aria-code 0.52.0__py3-none-any.whl

aria_code/ui/render/output.py

@@ -0,0 +1,775 @@
+"""Generic tool-result and error rendering for Aria Code.
+
+All functions accept console / has_rich as parameters so they stay
+import-free from aria_cli.py and testable in isolation.
+
+Public surface
+--------------
+    FINANCE_TOOL_NAMES          frozenset of tool names with dedicated renderers
+    clean_tool_error_message(e) short user-facing string from any exception
+    error_hint(msg, context)    actionable recovery suggestion
+    print_error(msg, context, *, console, has_rich, rich_box)
+    print_tool_result(...)
+"""
+
+from __future__ import annotations
+
+import difflib
+import pathlib
+import re
+import time
+
+
+# ── Finance tool name registry ─────────────────────────────────────────────────
+
+FINANCE_TOOL_NAMES: frozenset = frozenset({
+    "get_market_data", "get_market_history", "get_crypto_data", "get_forex_data",
+    "get_commodities_data", "get_futures_data", "calculate_factors",
+    "backtest_strategy", "cloud_backtest", "run_portfolio_backtest", "get_risk_metrics",
+    "optimize_positions", "get_sector_performance", "get_northbound_flow",
+    "screen_ashare", "get_limit_up_pool", "get_market_indices",
+    "analyze_news", "get_bonds_data", "get_ai_signal",
+    "get_market_insights", "get_predictions",
+    "broker_query", "broker_order",
+})
+
+
+# ── Narrow-terminal Markdown adaptation ──────────────────────────────────────
+
+_TABLE_SEPARATOR_CELL = re.compile(r"^:?-{3,}:?$")
+
+
+def _markdown_table_cells(line: str) -> list[str]:
+    """Split one simple GFM table row without leaking surrounding pipes."""
+    stripped = line.strip()
+    if not stripped.startswith("|"):
+        return []
+    return [cell.strip() for cell in stripped.strip("|").split("|")]
+
+
+def _is_markdown_table_separator(line: str) -> bool:
+    cells = _markdown_table_cells(line)
+    return bool(cells) and all(_TABLE_SEPARATOR_CELL.fullmatch(cell) for cell in cells)
+
+
+def adapt_markdown_for_width(markup: str, width: int, *, table_breakpoint: int = 96) -> str:
+    """Convert GFM tables to stacked records when the terminal is narrow.
+
+    Rich correctly renders tables, but a three-column research table inside an
+    80-column terminal has too little room for Chinese prose.  Cells are then
+    truncated or wrapped until row relationships become unreadable.  The source
+    Markdown remains unchanged in saved reports; only the terminal presentation
+    is adapted.
+    """
+    if width >= table_breakpoint or "|" not in (markup or ""):
+        return markup
+
+    lines = (markup or "").splitlines()
+    output: list[str] = []
+    index = 0
+    in_fence = False
+    while index < len(lines):
+        line = lines[index]
+        if line.lstrip().startswith("```"):
+            in_fence = not in_fence
+            output.append(line)
+            index += 1
+            continue
+
+        if (
+            not in_fence
+            and index + 1 < len(lines)
+            and line.strip().startswith("|")
+            and _is_markdown_table_separator(lines[index + 1])
+        ):
+            headers = _markdown_table_cells(line)
+            cursor = index + 2
+            rows: list[list[str]] = []
+            while cursor < len(lines) and lines[cursor].strip().startswith("|"):
+                row = _markdown_table_cells(lines[cursor])
+                if row:
+                    rows.append(row)
+                cursor += 1
+
+            if headers and rows:
+                for row in rows:
+                    pairs = [
+                        (header, row[pos] if pos < len(row) else "—")
+                        for pos, header in enumerate(headers)
+                        if header
+                    ]
+                    if not pairs:
+                        continue
+                    first_header, first_value = pairs[0]
+                    output.append(f"- **{first_header}**：{first_value or '—'}")
+                    for header, value in pairs[1:]:
+                        output.append(f"  - **{header}**：{value or '—'}")
+                index = cursor
+                continue
+
+        output.append(line)
+        index += 1
+    return "\n".join(output)
+
+
+# ── Tool display helpers ──────────────────────────────────────────────────────
+
+def tool_display_kind(tool_name: str) -> str:
+    """Return a user-facing service/tool kind without exposing local targets."""
+    if tool_name.startswith("mcp__"):
+        return "MCP"
+    if tool_name in FINANCE_TOOL_NAMES:
+        return "finance tool"
+    if tool_name in {"web_search", "search_web"}:
+        return "web search"
+    if tool_name == "web_fetch":
+        return "web fetch"
+    if tool_name in {"read_file", "write_file", "edit_file", "list_files", "search_code"}:
+        return "file tool"
+    if tool_name == "run_command":
+        return "shell tool"
+    if tool_name.startswith("skill") or tool_name in {"TaskCreate", "TaskUpdate"}:
+        return "skill"
+    if tool_name.startswith("broker_") or tool_name in {"broker_query", "broker_order"}:
+        return "broker tool"
+    return "tool"
+
+
+def tool_display_label(tool_name: str) -> str:
+    """Short label for activity UI: tool name plus its service kind."""
+    if tool_name.startswith("mcp__"):
+        parts = tool_name.split("__")
+        if len(parts) >= 3:
+            return f"{parts[1]} · {parts[2].replace('_', ' ')} · MCP"
+        return "MCP"
+    return f"{tool_name} · {tool_display_kind(tool_name)}"
+
+
+def format_turn_footer(metadata, *, mode: str = "compact", copy_available: bool = False) -> str:
+    """Return the post-response status line.
+
+    ``full`` keeps the historical token-heavy line for debugging.  ``compact``
+    is the default interactive UI: elapsed time, provider, tools, and /copy.
+    """
+    mode = (mode or "compact").strip().lower()
+    if mode in {"off", "none", "false", "0"}:
+        return ""
+    parts = list(getattr(metadata, "parts", []) or [])
+    if mode in {"full", "debug", "verbose"}:
+        footer = " · ".join(parts)
+        if copy_available:
+            footer = f"{footer}  /copy" if footer else "/copy"
+        return footer
+
+    elapsed = parts[0] if parts else ""
+    provider = str(getattr(metadata, "provider", "") or "").strip()
+    tools = list(getattr(metadata, "tools", []) or [])
+    out: list[str] = []
+    if elapsed:
+        out.append(elapsed)
+    if provider and provider not in {"aws", "local"}:
+        out.append(provider)
+    if tools:
+        shown = " ".join(str(t) for t in tools[:3])
+        if len(tools) > 3:
+            shown += f" +{len(tools) - 3}"
+        out.append(shown)
+    if copy_available:
+        out.append("/copy")
+    return " · ".join(out)
+
+
+def display_path(path: object, *, fallback: str = "file") -> str:
+    """Return a path-safe display value for user-facing UI."""
+    if not path:
+        return fallback
+    try:
+        name = pathlib.Path(str(path)).name
+    except Exception:
+        name = ""
+    return name or fallback
+
+
+# ── Error helpers ──────────────────────────────────────────────────────────────
+
+def clean_tool_error_message(error: object) -> str:
+    raw = str(error or "failed").strip()
+    low = raw.lower()
+    if not raw:
+        return "操作失败"
+    if "curl: (28)" in low or "timed out" in low or "timeout" in low:
+        return "请求超时，数据源暂时不可用。请稍后重试或运行 /health 检查服务。"
+    if "connection refused" in low:
+        return "连接被拒绝，服务暂时不可用。请检查本地服务或网络。"
+    if "connection aborted" in low or "remotedisconnected" in low:
+        return "网络连接中断，数据源未完成响应。请稍后重试。"
+    # Generic connection / proxy / DNS failures — collapse the verbose
+    # urllib3 HTTPSConnectionPool(...) dump into a single readable line.
+    if any(s in low for s in (
+        "httpsconnectionpool", "httpconnectionpool", "max retries exceeded",
+        "proxyerror", "failed to establish a new connection",
+        "nameresolutionerror", "getaddrinfo failed", "newconnectionerror",
+    )):
+        import re as _re3
+        _host = _re3.search(r"host=['\"]([^'\"]+)['\"]", raw)
+        _hint = f"（数据源 {_host.group(1)}）" if _host else ""
+        return f"数据源连接失败{_hint}，可能是网络或代理问题。请检查网络后重试。"
+    if "rate" in low or "429" in low or "too many requests" in low:
+        return "数据源请求频率受限，请稍后重试。"
+    # Collapse verbose HTTP error strings: "web_fetch failed: 401 Client Error: Unauthorized for url: https://..."
+    import re as _re
+    _http = _re.match(r"web_fetch failed:\s*(\d{3})\s+\w[\w\s]+?:\s*([\w\s]+?)(?:\s+for url:.*)?$", raw, _re.I)
+    if _http:
+        code, phrase = _http.group(1), _http.group(2).strip()
+        return f"HTTP {code} {phrase}"
+    if "traceback" in low:
+        return raw.splitlines()[-1][:160] if raw.splitlines() else "运行失败"
+    return raw[:200]
+
+
+def error_hint(error: str, context: str = "") -> str:
+    err_lower = error.lower() if error else ""
+    if "connection" in err_lower or "refused" in err_lower or "unreachable" in err_lower:
+        return "Hint: Backend unreachable. Try /health or check your network."
+    if "timeout" in err_lower or "timed out" in err_lower:
+        return "Hint: Request timed out. Try again or check /health."
+    # External web pages that block scraping (paywall / anti-bot) — NOT an Aria
+    # login problem, so /login must not be suggested.
+    _is_web = any(m in err_lower for m in (
+        "http://", "https://", "www.", ".com", ".org", ".net",
+        "web_fetch", "web fetch", "forbidden",
+    ))
+    if "401" in err_lower or "unauthorized" in err_lower:
+        if any(h in err_lower for h in ("finnhub", "alphavantage", "polygon", "api/v1", "api/v2/finance")):
+            return "Hint: API key required — /apikey set finnhub <KEY>  (free at finnhub.io)"
+        if _is_web:
+            return "Hint: This site blocks automated access (paywall/anti-bot). Try another source."
+        return "Hint: Authentication required. Run /login to sign in."
+    if "403" in err_lower or "forbidden" in err_lower:
+        if _is_web:
+            return "Hint: This site blocks automated access (paywall/anti-bot). Try another source."
+        return "Hint: Access denied. Check your API key or subscription."
+    if "429" in err_lower or "rate" in err_lower:
+        return "Hint: Rate limited. Wait a moment and try again."
+    if ("ollama" in err_lower or "ollama http" in err_lower) and (
+        "not found" in err_lower or "404" in err_lower
+    ):
+        m = re.search(r"model ['\"]?([^'\"]+)['\"]? not found", err_lower)
+        model_hint = m.group(1) if m else "the requested model"
+        try:
+            from local_llm_provider import list_ollama_models
+            available = list_ollama_models("http://localhost:11434")
+            if available:
+                suggestion = available[0]
+                return (
+                    f"Hint: Ollama model '{model_hint}' not found.\n"
+                    f"  Available: {', '.join(available[:4])}\n"
+                    f"  Run: /config model {suggestion}"
+                )
+        except Exception:
+            pass
+        return (
+            "Hint: Ollama model not found. Run `ollama list` to see available models.\n"
+            "  Or pull one: ollama pull qwen2.5-coder:7b"
+        )
+    # "File not found" is a path error. Tell the model firmly NOT to keep
+    # guessing filenames (it otherwise loops app.py→script.py→main.py…).
+    if "file not found" in err_lower or "no such file" in err_lower:
+        return ("Hint: This file does not exist. Do NOT guess other filenames — "
+                "list the directory first, or this question may not need a file at all.")
+    if "404" in err_lower and context == "tool":
+        return "Hint: Tool not available. Check /tools for available tools."
+    if "not found" in err_lower and context == "session":
+        return "Hint: Session not found. Run /sessions to list available."
+    if "404" in err_lower or ("not found" in err_lower and context not in ("tool", "")):
+        return "Hint: Resource not found. Check the symbol or path."
+    if "no data" in err_lower or "no result" in err_lower:
+        return "Hint: No data returned. Verify the symbol spelling."
+    if "500" in err_lower or "internal" in err_lower:
+        return "Hint: Server error. Try again in a moment or /health to check."
+    if context == "login":
+        return "Hint: Check email/password. Usage: /login email password"
+    return ""
+
+
+# ── Error panel & tree formatting ──────────────────────────────────────────────
+
+def print_error(
+    msg: str,
+    context: str = "",
+    *,
+    console,
+    has_rich: bool,
+    rich_box=None,
+    use_panel: bool = False,
+) -> None:
+    hint = error_hint(msg, context)
+    if not use_panel:
+        # Claude Code style: clean └ tree connector for inline error guidance
+        if has_rich:
+            console.print(f"  [red]└ {msg}[/red]")
+            if hint:
+                for hline in hint.splitlines():
+                    console.print(f"    [dim]{hline}[/dim]")
+        else:
+            print(f"  └ {msg}")
+            if hint:
+                for hline in hint.splitlines():
+                    print(f"    {hline}")
+        return
+
+    if has_rich:
+        from rich.panel import Panel
+        box_style = getattr(rich_box, "ROUNDED", None) if rich_box else None
+        body = f"[red]{msg}[/red]"
+        if hint:
+            body += f"\n[dim]{hint}[/dim]"
+        console.print(Panel(body, border_style="red", box=box_style, padding=(0, 1)))
+    else:
+        print(msg)
+
+
+# ── Tool result ────────────────────────────────────────────────────────────────
+
+def print_tool_result(
+    tool_name: str,
+    result: dict,
+    elapsed: float = 0,
+    params: dict = None,
+    *,
+    console,
+    has_rich: bool,
+    rich_box,
+    print_finance_fn,   # callable(tool_name, result) for finance tools
+    bot_mode: bool = False,
+) -> None:
+    """Render a tool result summary — Codex-style ⎿ tree connector."""
+    if bot_mode:
+        return
+
+    ts       = f"  [dim]{elapsed:.1f}s[/dim]" if elapsed >= 0.1 else ""
+    ts_plain = f"  {elapsed:.1f}s" if elapsed >= 0.1 else ""
+    params   = params or {}
+
+    if tool_name in FINANCE_TOOL_NAMES:
+        print_finance_fn(tool_name, result)
+        if ts and has_rich:
+            console.print(f"  [dim]⎿[/dim]{ts}")
+        return
+
+    if result.get("success"):
+        data = result.get("data", {})
+
+        if tool_name == "write_file":
+            lines     = data.get("lines") or (params.get("content", "").count("\n") + 1 if params.get("content") else 0)
+            size      = data.get("size_bytes") or len((params.get("content", "") or "").encode())
+            size_str  = f"{size}B" if size < 1024 else f"{size // 1024}KB"
+            if has_rich:
+                console.print(f"  [dim]⎿[/dim]  [green]✓[/green]  [dim]file tool  {lines} lines  {size_str}[/dim]{ts}")
+            else:
+                print(f"  ⎿  ✓ file tool  {lines} lines  {size_str}{ts_plain}")
+
+        elif tool_name == "edit_file":
+            old = params.get("old_string", "")
+            new = params.get("new_string", "")
+            if old and new and has_rich:
+                diff = list(difflib.unified_diff(
+                    old.splitlines(keepends=True),
+                    new.splitlines(keepends=True),
+                    lineterm="",
+                ))
+                if diff:
+                    _hdr = "  [dim]⎿[/dim]  [#C08050]file tool[/#C08050]"
+                    console.print(f"{_hdr}{ts}")
+                    from rich.syntax import Syntax
+                    from rich.panel import Panel
+                    diff_text = "".join(diff[2:])
+                    console.print(Panel(
+                        Syntax(diff_text, "diff", theme="monokai", padding=(0, 1)),
+                        border_style="dim",
+                        box=rich_box.SIMPLE,
+                    ))
+                else:
+                    console.print(f"  [dim]⎿  no change[/dim]{ts}")
+            elif has_rich:
+                console.print(f"  [dim]⎿  edited[/dim]{ts}")
+            else:
+                print(f"  ⎿  edited{ts_plain}")
+
+        elif tool_name == "run_command":
+            stdout     = data.get("stdout", "").strip()
+            returncode = data.get("returncode", data.get("exit_code", 0))
+            if has_rich:
+                from rich.panel import Panel
+                rc_color = "green" if returncode == 0 else "red"
+                rc_icon  = "✓" if returncode == 0 else "✗"
+                console.print(f"  [dim]⎿[/dim]  [{rc_color}]{rc_icon} exit {returncode}[/{rc_color}]{ts}")
+                if stdout:
+                    out_lines = stdout.splitlines()
+                    if len(out_lines) > 15:
+                        head = out_lines[:5]
+                        tail = out_lines[-10:]
+                        truncated = "\n".join(head) + f"\n[dim]… ({len(out_lines) - 15} lines hidden) …[/dim]\n" + "\n".join(tail)
+                        if data.get("full_output_path"):
+                            truncated += f"\n[dim]full output saved: {data.get('full_output_path')}[/dim]"
+                        console.print(Panel(
+                            f"[dim]{truncated}[/dim]",
+                            border_style="dim",
+                            box=rich_box.SIMPLE,
+                            padding=(0, 1),
+                        ))
+                    else:
+                        truncated = "\n".join(out_lines)
+                        if data.get("full_output_path"):
+                            truncated += f"\n[dim]full output saved: {data.get('full_output_path')}[/dim]"
+                        console.print(Panel(
+                            f"[dim]{truncated}[/dim]",
+                            border_style="dim",
+                            box=rich_box.SIMPLE,
+                            padding=(0, 1),
+                        ))
+            else:
+                print(f"  ⎿  exit {returncode}{ts_plain}")
+                for ol in stdout.splitlines()[:4]:
+                    print(f"    {ol[:100]}")
+
+        elif tool_name == "read_file":
+            lines = data.get("lines", 0)
+            if has_rich:
+                console.print(f"  [dim]⎿  file tool  {lines} lines[/dim]{ts}")
+            else:
+                print(f"  ⎿  file tool  {lines} lines{ts_plain}")
+
+        elif tool_name == "list_files":
+            count = data.get("count", 0)
+            if has_rich:
+                color = "yellow" if count == 0 else "dim"
+                msg   = "0 items — no matches" if count == 0 else f"{count} items"
+                console.print(f"  [{color}]⎿  {msg}[/{color}]{ts}")
+            else:
+                print(f"  ⎿  {count} items{ts_plain}")
+
+        elif tool_name == "search_code":
+            matches = len(data.get("matches", []))
+            if has_rich:
+                console.print(f"  [dim]⎿  {matches} matches[/dim]{ts}")
+            else:
+                print(f"  ⎿  {matches} matches{ts_plain}")
+
+        elif tool_name == "web_fetch":
+            length = data.get("length", 0)
+            trunc  = data.get("truncated", False)
+            len_str = f"  {length:,} chars" if length else ""
+            trunc_str = "  [dim]truncated[/dim]" if trunc else ""
+            if has_rich:
+                console.print(f"  [dim]⎿  web fetch{len_str}[/dim]{trunc_str}{ts}")
+            else:
+                print(f"  ⎿  web fetch{ts_plain}")
+
+        elif tool_name in ("web_search", "search_web"):
+            results = data.get("results", [])
+            count   = len(results)
+            if has_rich:
+                console.print(f"  [dim]⎿  {count} results[/dim]{ts}")
+            else:
+                print(f"  ⎿  {count} results{ts_plain}")
+
+        else:
+            short = tool_display_kind(tool_name)
+            if has_rich:
+                console.print(f"  [dim]⎿  {short} done[/dim]{ts}")
+            else:
+                print(f"  ⎿  done{ts_plain}")
+
+    else:
+        error = clean_tool_error_message(result.get("error", "failed"))
+        hint  = error_hint(str(error), context="tool")
+        if has_rich:
+            console.print(f"  [dim]⎿[/dim]  [red]✗ {error[:120]}[/red]")
+            if hint:
+                console.print(f"    [dim]{hint}[/dim]")
+        else:
+            print(f"  ⎿  ✗ {error[:80]}")
+
+
+# ── Activity group (OpenClaw-style batch summary) ──────────────────────────────
+
+def _one_line_tool_summary(
+    tool_name: str,
+    result: dict,
+    elapsed: float,
+    params: dict,
+) -> tuple[str, str]:
+    """Return (status_markup, detail_markup) for one tool in an activity table."""
+    params = params or {}
+    ts = f"[dim]  {elapsed:.1f}s[/dim]" if elapsed >= 0.1 else ""
+
+    if not result.get("success"):
+        error = clean_tool_error_message(result.get("error", "failed"))
+        return "[red]✗[/red]", f"[red]{error[:80]}[/red]{ts}"
+
+    data = result.get("data", {})
+    kind = tool_display_kind(tool_name)
+
+    if tool_name == "write_file":
+        lines = data.get("lines") or (params.get("content", "").count("\n") + 1 if params.get("content") else 0)
+        size  = data.get("size_bytes") or len((params.get("content", "") or "").encode())
+        size_str = f"{size}B" if size < 1024 else f"{size // 1024}KB"
+        return "[green]✓[/green]", f"[dim]{kind}  {lines} lines  {size_str}[/dim]{ts}"
+
+    elif tool_name == "edit_file":
+        return "[green]✓[/green]", f"[dim]edited  {kind}[/dim]{ts}"
+
+    elif tool_name == "run_command":
+        rc = data.get("returncode", data.get("exit_code", 0))
+        icon  = "[green]✓[/green]" if rc == 0 else "[red]✗[/red]"
+        color = "green" if rc == 0 else "red"
+        suffix = " [dim]· full output saved[/dim]" if data.get("full_output_path") else ""
+        return icon, f"[{color}]exit {rc}[/{color}]{suffix}{ts}"
+
+    elif tool_name == "read_file":
+        lines = data.get("lines", 0)
+        return "[green]✓[/green]", f"[dim]{kind}  {lines} lines[/dim]{ts}"
+
+    elif tool_name == "list_files":
+        count = data.get("count", 0)
+        color = "yellow" if count == 0 else "dim"
+        msg   = "no matches" if count == 0 else f"{count} items"
+        return "[green]✓[/green]", f"[{color}]{msg}[/{color}]{ts}"
+
+    elif tool_name == "search_code":
+        matches = len(data.get("matches", []))
+        return "[green]✓[/green]", f"[dim]{matches} matches[/dim]{ts}"
+
+    elif tool_name == "web_fetch":
+        length = data.get("length", 0)
+        len_s  = f"  {length:,}c" if length else ""
+        return "[green]✓[/green]", f"[dim]{kind}{len_s}[/dim]{ts}"
+
+    elif tool_name in ("web_search", "search_web"):
+        count = len(data.get("results", []))
+        return "[green]✓[/green]", f"[dim]{count} results[/dim]{ts}"
+
+    else:
+        return "[green]✓[/green]", f"[dim]{kind} done[/dim]{ts}"
+
+
+def print_tool_activity_group(
+    results: list,       # list of (tool_name, result, elapsed, params)
+    *,
+    console,
+    has_rich: bool,
+    rich_box,
+    print_finance_fn,
+    bot_mode: bool = False,
+) -> None:
+    """Render multiple tool results as a compact Activity block (OpenClaw style).
+
+    For N >= 2 tools: prints a titled table.
+    For N == 1: delegates to print_tool_result (single-line).
+    """
+    if bot_mode or not results:
+        return
+
+    if len(results) == 1:
+        tool_name, result, elapsed, params = results[0]
+        print_tool_result(tool_name, result, elapsed, params,
+                          console=console, has_rich=has_rich, rich_box=rich_box,
+                          print_finance_fn=print_finance_fn, bot_mode=bot_mode)
+        return
+
+    total_elapsed = sum(e for _, _, e, _ in results)
+    n = len(results)
+
+    # Finance tools: print with dedicated renderer, then add to activity table
+    finance_rows = []
+    for tool_name, result, elapsed, params in results:
+        if tool_name in FINANCE_TOOL_NAMES:
+            print_finance_fn(tool_name, result)
+            finance_rows.append(tool_name)
+
+    if has_rich:
+        from rich.table import Table
+        ts_total = f"  [dim]{total_elapsed:.1f}s[/dim]" if total_elapsed >= 0.1 else ""
+        header = f"[dim]Activity · {n} tools[/dim]{ts_total}"
+        console.print(f"\n  {header}")
+        tbl = Table.grid(padding=(0, 2))
+        tbl.add_column(no_wrap=True, min_width=14, style="dim")   # tool name
+        tbl.add_column(no_wrap=True, min_width=2)                  # status icon
+        tbl.add_column()                                            # detail
+
+        from collections import OrderedDict
+        _mcp_groups: "OrderedDict[str, list]" = OrderedDict()
+        for tool_name, result, elapsed, params in results:
+            if tool_name in finance_rows:
+                icon = "[green]✓[/green]" if result.get("success") else "[red]✗[/red]"
+                tbl.add_row(tool_name, icon, "")
+            elif tool_name.startswith("mcp__"):
+                # Defer MCP calls — collapse per server below
+                _server = tool_name.split("__")[1] if len(tool_name.split("__")) >= 2 else "mcp"
+                _mcp_groups.setdefault(_server, []).append((tool_name, result))
+            else:
+                icon, detail = _one_line_tool_summary(tool_name, result, elapsed, params)
+                tbl.add_row(f"[dim]{tool_name}[/dim]", icon, detail)
+
+        # Collapsed MCP rows: "server · tool" for one, "called N times" for many
+        for _server, _calls in _mcp_groups.items():
+            _all_ok = all(r.get("success") for _, r in _calls)
+            _icon = "[green]✓[/green]" if _all_ok else "[red]✗[/red]"
+            if len(_calls) == 1:
+                _tn = _calls[0][0].split("__")
+                _label = _tn[2].replace("_", " ") if len(_tn) >= 3 else _server
+                tbl.add_row(f"[dim]{_server}[/dim]", _icon, f"[dim]{_label}  · MCP[/dim]")
+            else:
+                tbl.add_row(f"[dim]{_server}[/dim]", _icon,
+                            f"[dim]called {len(_calls)} times · MCP[/dim]")
+
+        from rich.padding import Padding
+        console.print(Padding(tbl, (0, 0, 0, 4)))
+
+        # For run_command with stdout, still print the output panel
+        for tool_name, result, elapsed, params in results:
+            if tool_name == "run_command" and result.get("success"):
+                stdout = result.get("data", {}).get("stdout", "").strip()
+                if stdout:
+                    from rich.panel import Panel
+                    out_lines = stdout.splitlines()
+                    if len(out_lines) > 15:
+                        head = out_lines[:5]
+                        tail = out_lines[-10:]
+                        truncated = "\n".join(head) + f"\n[dim]… ({len(out_lines) - 15} lines hidden) …[/dim]\n" + "\n".join(tail)
+                        if result.get("data", {}).get("full_output_path"):
+                            truncated += f"\n[dim]full output saved: {result.get('data', {}).get('full_output_path')}[/dim]"
+                        console.print(Panel(f"[dim]{truncated}[/dim]",
+                                            border_style="dim", box=rich_box.SIMPLE,
+                                            padding=(0, 1)))
+                    else:
+                        truncated = "\n".join(out_lines)
+                        if result.get("data", {}).get("full_output_path"):
+                            truncated += f"\n[dim]full output saved: {result.get('data', {}).get('full_output_path')}[/dim]"
+                        console.print(Panel(f"[dim]{truncated}[/dim]",
+                                            border_style="dim", box=rich_box.SIMPLE,
+                                            padding=(0, 1)))
+
+            # For edit_file, still print diff
+            elif tool_name == "edit_file" and result.get("success"):
+                old = (params or {}).get("old_string", "")
+                new = (params or {}).get("new_string", "")
+                if old and new:
+                    diff = list(difflib.unified_diff(
+                        old.splitlines(keepends=True),
+                        new.splitlines(keepends=True),
+                        lineterm="",
+                    ))
+                    if diff:
+                        from rich.syntax import Syntax
+                        from rich.padding import Padding
+                        diff_text = "".join(diff[2:])
+                        console.print(Padding(Panel(
+                            Syntax(diff_text, "diff", theme="monokai", padding=(0, 1)),
+                            border_style="dim", box=rich_box.SIMPLE
+                        ), (0, 0, 0, 4)))
+    else:
+        ts_total = f"  {total_elapsed:.1f}s" if total_elapsed >= 0.1 else ""
+        print(f"\n  Activity · {n} tools{ts_total}")
+        for tool_name, result, elapsed, params in results:
+            icon, detail = _one_line_tool_summary(tool_name, result, elapsed, params)
+            detail_plain = re.sub(r"\[/?[^\]]+\]", "", detail)
+            icon_plain   = "✓" if result.get("success") else "✗"
+            print(f"    {tool_name:<18}{icon_plain}  {detail_plain}")
+
+
+# ── Fallback / model-switch toast ──────────────────────────────────────────────
+
+def print_fallback_toast(
+    from_provider: str,
+    to_provider: str,
+    reason: str = "",
+    *,
+    console,
+    has_rich: bool,
+) -> None:
+    """Show a transient yellow notification when the active model/provider switches."""
+    if not has_rich:
+        print(f"\n  ⚡ 模型切换  {from_provider} → {to_provider}{('  ' + reason) if reason else ''}")
+        return
+    body = f"[bold #C08050]⚡[/bold #C08050]  [#C08050]{from_provider}[/#C08050] [dim]→[/dim] [#C08050]{to_provider}[/#C08050]"
+    if reason:
+        body += f"\n  [dim]{reason}[/dim]"
+    console.print(f"\n  {body}")
+
+
+# ── Context pressure warning ───────────────────────────────────────────────────
+
+_CTX_WARNED: dict[str, float] = {}   # session_id → last warn time
+
+def print_context_warning(
+    est_tokens: int,
+    max_tokens: int,
+    *,
+    console,
+    has_rich: bool,
+    session_id: str = "",
+    cooldown: float = 120.0,         # only warn once every 2 min per session
+) -> None:
+    """Warn when context is >85% full; rate-limited to avoid spam."""
+    if max_tokens <= 0:
+        return
+    ratio = est_tokens / max_tokens
+    if ratio < 0.85:
+        return
+    now = time.monotonic()
+    if now - _CTX_WARNED.get(session_id, 0) < cooldown:
+        return
+    _CTX_WARNED[session_id] = now
+
+    def _k(n: int) -> str:
+        return f"{n // 1000}K" if n >= 1000 else str(n)
+
+    pct = int(ratio * 100)
+    if has_rich:
+        color  = "red" if ratio >= 0.95 else "#C08050"
+        icon   = "●" if ratio >= 0.95 else "⚠"
+        msg    = f"  [{color}]{icon} 上下文 {pct}% 已满  ({_k(est_tokens)}/{_k(max_tokens)} tokens)[/{color}]"
+        msg   += "  [dim]→ /compact 压缩历史  /clear 重置[/dim]"
+        console.print(msg)
+    else:
+        print(f"  ⚠ 上下文 {pct}% ({_k(est_tokens)}/{_k(max_tokens)} tokens) — /compact 或 /clear")
+
+
+# ── Blocked / cancelled tool visual ───────────────────────────────────────────
+
+def print_tool_blocked(
+    tool_name: str,
+    reason: str = "用户取消",
+    *,
+    console,
+    has_rich: bool,
+) -> None:
+    """Show a styled 'Blocked' line when tool execution is denied or cancelled."""
+    if has_rich:
+        console.print(
+            f"  [dim]⎿[/dim]  [#C08050]⊘  {tool_name}[/#C08050]  [dim]{reason}[/dim]"
+        )
+    else:
+        print(f"  ⎿  ⊘ {tool_name}  {reason}")
+
+
+# ── Robot thinking / response header ──────────────────────────────────────────
+
+def print_thinking_header(*, console, has_rich: bool, agent_name: str = "Aria", color: str = "#C08050") -> None:
+    """Print a subtle copper 'Aria ▸' header before each AI response stream.
+
+    Gives the response a clear starting-point rather than appearing inline.
+    Called once per turn, right before the first streaming token is printed.
+    """
+    if not has_rich:
+        return
+    console.print(f"[bold {color}]▣[/bold {color}]  [dim {color}]{agent_name}[/dim {color}]", end="  ")
+
+
+def print_done_footer(elapsed: float, *, console, has_rich: bool) -> None:
+    """Print a dim elapsed-time line after the response stream ends."""
+    if not has_rich:
+        return
+    console.print(f"\n[dim]  ✓  {elapsed:.1f}s[/dim]")