aria-code 0.167.0__py3-none-win_amd64.whl

aria_code/apps/cli/skills_catalog.py

@@ -0,0 +1,273 @@
+"""Skills 目录 —— 从 aria_cli.py 抽出的纯数据。
+
+SKILLS 字面量原本在 aria_cli.py 的 1772-2035 行（264 行）。同 model_catalog：
+纯字面量、零耦合，用普通 import 即可，core_cmds / diagnostic_cmds 的裸名
+引用不受影响。
+"""
+
+from __future__ import annotations
+
+SKILLS = [
+    {
+        "command": "/morning-brief",
+        "name": "Morning Brief",
+        "category": "research",
+        "description": "Daily market briefing with key events and outlook",
+        "args": "[focus_area]",
+        "prompt": (
+            "Generate a comprehensive morning market briefing:\n"
+            "1. US market futures and overnight moves\n"
+            "2. Key economic events and earnings today\n"
+            "3. Global markets overview (Asia, Europe)\n"
+            "4. Top sector movers and themes\n"
+            "5. Trading outlook and key levels to watch\n"
+            "{extra}"
+        ),
+        "tools_hint": ["web_search", "get_market_indices", "get_sector_performance", "analyze_news"],
+    },
+    {
+        "command": "/deep-analysis",
+        "name": "Deep Analysis",
+        "category": "analysis",
+        "description": "Multi-factor stock deep dive (technical + fundamental + sentiment)",
+        "args": "<symbol>",
+        "prompt": (
+            "Perform a comprehensive multi-factor analysis of {symbol}:\n"
+            "1. Technical Analysis: trend, support/resistance, indicators (RSI, MACD, Bollinger)\n"
+            "2. Fundamental Analysis: PE, PB, revenue growth, margins, debt ratios\n"
+            "3. Sentiment Analysis: recent news sentiment, analyst ratings, social buzz\n"
+            "4. Risk Assessment: VaR, beta, max drawdown potential\n"
+            "5. Verdict: Bull/Bear/Neutral with confidence level and price targets"
+        ),
+        "tools_hint": ["web_search", "get_market_data", "calculate_factors", "analyze_news", "peer_comparison", "piotroski_fscore", "get_risk_metrics"],
+    },
+    {
+        "command": "/trade-idea",
+        "name": "Trade Idea",
+        "category": "strategy",
+        "description": "AI-generated trade ideas with entry/exit levels",
+        "args": "[market_or_sector]",
+        "prompt": (
+            "Generate 3 actionable trade ideas{context}:\n"
+            "For each idea provide:\n"
+            "1. Symbol and direction (Long/Short)\n"
+            "2. Entry zone, stop loss, and 2 take-profit levels\n"
+            "3. Risk-reward ratio\n"
+            "4. Catalyst: what's driving the trade\n"
+            "5. Timeframe (swing/position/day)\n"
+            "6. Confidence level (1-10)"
+        ),
+        "tools_hint": ["web_search", "get_market_data", "analyze_news", "recommend_strategy"],
+    },
+    {
+        "command": "/risk-report",
+        "name": "Risk Report",
+        "category": "risk",
+        "description": "Portfolio risk analysis with VaR, stress tests, and correlation",
+        "args": "[symbols...]",
+        "prompt": (
+            "Generate a comprehensive risk report for portfolio: {symbols}\n"
+            "1. Portfolio VaR (95%, 99%) — daily and monthly\n"
+            "2. Correlation matrix between holdings\n"
+            "3. Concentration risk by sector/geography\n"
+            "4. Stress test scenarios (2008 crisis, COVID crash, rate hike)\n"
+            "5. Tail risk analysis\n"
+            "6. Recommendations: rebalancing suggestions to reduce risk"
+        ),
+        "tools_hint": ["assess_portfolio_risk", "get_risk_metrics", "stress_test_strategy"],
+    },
+    {
+        "command": "/sector-rotation",
+        "name": "Sector Rotation",
+        "category": "strategy",
+        "description": "Sector rotation analysis with economic cycle positioning",
+        "args": "",
+        "prompt": (
+            "Analyze current sector rotation dynamics:\n"
+            "1. Current economic cycle phase (early/mid/late/recession)\n"
+            "2. All 11 GICS sectors: performance, momentum, relative strength\n"
+            "3. Leading vs lagging sectors and why\n"
+            "4. Sector rotation strategy: which sectors to overweight/underweight\n"
+            "5. Top stock picks from the strongest sectors\n"
+            "6. Historical analog: which past period is most similar"
+        ),
+        "tools_hint": ["get_sector_performance", "get_market_indices", "analyze_news"],
+    },
+    {
+        "command": "/macro-outlook",
+        "name": "Macro Outlook",
+        "category": "research",
+        "description": "Macroeconomic analysis: rates, inflation, growth & cycle",
+        "args": "[region]",
+        "prompt": (
+            "Provide a macroeconomic outlook{context}:\n"
+            "1. GDP growth forecast and trends\n"
+            "2. Inflation trajectory (CPI, PCE) and central bank response\n"
+            "3. Interest rate path: current level and expectations\n"
+            "4. Employment situation: jobs, wages, participation\n"
+            "5. Key risks: geopolitical, financial, systemic\n"
+            "6. Asset class implications: equities, bonds, commodities, crypto"
+        ),
+        "tools_hint": ["web_search", "get_world_bank_reports", "get_bonds_data", "analyze_news"],
+    },
+    {
+        "command": "/factor-screen",
+        "name": "Factor Screen",
+        "category": "quant",
+        "description": "Factor-based stock screening (value, momentum, quality, etc.)",
+        "args": "<factor_type>",
+        "prompt": (
+            "Screen US stocks using {factor} factor strategy:\n"
+            "1. Define the factor criteria and thresholds\n"
+            "2. Top 10 stocks ranking highest on {factor}\n"
+            "3. For each: symbol, score, key metrics, sector\n"
+            "4. Historical factor performance: how has {factor} performed\n"
+            "5. Current factor environment: is {factor} in favor?\n"
+            "6. Combined multi-factor overlay suggestion"
+        ),
+        "tools_hint": ["calculate_factors", "get_alpha158_factors", "get_market_data"],
+    },
+    {
+        "command": "/crypto-scan",
+        "name": "Crypto Scanner",
+        "category": "crypto",
+        "description": "Cryptocurrency market scan with top movers and DeFi trends",
+        "args": "[focus]",
+        "prompt": (
+            "Scan the cryptocurrency market:\n"
+            "1. BTC and ETH: price, trend, dominance, key levels\n"
+            "2. Top 5 gainers and top 5 losers (24h)\n"
+            "3. Market sentiment: Fear & Greed index, funding rates\n"
+            "4. DeFi and Layer-2 highlights\n"
+            "5. Upcoming catalysts: halvings, upgrades, token unlocks\n"
+            "6. Trading opportunities with risk levels\n"
+            "{extra}"
+        ),
+        "tools_hint": ["get_crypto_data", "analyze_news"],
+    },
+    {
+        "command": "/backtest-report",
+        "name": "Backtest Report",
+        "category": "quant",
+        "description": "Run and analyze a strategy backtest with detailed metrics",
+        "args": "<strategy> <symbol> [start] [end]",
+        "prompt": (
+            "Run a detailed backtest of '{strategy}' strategy on {symbol} from {start} to {end}:\n"
+            "1. Performance summary: total return, annualized, Sharpe, Sortino\n"
+            "2. Risk metrics: max drawdown, VaR, downside deviation\n"
+            "3. Trade analysis: win rate, avg win/loss, profit factor\n"
+            "4. Monthly returns breakdown\n"
+            "5. Comparison vs buy-and-hold and benchmark (SPY)\n"
+            "6. Optimization suggestions: parameter sensitivity"
+        ),
+        "tools_hint": ["backtest_strategy", "get_market_data", "get_risk_metrics"],
+    },
+    {
+        "command": "/watchlist-scan",
+        "name": "Watchlist Scan",
+        "category": "tools",
+        "description": "Scan all watchlist stocks for signals and alerts",
+        "args": "",
+        "prompt": (
+            "Scan my watchlist ({symbols}) and for each stock provide:\n"
+            "1. Current price and daily change\n"
+            "2. Technical signal: Buy/Sell/Hold based on key indicators\n"
+            "3. Any earnings or events upcoming\n"
+            "4. News sentiment (positive/neutral/negative)\n"
+            "5. Overall alert level: Green/Yellow/Red\n"
+            "Sort by urgency of action needed."
+        ),
+        "tools_hint": ["get_market_data", "analyze_news"],
+    },
+    {
+        "command": "/gen-strategy",
+        "name": "Generate Strategy Code",
+        "category": "code",
+        "description": "Generate complete Python trading strategy code",
+        "args": "<strategy_type> [symbol]",
+        "prompt": (
+            "Generate a complete, production-ready Python backtrader trading strategy.\n"
+            "Strategy type: {strategy}\n"
+            "Target symbol: {symbol}\n\n"
+            "Requirements:\n"
+            "1. Full backtrader Strategy class with __init__, next, notify_order\n"
+            "2. Proper indicator initialization (use bt.indicators)\n"
+            "3. Entry/exit logic with clear conditions\n"
+            "4. Position sizing (percent sizer or fixed)\n"
+            "5. Risk management: stop-loss and take-profit\n"
+            "6. Logging via self.log()\n"
+            "7. Complete cerebro setup code at the bottom\n\n"
+            "Return ONLY the Python code wrapped in ```python``` fences. "
+            "Include inline comments explaining the logic."
+        ),
+        "tools_hint": ["recommend_strategy", "backtest_strategy"],
+    },
+    {
+        "command": "/gen-analysis",
+        "name": "Generate Analysis Script",
+        "category": "code",
+        "description": "Generate a Python analysis/visualization script",
+        "args": "<topic> [symbols...]",
+        "prompt": (
+            "Generate a Python script for financial analysis and visualization.\n"
+            "Topic: {topic}\n"
+            "Symbols: {symbols}\n\n"
+            "Requirements:\n"
+            "1. Use pandas, numpy, matplotlib/plotly, yfinance\n"
+            "2. Fetch real market data with yfinance\n"
+            "3. Compute relevant metrics/indicators\n"
+            "4. Create informative charts/plots\n"
+            "5. Print a summary table of key findings\n"
+            "6. Include error handling for data fetching\n\n"
+            "Return ONLY the Python code wrapped in ```python``` fences. "
+            "Include inline comments."
+        ),
+        "tools_hint": ["get_market_data", "calculate_factors"],
+    },
+    {
+        "command": "/gen-bot",
+        "name": "Generate Trading Bot",
+        "category": "code",
+        "description": "Generate live trading bot with exchange API (ccxt)",
+        "args": "<exchange> <strategy>",
+        "prompt": (
+            "Generate a Python trading bot for live execution.\n"
+            "Exchange: {exchange}\n"
+            "Strategy: {strategy}\n\n"
+            "Requirements:\n"
+            "1. Use ccxt library for exchange connection\n"
+            "2. Market data fetching and order execution\n"
+            "3. Signal generation based on the strategy logic\n"
+            "4. Risk management: max position size, daily loss limit\n"
+            "5. Logging with timestamps\n"
+            "6. Graceful shutdown handling (SIGINT)\n"
+            "7. Configuration via environment variables (API keys)\n"
+            "8. Paper trading mode toggle\n\n"
+            "Return ONLY the Python code wrapped in ```python``` fences. "
+            "NEVER include actual API keys. Use env vars."
+        ),
+        "tools_hint": ["recommend_strategy"],
+    },
+    {
+        "command": "/train-status",
+        "name": "Training Status",
+        "category": "tools",
+        "description": "Check Aria model training and data pipeline status",
+        "args": "",
+        "prompt": (
+            "Check the current Aria model training status.\n"
+            "1. Locate the project root via the ARIA_PROJECT_ROOT environment variable, or "
+            "search upward from the current directory for a 'packages/ml/llm/training' folder.\n"
+            "2. List checkpoint directories inside 'packages/ml/llm/training/outputs/' "
+            "(any subdirectory containing 'trainer_state.json').\n"
+            "3. Read the latest checkpoint's trainer_state.json: report current step, "
+            "total steps, epoch, eval_loss, and best_model_checkpoint.\n"
+            "4. Check for model_versions.json in the training outputs and report the "
+            "currently deployed version if present.\n"
+            "5. List recent training data files under 'data/training/' (newest 5 files).\n"
+            "Summarize: training progress (step/total, %), eval_loss trend, "
+            "deployed version, and data pipeline status."
+        ),
+        "tools_hint": ["read_file", "list_files"],
+    },
+]

aria_code/apps/cli/task_isolation.py

@@ -0,0 +1,211 @@
+"""The REPL's side of task worktrees: start one per task, offer to apply it.
+
+``runtime/task_worktree.py`` does the git work; this decides when a turn runs
+in a worktree, tells the executor where it is, and asks after a turn that
+changed files whether those changes should reach the user's own copy.
+
+Only the interactive REPL isolates. Headless ``-p`` runs keep editing in
+place: a script that asked for a change expects to find it on disk when the
+command exits, and there is nobody to ask.
+
+``task_isolation: off`` in the config, or ``ARIA_TASK_ISOLATION=off`` in the
+environment, turns it off.
+"""
+
+from __future__ import annotations
+
+import atexit
+import logging
+import os
+from pathlib import Path
+from typing import Any, Callable, Optional
+
+from aria_code.runtime.task_worktree import (
+    TaskWorktree,
+    TaskWorktreeError,
+    TaskWorktrees,
+    default_task_worktrees,
+)
+
+logger = logging.getLogger(__name__)
+
+_READ_ONLY_MODES = frozenset({"read-only", "readonly", "read_only", "plan"})
+
+_MANAGER: list[Optional[TaskWorktrees]] = [None]
+_CLEANUP_REGISTERED = [False]
+
+
+def manager() -> TaskWorktrees:
+    if _MANAGER[0] is None:
+        _MANAGER[0] = default_task_worktrees()
+    return _MANAGER[0]
+
+
+_OFF = {"off", "false", "none", "0"}
+
+
+def isolation_enabled(config: dict) -> bool:
+    setting = os.environ.get("ARIA_TASK_ISOLATION") or config.get("task_isolation", "worktree")
+    if str(setting or "").strip().lower() in _OFF:
+        return False
+    return str(config.get("permission_mode", "workspace-write") or "") not in _READ_ONLY_MODES
+
+
+def _workspace(config: dict) -> Path:
+    return Path(config.get("_session_workspace_root") or os.getcwd()).expanduser().resolve()
+
+
+def prepare(config: dict, *, notify: Callable[[str], None] = print) -> Optional[TaskWorktree]:
+    """The worktree this turn runs in, or None to edit in place.
+
+    Never raises: a repository git cannot snapshot falls back to editing in
+    place with a one-line note, as before this existed.
+    """
+    if not isolation_enabled(config):
+        return None
+    tasks = manager()
+    workspace = _workspace(config)
+    try:
+        before = tasks.active(workspace)
+        task = tasks.ensure(workspace)
+    except TaskWorktreeError as exc:
+        logger.debug("task worktree unavailable: %s", exc)
+        notify(f"Task isolation unavailable, editing your files directly: {exc}")
+        return None
+    if task is None:
+        return None
+    if before is None or before.task_id != task.task_id:
+        _register_cleanup()
+    return task
+
+
+def execution_context(task: Optional[TaskWorktree], config: dict) -> dict:
+    """What the executor needs to work in ``task`` instead of the workspace."""
+    if task is None:
+        return {}
+    workspace = _workspace(config)
+    origin = Path(task.repository)
+    try:
+        inside = workspace.relative_to(origin)
+    except ValueError:
+        inside = Path()
+    target = Path(task.path) / inside
+    # git does not carry empty or ignored directories; the session's own
+    # directory has to exist for commands to run in it.
+    target.mkdir(parents=True, exist_ok=True)
+    return {"_workspace": str(target), "_workspace_origin": str(origin), "_task_base": task.base}
+
+
+def summary_lines(task: TaskWorktree, changes) -> list[str]:
+    lines = [f"Task {task.task_id} changed {len(changes)} file{'s' if len(changes) != 1 else ''} "
+             "in its worktree; your files are unchanged."]
+    for change in changes[:12]:
+        lines.append(f"  {change.status} {change.path}")
+    if len(changes) > 12:
+        lines.append(f"  … and {len(changes) - 12} more")
+    return lines
+
+
+def apply(task: TaskWorktree, *, session_id: str = "") -> str:
+    applied = manager().apply(task, session_id=session_id)
+    return (f"Applied task {task.task_id}: {len(applied)} file{'s' if len(applied) != 1 else ''} "
+            "changed in your workspace; /rewind list shows how to undo it.")
+
+
+def discard(task: TaskWorktree) -> str:
+    manager().discard(task)
+    return f"Discarded task {task.task_id}; your files were not changed."
+
+
+APPLY, KEEP, DISCARD = 0, 1, 2
+
+
+def offer(task: Optional[TaskWorktree], *, choose: Callable[[list, str], int],
+          say: Callable[[str], None], session_id: str = "") -> Optional[int]:
+    """After a turn: if the task changed files, ask what to do with them.
+
+    ``choose(options, title)`` returns an index, or -1 when cancelled, which
+    keeps the task — nothing is applied or thrown away without an answer.
+    """
+    if task is None:
+        return None
+    try:
+        changes = manager().changes(task)
+    except TaskWorktreeError as exc:
+        logger.debug("task changes unavailable: %s", exc)
+        return None
+    if not changes:
+        return None
+    for line in summary_lines(task, changes):
+        say(line)
+    options = [
+        ("Apply to my files", "copy the changes into your workspace"),
+        ("Keep working", "the next message continues in this task"),
+        ("Discard", "drop the task's changes"),
+    ]
+    choice = choose(options, "Apply this task?")
+    if choice == APPLY:
+        try:
+            say(apply(task, session_id=session_id))
+        except TaskWorktreeError as exc:
+            say(str(exc))
+            return KEEP
+        return APPLY
+    if choice == DISCARD:
+        say(discard(task))
+        return DISCARD
+    say(f"Kept task {task.task_id}. /task apply or /task discard when you are ready.")
+    return KEEP
+
+
+def command(args: str, config: dict, *, session_id: str = "") -> str:
+    """``/task [status|diff|apply|discard]`` for the workspace's open task."""
+    sub = (args.strip().split() or ["status"])[0].lower()
+    tasks = manager()
+    try:
+        task = tasks.active(_workspace(config))
+    except TaskWorktreeError as exc:
+        return str(exc)
+    if task is None:
+        state = "on" if isolation_enabled(config) else "off (task_isolation: off)"
+        return f"No open task. Task isolation is {state}."
+    try:
+        if sub == "apply":
+            return apply(task, session_id=session_id)
+        if sub == "discard":
+            return discard(task)
+        if sub == "diff":
+            return tasks.diff(task) or "The task has no changes yet."
+        changes = tasks.changes(task)
+    except TaskWorktreeError as exc:
+        return str(exc)
+    if not changes:
+        return f"Task {task.task_id} has no changes yet.\n  worktree {task.path}"
+    lines = summary_lines(task, changes)
+    lines += [f"  worktree {task.path}", "/task diff · /task apply · /task discard"]
+    return "\n".join(lines)
+
+
+def _register_cleanup() -> None:
+    if _CLEANUP_REGISTERED[0]:
+        return
+    _CLEANUP_REGISTERED[0] = True
+    atexit.register(_discard_unchanged)
+
+
+def _discard_unchanged() -> None:
+    """At exit, drop open tasks with nothing in them; keep any with changes."""
+    tasks = _MANAGER[0]
+    if tasks is None:
+        return
+    for record in list(tasks._load().values()):
+        try:
+            task = TaskWorktree.from_dict(record)
+            if not Path(task.path).is_dir() or not tasks.changes(task):
+                tasks.discard(task)
+        except Exception:
+            continue
+
+
+def reset_for_tests(replacement: Any = None) -> None:
+    _MANAGER[0] = replacement