch()
        print(f"[tool-result] {_summarize_tool_result(tool_result)}", flush=True)

    def on_final(self, final_response: str) -> None:
        self._touch()
        if final_response and not self._saw_text:
            if not self._printed_answer:
                prefix = "\n[answer] " if self._printed_thinking else ""
                print(prefix, end="", flush=True)
                self._printed_answer = True
            print(final_response, end="", flush=True)

    def _touch(self) -> None:
        self._last_activity = time.monotonic()

    def _heartbeat_loop(self) -> None:
        while not self._stop.wait(0.25):
            idle_for = time.monotonic() - self._last_activity
            if idle_for < self.heartbeat_seconds:
                continue
            label = "thinking" if self._printed_thinking and not self._printed_answer else "working"
            print(f"\n[{label}...]", end="", flush=True)
            self._last_activity = time.monotonic()


def main() -> None:
    workspace = Path.cwd()
    load_core_agent_env()
    apply_compat_env_aliases()
    config = build_core_agent_config()

    provider = OpenAICompatibleProvider(
        model=config.model,
        api_key=config.api_key,
        base_url=config.base_url,
        timeout=config.timeout,
    )
    agent = CoreAgent(
        provider=provider,
        workspace=workspace,
        skill_dirs=[workspace / "skills", Path(__file__).resolve().parent / "sample_skills"],
    )
    session = agent.new_session()
    renderer = TerminalRenderer()

    print("Core Agent multi-turn chat. Type `exit` to quit.")
    while True:
        try:
            user_input = input("\nYou: ").strip()
        except EOFError:
            print()
            break
        if user_input.lower() in {"exit", "quit", "q"}:
            break
        if not user_input:
            continue

        renderer.start_turn()
        try:
            for event in session.stream_ask(user_input):
                if event.type == "reasoning":
                    renderer.on_reasoning(event.delta)
                elif event.type == "content":
                    renderer.on_content(event.delta)
                elif event.type == "tool_call":
                    renderer.on_tool_call(event.tool_name, event.tool_args)
                elif event.type == "tool_result":
                    renderer.on_tool_result(event.tool_result)
                elif event.type == "final":
                    renderer.on_final(event.final_response)
        except Exception as exc:
            print(f"\n[error] {exc}", flush=True)
        finally:
            renderer.stop_turn()


def _compact_json(value: Any, limit: int = 200) -> str:
    try:
        text = json.dumps(value, ensure_ascii=False)
    except Exception:
        text = str(value)
    if len(text) <= limit:
        return text
    return text[:limit] + "..."


def _summarize_tool_result(tool_result: str, limit: int = 240) -> str:
    try:
        payload = json.loads(tool_result)
        if isinstance(payload, dict):
            if "stdout" in payload and payload["stdout"]:
                text = str(payload["stdout"]).strip()
            elif "content" in payload and payload["content"]:
                text = str(payload["content"]).strip()
            elif "entries" in payload and isinstance(payload["entries"], list):
                text = f"{len(payload['entries'])} entries"
            elif "tasks" in payload and isinstance(payload["tasks"], list):
                text = f"{len(payload['tasks'])} tasks"
            else:
                text = json.dumps(payload, ensure_ascii=False)
        else:
            text = str(payload)
    except Exception:
        text = tool_result
    text = " ".join(text.split())
    if len(text) <= limit:
        return text
    return text[:limit] + "..."


if __name__ == "__main__":
    main()