sigma-terminal 3.4.0__py3-none-any.whl → 3.5.0__py3-none-any.whl

sigma/app.py

@@ -1,1186 +1,445 @@
-"""Sigma v3.4.0 - Finance Research Agent."""
+"""Sigma v3.5.0 - Finance Research Agent."""
 
 import asyncio
-import os
-import re
 from datetime import datetime
-from typing import Optional, List
+from typing import Any, Optional
+
+import re  # Added top-level import
 
 from rich.markdown import Markdown
-from rich.panel import Panel
+from rich.markup import escape
 from rich.table import Table
 from rich.text import Text
+from rich.panel import Panel
+from rich.align import Align
+
 from textual import on, work
 from textual.app import App, ComposeResult
 from textual.binding import Binding
-from textual.containers import Container, Horizontal, Vertical, ScrollableContainer
-from textual.widgets import Footer, Input, RichLog, Static
-from textual.suggester import Suggester
-
-from .config import LLMProvider, get_settings, save_api_key, AVAILABLE_MODELS, SigmaError, ErrorCode
-from .llm import get_llm
-from .tools import TOOLS, execute_tool
-from .backtest import run_backtest, get_available_strategies, BACKTEST_TOOL
-
-
-__version__ = "3.4.0"
-SIGMA = "σ"
-
-# Common stock tickers for recognition
-COMMON_TICKERS = {
-    "AAPL", "MSFT", "GOOGL", "GOOG", "AMZN", "NVDA", "META", "TSLA", "BRK.A", "BRK.B",
-    "JPM", "JNJ", "V", "PG", "UNH", "HD", "MA", "DIS", "PYPL", "BAC", "ADBE", "NFLX",
-    "CRM", "INTC", "AMD", "CSCO", "PEP", "KO", "ABT", "NKE", "MRK", "PFE", "TMO",
-    "COST", "AVGO", "WMT", "ACN", "LLY", "MCD", "DHR", "TXN", "NEE", "PM", "HON",
-    "UPS", "BMY", "QCOM", "LOW", "MS", "RTX", "UNP", "ORCL", "IBM", "GE", "CAT",
-    "SBUX", "AMAT", "GS", "BLK", "DE", "AMT", "NOW", "ISRG", "LMT", "MDLZ", "AXP",
-    "SYK", "BKNG", "PLD", "GILD", "ADI", "TMUS", "CVS", "MMC", "ZTS", "CB", "C",
-    "SPY", "QQQ", "IWM", "DIA", "VTI", "VOO", "VXX", "ARKK", "XLF", "XLK", "XLE",
-}
-
-# Sigma animation frames - color cycling for thinking/processing state
-SIGMA_FRAMES = [
-    "[bold #3b82f6]s[/bold #3b82f6]",
-    "[bold #60a5fa]si[/bold #60a5fa]",
-    "[bold #93c5fd]sig[/bold #93c5fd]",
-    "[bold #bfdbfe]sigm[/bold #bfdbfe]",
-    "[bold white]sigma[/bold white]",
-    "[bold #bfdbfe]sigm[/bold #bfdbfe]",
-    "[bold #93c5fd]sig[/bold #93c5fd]",
-    "[bold #60a5fa]si[/bold #60a5fa]",
-    "[bold #3b82f6]s[/bold #3b82f6]",
-    "[bold cyan].[/bold cyan]",
-]
-
-# Sigma pulse animation (color breathing)
-SIGMA_PULSE_FRAMES = [
-    "[bold #1e40af]o[/bold #1e40af]",
-    "[bold #2563eb]o[/bold #2563eb]",
-    "[bold #3b82f6]o[/bold #3b82f6]",
-    "[bold #60a5fa]o[/bold #60a5fa]",
-    "[bold #93c5fd]o[/bold #93c5fd]",
-    "[bold #bfdbfe]o[/bold #bfdbfe]",
-    "[bold #93c5fd]o[/bold #93c5fd]",
-    "[bold #60a5fa]o[/bold #60a5fa]",
-    "[bold #3b82f6]o[/bold #3b82f6]",
-    "[bold #2563eb]o[/bold #2563eb]",
-]
-
-# Tool call spinner frames - ASCII based
-TOOL_SPINNER_FRAMES = [
-    "|", "/", "-", "\\"
-]
-
-# Welcome banner - clean design
-WELCOME_BANNER = """
-[bold #3b82f6]███████╗██╗ ██████╗ ███╗   ███╗ █████╗ [/bold #3b82f6]
-[bold #60a5fa]██╔════╝██║██╔════╝ ████╗ ████║██╔══██╗[/bold #60a5fa]
-[bold #93c5fd]███████╗██║██║  ███╗██╔████╔██║███████║[/bold #93c5fd]
-[bold #60a5fa]╚════██║██║██║   ██║██║╚██╔╝██║██╔══██║[/bold #60a5fa]
-[bold #3b82f6]███████║██║╚██████╔╝██║ ╚═╝ ██║██║  ██║[/bold #3b82f6]
-[bold #1d4ed8]╚══════╝╚═╝ ╚═════╝ ╚═╝     ╚═╝╚═╝  ╚═╝[/bold #1d4ed8]
-
-[bold cyan]Finance Research Agent[/bold cyan]  [dim]v3.4.0[/dim]
+from textual.containers import Vertical, Horizontal, Container
+from textual.screen import Screen
+from textual.widgets import Input, RichLog, Label, Static, LoadingIndicator
+from textual.reactive import reactive
+from textual.suggester import SuggestFromList
+
+from .config import get_settings
+from .core.command_router import CommandRouter, Request
+from .llm.router import get_router
+from .tools.registry import TOOL_REGISTRY
+
+__version__ = "3.5.0"
+
+# -----------------------------------------------------------------------------
+# CONSTANTS & ASSETS
+# -----------------------------------------------------------------------------
+
+SIGMA_ASCII = """
+[#d97757]  ██████  ██[/]  [#e08e79]██████  ███    ███  █████ [/]
+[#d97757]  ██      ██[/] [#e08e79]██       ████  ████ ██   ██[/]
+[#d97757]  ███████ ██[/] [#e08e79]██   ███ ██ ████ ██ ███████[/]
+[#d97757]       ██ ██[/] [#e08e79]██    ██ ██  ██  ██ ██   ██[/]
+[#d97757]  ██████  ██[/] [#e08e79] ██████  ██      ██ ██   ██[/]
 """
 
-SYSTEM_PROMPT = """You are Sigma (σ), an elite AI-powered Finance Research Agent. You combine the analytical rigor of a quantitative analyst, the market intuition of a seasoned portfolio manager, and the communication clarity of a top financial advisor.
-
-CORE CAPABILITIES:
-- Real-time market data analysis (quotes, charts, technicals) via yfinance and Polygon.io
-- Fundamental analysis (financials, ratios, earnings, valuations)
-- Technical analysis (RSI, MACD, Bollinger Bands, moving averages, support/resistance)
-- Backtesting strategies (SMA crossover, RSI, MACD, Bollinger, momentum, breakout)
-- Portfolio analysis and optimization
-- Sector and market overview with economic indicators
-- Insider trading and institutional activity tracking
-- Financial news search and SEC filings analysis
-
-RESPONSE PHILOSOPHY:
-1. BE PROACTIVE: Anticipate follow-up questions and address them
-2. BE THOROUGH: When analyzing, cover fundamentals + technicals + sentiment
-3. BE ACTIONABLE: Always end with clear recommendations or next steps
-4. BE DATA-DRIVEN: Cite specific numbers, dates, and percentages
-5. BE HONEST: Acknowledge uncertainty and risks
-
-RESPONSE FORMAT:
-- Start with a 1-2 sentence executive summary (the key takeaway)
-- Use structured sections with clear headers (## format)
-- Present comparative data in markdown tables
-- Highlight key metrics: **bold** for critical numbers
-- Use bullet points for easy scanning
-- End with "**Bottom Line:**" or "**Recommendation:**"
-
-RATING SYSTEM (when asked for recommendations):
-- STRONG BUY [A+]: Exceptional opportunity, high conviction
-- BUY [A]: Favorable outlook, solid fundamentals
-- HOLD [B]: Fair value, wait for better entry
-- SELL [C]: Concerns outweigh positives
-- STRONG SELL [D]: Significant risks, avoid
-
-DATA GATHERING RULES:
-- ALWAYS use tools to fetch current data before answering stock questions
-- Use multiple tools for comprehensive analysis (quote + technicals + fundamentals)
-- Cross-reference data points for accuracy
-- If a tool fails, try alternative approaches or acknowledge the limitation
-
-PROACTIVE INSIGHTS:
-When analyzing any stock, proactively mention:
-- Recent earnings surprises or upcoming earnings dates
-- Major analyst rating changes
-- Unusual volume or price movements
-- Relevant sector trends
-- Key support/resistance levels
-- Comparison to peers when relevant
-
-FORBIDDEN:
-- Never fabricate data or prices
-- Never provide data without fetching it first
-- Never give buy/sell advice without disclosure of risks
-- Never claim certainty about future performance
-
-Remember: Users trust you for professional-grade financial research. Exceed their expectations with every response."""
-
-# Enhanced autocomplete suggestions with more variety
 SUGGESTIONS = [
-    # Analysis commands
-    "analyze AAPL",
-    "analyze MSFT", 
-    "analyze GOOGL",
-    "analyze NVDA",
-    "analyze TSLA",
-    "analyze META",
-    "analyze AMZN",
-    "analyze AMD",
-    "analyze SPY",
-    # Comparisons
-    "compare AAPL MSFT GOOGL",
-    "compare NVDA AMD INTC",
-    "compare META GOOGL AMZN",
-    "compare TSLA RIVN LCID",
-    # Technical
-    "technical analysis of AAPL",
-    "technical analysis of SPY",
-    "technical analysis of NVDA",
-    "technical analysis of QQQ",
-    # Backtesting
-    "backtest SMA crossover on AAPL",
-    "backtest RSI strategy on SPY",
-    "backtest MACD on NVDA",
-    "backtest momentum on QQQ",
-    # Market
-    "market overview",
-    "sector performance",
-    "what sectors are hot today",
-    # Quotes
-    "get quote for AAPL",
-    "price of NVDA",
-    "how is TSLA doing",
-    # Fundamentals
-    "fundamentals of MSFT",
-    "financials for AAPL",
-    "earnings of NVDA",
-    # Activity
-    "insider trading for AAPL",
-    "institutional holders of NVDA",
-    "analyst recommendations for TSLA",
-    # Natural language queries
-    "what should I know about AAPL",
-    "is NVDA overvalued",
-    "best tech stocks right now",
-    "should I buy TSLA",
-    # Commands
-    "/help",
-    "/clear",
-    "/keys",
-    "/models",
-    "/status",
-    "/backtest",
-]
-
-# Extended action verbs for smart completion
-ACTION_VERBS = [
-    "analyze", "compare", "show", "get", "what is", "tell me about",
-    "technical analysis", "fundamentals", "price", "quote", "chart",
-    "backtest", "insider trading", "institutional", "analyst", "earnings",
-    "financials", "valuation", "news", "sector", "market", "portfolio"
+    "analyze", "backtest", "compare", "chart", "quote",
+    "AAPL", "MSFT", "GOOGL", "AMZN", "NVDA", "TSLA", "META",
+    "strategy", "momentum", "mean_reversion"
 ]
 
-# Ticker categories for smart suggestions
-TICKER_CATEGORIES = {
-    "tech": ["AAPL", "MSFT", "GOOGL", "AMZN", "META", "NVDA", "AMD", "INTC", "CRM", "ADBE"],
-    "finance": ["JPM", "BAC", "GS", "MS", "V", "MA", "BRK.B", "C", "WFC", "AXP"],
-    "healthcare": ["JNJ", "UNH", "PFE", "MRK", "ABBV", "LLY", "TMO", "ABT", "BMY", "GILD"],
-    "consumer": ["AMZN", "TSLA", "HD", "NKE", "MCD", "SBUX", "COST", "WMT", "TGT", "LOW"],
-    "energy": ["XOM", "CVX", "COP", "SLB", "EOG", "MPC", "PSX", "VLO", "OXY", "HAL"],
-    "etf": ["SPY", "QQQ", "IWM", "DIA", "VTI", "VOO", "VXX", "ARKK", "XLF", "XLK"],
-    "crypto": ["COIN", "MSTR", "RIOT", "MARA", "HUT"],
-    "ev": ["TSLA", "RIVN", "LCID", "NIO", "XPEV", "LI"],
-    "ai": ["NVDA", "AMD", "MSFT", "GOOGL", "META", "PLTR", "AI", "PATH", "SNOW"],
-    "semiconductor": ["NVDA", "AMD", "INTC", "AVGO", "QCOM", "TXN", "MU", "AMAT", "LRCX", "ASML"],
-}
 
+# -----------------------------------------------------------------------------
+# SCREENS
+# -----------------------------------------------------------------------------
 
-def extract_tickers(text: str) -> List[str]:
-    """Extract stock tickers from text."""
-    # Look for common patterns: $AAPL, or standalone uppercase words
-    # Only match if it's a known ticker or starts with $
-    words = text.upper().split()
-    tickers = []
+class ThinkingStatus(Static):
+    """Animated thinking indicator like Claude Code."""
     
-    for word in words:
-        # Clean the word
-        clean = word.strip('.,!?()[]{}":;')
-        
-        # Check for $TICKER format
-        if clean.startswith('$'):
-            ticker = clean[1:]
-            if ticker and ticker.isalpha() and len(ticker) <= 5:
-                tickers.append(ticker)
-        # Check if it's a known ticker
-        elif clean in COMMON_TICKERS:
-            tickers.append(clean)
-    
-    return list(dict.fromkeys(tickers))  # Dedupe while preserving order
-
-
-class SigmaSuggester(Suggester):
-    """Smart autocomplete with fuzzy matching and context awareness."""
-    
-    def __init__(self):
-        super().__init__(use_cache=False, case_sensitive=False)
-        self._build_suggestion_index()
-    
-    def _build_suggestion_index(self):
-        """Build an index of all possible suggestions for fast lookup."""
-        self.all_suggestions = []
-        
-        # Add static suggestions
-        self.all_suggestions.extend(SUGGESTIONS)
-        
-        # Generate dynamic suggestions for all tickers
-        for ticker in COMMON_TICKERS:
-            self.all_suggestions.extend([
-                f"analyze {ticker}",
-                f"technical analysis of {ticker}",
-                f"fundamentals of {ticker}",
-                f"price of {ticker}",
-                f"quote {ticker}",
-                f"insider trading {ticker}",
-                f"earnings {ticker}",
-                f"news {ticker}",
-            ])
-        
-        # Add category-based suggestions
-        for category, tickers in TICKER_CATEGORIES.items():
-            self.all_suggestions.append(f"best {category} stocks")
-            self.all_suggestions.append(f"{category} sector performance")
-            if len(tickers) >= 3:
-                self.all_suggestions.append(f"compare {' '.join(tickers[:3])}")
-    
-    def _fuzzy_match(self, query: str, target: str) -> float:
-        """Calculate fuzzy match score (0-1). Higher is better."""
-        query = query.lower()
-        target = target.lower()
-        
-        # Exact prefix match is best
-        if target.startswith(query):
-            return 1.0 + len(query) / len(target)
-        
-        # Check if all query chars appear in order
-        q_idx = 0
-        matches = 0
-        consecutive = 0
-        max_consecutive = 0
-        last_match = -1
-        
-        for t_idx, char in enumerate(target):
-            if q_idx < len(query) and char == query[q_idx]:
-                matches += 1
-                if last_match == t_idx - 1:
-                    consecutive += 1
-                    max_consecutive = max(max_consecutive, consecutive)
-                else:
-                    consecutive = 1
-                last_match = t_idx
-                q_idx += 1
-        
-        if q_idx < len(query):
-            # Not all chars matched
-            # Try substring match
-            if query in target:
-                return 0.5
-            return 0
-        
-        # Score based on match quality
-        score = (matches / len(query)) * 0.5 + (max_consecutive / len(query)) * 0.3
-        # Bonus for shorter targets (more relevant)
-        score += (1 - len(target) / 100) * 0.2
-        
-        return score
-    
-    def _get_context_suggestions(self, value: str) -> List[str]:
-        """Get context-aware suggestions based on partial input."""
-        suggestions = []
-        value_lower = value.lower().strip()
-        
-        # Command shortcuts
-        if value.startswith("/"):
-            cmd = value_lower[1:]
-            commands = ["/help", "/clear", "/keys", "/models", "/status", "/backtest", 
-                       "/provider", "/model", "/setkey", "/tickers"]
-            for c in commands:
-                if c[1:].startswith(cmd):
-                    suggestions.append(c)
-            return suggestions
-        
-        # Detect if user is typing a ticker
-        words = value.split()
-        last_word = words[-1] if words else ""
-        
-        if last_word.startswith("$") or (last_word.isupper() and len(last_word) >= 1):
-            ticker_prefix = last_word.lstrip("$").upper()
-            matching_tickers = [t for t in COMMON_TICKERS if t.startswith(ticker_prefix)]
-            
-            # If we have an action verb, complete with ticker
-            action_words = ["analyze", "compare", "technical", "price", "quote", 
-                           "fundamentals", "insider", "earnings", "news"]
-            prefix = " ".join(words[:-1]).lower() if len(words) > 1 else ""
-            
-            for ticker in matching_tickers[:5]:
-                if prefix:
-                    suggestions.append(f"{prefix} {ticker}")
-                else:
-                    suggestions.append(f"analyze {ticker}")
-            return suggestions
-        
-        # Natural language patterns
-        patterns = [
-            ("ana", "analyze"),
-            ("tech", "technical analysis of"),
-            ("fun", "fundamentals of"),
-            ("comp", "compare"),
-            ("back", "backtest"),
-            ("mark", "market overview"),
-            ("sect", "sector performance"),
-            ("pri", "price of"),
-            ("quo", "quote"),
-            ("ins", "insider trading"),
-            ("ear", "earnings"),
-            ("wha", "what should I know about"),
-            ("sho", "should I buy"),
-            ("is ", "is NVDA overvalued"),
-            ("how", "how is"),
-            ("bes", "best tech stocks"),
-        ]
-        
-        for prefix, expansion in patterns:
-            if value_lower.startswith(prefix):
-                if expansion.endswith("of") or expansion.endswith("about"):
-                    # Add popular tickers
-                    for ticker in ["AAPL", "NVDA", "MSFT", "TSLA", "GOOGL"]:
-                        suggestions.append(f"{expansion} {ticker}")
-                else:
-                    suggestions.append(expansion)
-        
-        return suggestions
+    DEFAULT_CSS = """
+    ThinkingStatus {
+        height: 1;
+        width: 100%;
+        color: #a1a1aa;
+        display: none;
+        padding-left: 2;
+    }
+    ThinkingStatus.visible {
+        display: block;
+    }
+    """
     
-    async def get_suggestion(self, value: str) -> Optional[str]:
-        """Get the best autocomplete suggestion."""
-        if not value or len(value) < 1:
-            return None
-        
-        value_lower = value.lower().strip()
-        
-        # Get context-aware suggestions first
-        context_suggestions = self._get_context_suggestions(value)
-        if context_suggestions:
-            return context_suggestions[0]
-        
-        # Fuzzy match against all suggestions
-        scored = []
-        for suggestion in self.all_suggestions:
-            score = self._fuzzy_match(value_lower, suggestion)
-            if score > 0:
-                scored.append((score, suggestion))
-        
-        # Sort by score and return best
-        scored.sort(key=lambda x: x[0], reverse=True)
-        
-        if scored:
-            return scored[0][1]
-        
-        # Last resort: if looks like a ticker, suggest analyze
-        if value.isupper() and len(value) <= 5 and value.isalpha():
-            return f"analyze {value}"
+    def on_mount(self):
+        self.loading = False
+        self.frame = 0
+        self.frames = ["σ", "o", "O", "0", "O", "o"]
+        self.set_interval(0.1, self.animate)
+        
+    def animate(self):
+        if self.loading:
+            self.frame = (self.frame + 1) % len(self.frames)
+            # symbol = self.frames[self.frame]
+            # Just pulse the sigma
+            if self.frame < 3:
+                self.update(Text("σ", style="bold #d97757") + " Analysis in progress...")
+            else:
+                self.update(Text("σ", style="dim #d97757") + " Analysis in progress...")
+                
+    def start(self):
+        self.loading = True
+        self.add_class("visible")
         
-        return None
-
-
-CSS = """
-Screen {
-    background: #0a0a0f;
-}
-
-* {
-    scrollbar-size: 1 1;
-    scrollbar-color: #3b82f6 30%;
-    scrollbar-color-hover: #60a5fa 50%;
-    scrollbar-color-active: #93c5fd 70%;
-}
-
-#main-container {
-    width: 100%;
-    height: 100%;
-    background: #0a0a0f;
-}
-
-#chat-area {
-    height: 1fr;
-    margin: 1 2;
-    background: #0a0a0f;
-}
-
-#chat-log {
-    background: #0a0a0f;
-    padding: 1 0;
-}
-
-#status-bar {
-    height: 3;
-    background: #0f1419;
-    border-top: solid #1e293b;
-    padding: 0 2;
-    dock: bottom;
-}
-
-#status-content {
-    width: 100%;
-    height: 100%;
-    content-align: left middle;
-}
-
-#thinking-indicator {
-    width: auto;
-    height: 1;
-    content-align: center middle;
-    display: none;
-}
-
-#thinking-indicator.visible {
-    display: block;
-}
-
-#tool-calls-display {
-    width: 100%;
-    height: auto;
-    max-height: 8;
-    background: #0f1419;
-    border: round #1e293b;
-    margin: 0 2;
-    padding: 0 1;
-    display: none;
-}
-
-#tool-calls-display.visible {
-    display: block;
-}
-
-#input-area {
-    height: 5;
-    padding: 1 2;
-    background: #0f1419;
-}
-
-#input-row {
-    height: 3;
-    width: 100%;
-}
-
-#sigma-indicator {
-    width: 4;
-    height: 3;
-    content-align: center middle;
-    background: transparent;
-}
-
-#prompt-input {
-    width: 1fr;
-    background: #1e293b;
-    border: tall #3b82f6;
-    color: #f8fafc;
-    padding: 0 1;
-}
-
-#prompt-input:focus {
-    border: tall #60a5fa;
-    background: #1e3a5f;
-}
-
-#prompt-input.-autocomplete {
-    border: tall #22c55e;
-}
-
-#ticker-highlight {
-    width: auto;
-    height: 1;
-    padding: 0 1;
-    background: transparent;
-    color: #22d3ee;
-}
-
-Footer {
-    background: #0d1117;
-    height: 1;
-    dock: bottom;
-}
+    def stop(self):
+        self.loading = False
+        self.remove_class("visible")
 
-Footer > .footer--highlight {
-    background: transparent;
-}
 
-Footer > .footer--key {
-    background: #1a1a2e;
-    color: #f59e0b;
-    text-style: bold;
-}
+class SplashScreen(Screen):
 
-Footer > .footer--description {
-    color: #6b7280;
-}
+    BINDINGS = [("enter", "start_app", "Start")]
 
-#help-panel {
-    width: 100%;
-    height: auto;
-    padding: 1;
-    background: #0d1117;
-    border: solid #3b82f6;
-    margin: 1 2;
-    display: none;
-}
+    def compose(self) -> ComposeResult:
+        yield Container(
+            Static(Panel(Align.center("[#e4e4e7]Welcome to Sigma[/]"), border_style="#d97757", padding=(0, 2)), id="welcome-badge"),
+            Static(Align.center(SIGMA_ASCII), id="ascii-art"),
+            Label("Press Enter to continue", id="press-enter"),
+            id="splash-container"
+        )
 
-#help-panel.visible {
-    display: block;
-}
-"""
+    def action_start_app(self):
+        self.app.switch_screen("main")
 
 
-class ToolCallDisplay(Static):
-    """Animated display for tool calls."""
-    
-    def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
-        self.tool_calls: List[dict] = []
-        self.frame = 0
-        self.timer = None
-    
-    def add_tool_call(self, name: str, status: str = "running"):
-        """Add a tool call to the display."""
-        self.tool_calls.append({"name": name, "status": status, "frame": 0})
-        self.add_class("visible")
-        self._render()
-        if not self.timer:
-            self.timer = self.set_interval(0.1, self._animate)
+class MainScreen(Screen):
+    """Main chat interface."""
     
-    def complete_tool_call(self, name: str):
-        """Mark a tool call as complete."""
-        for tc in self.tool_calls:
-            if tc["name"] == name and tc["status"] == "running":
-                tc["status"] = "complete"
-                break
-        self._render()
-    
-    def clear(self):
-        """Clear all tool calls."""
-        self.tool_calls = []
-        if self.timer:
-            self.timer.stop()
-            self.timer = None
-        self.remove_class("visible")
-        self.update("")
-    
-    def _animate(self):
-        """Animate the spinner."""
-        self.frame = (self.frame + 1) % len(TOOL_SPINNER_FRAMES)
-        for tc in self.tool_calls:
-            if tc["status"] == "running":
-                tc["frame"] = self.frame
-        self._render()
-    
-    def _render(self):
-        """Render the tool calls display."""
-        if not self.tool_calls:
-            return
-        
-        lines = []
-        for tc in self.tool_calls:
-            if tc["status"] == "running":
-                spinner = TOOL_SPINNER_FRAMES[tc["frame"] % len(TOOL_SPINNER_FRAMES)]
-                lines.append(f"  [cyan]{spinner}[/cyan] [bold]{tc['name']}[/bold] [dim]executing...[/dim]")
-            else:
-                lines.append(f"  [green][ok][/green] [bold]{tc['name']}[/bold] [green]done[/green]")
-        
-        self.update(Text.from_markup("\n".join(lines)))
+    BINDINGS = [
+        Binding("ctrl+l", "clear_chat", "Clear"),
+        Binding("ctrl+b", "toggle_sidebar", "Sidebar"),
+    ]
 
+    def compose(self) -> ComposeResult:
+        with Horizontal():
+            # Main Chat Area
+            with Vertical(id="chat-area"):
+                # Use a specific class or check if can_focus can be disabled here for RichLog
+                # In Textual, RichLog is focusable by default. We disable it to prevent it stealing focus from Input.
+                log = RichLog(id="chat-log", wrap=True, highlight=True, markup=True)
+                log.can_focus = False
+                yield log
+                yield ThinkingStatus(id="thinking")
+                yield Input(
+                    placeholder="Ask Sigma...", 
+                    id="prompt",
+                    suggester=SuggestFromList(SUGGESTIONS, case_sensitive=False)
+                )
+            
+            # Sidebar (Hidden by default)
+            with Vertical(id="sidebar"):
+                yield Label("TRACE", classes="sidebar-header")
+                tracelog = RichLog(id="trace-log", wrap=True, highlight=True, markup=True)
+                tracelog.can_focus = False
+                yield tracelog
 
-class SigmaIndicator(Static):
-    """Animated sigma indicator - typewriter style when active."""
-    
-    def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
-        self.active = False
-        self.mode = "idle"  # idle, thinking, tool
-        self.frame = 0
-        self.timer = None
-    
     def on_mount(self):
-        self.update(Text.from_markup(f"[bold blue]{SIGMA}[/bold blue]"))
-    
-    def set_active(self, active: bool, mode: str = "thinking"):
-        self.active = active
-        self.mode = mode if active else "idle"
-        if active and not self.timer:
-            self.frame = 0
-            interval = 0.08 if mode == "thinking" else 0.12
-            self.timer = self.set_interval(interval, self._animate)
-        elif not active and self.timer:
-            self.timer.stop()
-            self.timer = None
-            self.update(Text.from_markup(f"[bold blue]{SIGMA}[/bold blue]"))
-    
-    def _animate(self):
-        if self.mode == "thinking":
-            # Typewriter effect: s -> si -> sig -> sigm -> sigma -> sigm -> ...
-            self.frame = (self.frame + 1) % len(SIGMA_FRAMES)
-            self.update(Text.from_markup(SIGMA_FRAMES[self.frame]))
-        else:
-            # Pulse effect for tool calls
-            self.frame = (self.frame + 1) % len(SIGMA_PULSE_FRAMES)
-            self.update(Text.from_markup(SIGMA_PULSE_FRAMES[self.frame]))
-
-
-class TickerHighlight(Static):
-    """Display detected tickers in real-time."""
-    
-    def update_tickers(self, text: str):
-        """Update displayed tickers based on input."""
-        tickers = extract_tickers(text)
-        if tickers:
-            ticker_text = " ".join([f"[cyan]${t}[/cyan]" for t in tickers[:3]])
-            self.update(Text.from_markup(ticker_text))
-        else:
-            self.update("")
-
-
-class ChatLog(RichLog):
-    """Chat log with rich formatting."""
-    
-    def write_user(self, message: str):
-        # Highlight any tickers in user message
-        highlighted = message
-        for ticker in extract_tickers(message):
-            highlighted = re.sub(
-                rf'\b{ticker}\b',
-                f'[cyan]{ticker}[/cyan]',
-                highlighted,
-                flags=re.IGNORECASE
+        self.query_one("#prompt").focus()
+        # Initial greeting
+        self.query_one("#chat-log", RichLog).write(
+            Text.assemble(
+                ("σ ", "bold #d97757"),
+                ("Sigma initialized. ", "bold #e4e4e7"),
+                ("Ready for research.", "dim #a1a1aa")
             )
-        
-        self.write(Panel(
-            Text.from_markup(highlighted) if '[cyan]' in highlighted else Text(message, style="white"),
-            title="[bold blue]You[/bold blue]",
-            border_style="blue",
-            padding=(0, 1),
-        ))
-    
-    def write_assistant(self, message: str):
-        self.write(Panel(
-            Markdown(message),
-            title=f"[bold cyan]{SIGMA} Sigma[/bold cyan]",
-            border_style="cyan",
-            padding=(0, 1),
-        ))
-    
-    def write_tool(self, tool_name: str):
-        # This is now handled by ToolCallDisplay
-        pass
-    
-    def write_error(self, message: str):
-        self.write(Panel(Text(message, style="red"), title="[red]Error[/red]", border_style="red"))
-    
-    def write_system(self, message: str):
-        self.write(Text.from_markup(f"[dim]{message}[/dim]"))
-    
-    def write_welcome(self):
-        self.write(Text.from_markup(WELCOME_BANNER))
+        )
+
 
+# -----------------------------------------------------------------------------
+# MAIN APPLICATION
+# -----------------------------------------------------------------------------
 
 class SigmaApp(App):
-    """Sigma Finance Research Agent."""
-    
-    TITLE = "Sigma"
-    CSS = CSS
-    
-    BINDINGS = [
-        Binding("ctrl+l", "clear", "Clear"),
-        Binding("ctrl+m", "models", "Models"),
-        Binding("ctrl+h", "help_toggle", "Help"),
-        Binding("ctrl+p", "palette", "palette", show=True),
-        Binding("escape", "cancel", show=False),
-    ]
-    
-    def __init__(self):
-        super().__init__()
-        self.settings = get_settings()
-        self.llm = None
-        self.conversation = []
-        self.is_processing = False
-        self.history: List[str] = []
-        self.history_idx = -1
-        self.show_help = False
-    
-    def compose(self) -> ComposeResult:
-        yield Container(
-            ScrollableContainer(
-                ChatLog(id="chat-log", highlight=True, markup=True),
-                id="chat-area",
-            ),
-            ToolCallDisplay(id="tool-calls-display"),
-            Static(id="help-panel"),
-            Container(
-                Horizontal(
-                    SigmaIndicator(id="sigma-indicator"),
-                    Input(
-                        placeholder="Ask about any stock, market, or strategy... (Tab to autocomplete)",
-                        id="prompt-input",
-                        suggester=SigmaSuggester(),
-                    ),
-                    TickerHighlight(id="ticker-highlight"),
-                    id="input-row",
-                ),
-                id="input-area",
-            ),
-            id="main-container",
-        )
-        yield Footer()
-    
+    CSS = """
+    /* --- COLOR PALETTE (Claude Code Dark) --- */
+    $bg: #0f0f0f;       /* Very dark grey/black */
+    $surface: #1a1a1a;  /* Slightly lighter surface */
+    $text: #e4e4e7;     /* Off-white */
+    $dim: #a1a1aa;      /* Muted text */
+    $accent: #d97757;   /* Claude-like Orange/Peach */
+    $blue: #3b82f6;     /* Link/Action Blue */
+
+    Screen {
+        background: $bg;
+        color: $text;
+    }
+
+    /* --- SPLASH SCREEN --- */
+    #splash-container {
+        align: center middle;
+        height: 100%;
+    }
+
+    #welcome-badge {
+        width: auto;
+        margin-bottom: 2;
+        background: $bg;
+    }
+    
+    #ascii-art {
+        margin-bottom: 4;
+        color: $accent;
+    }
+
+    #press-enter {
+        color: $dim;
+        text-style: bold;
+    }
+
+    /* --- MAIN CHAT LAYOUT --- */
+    #chat-area {
+        width: 1fr;
+        height: 100%;
+        padding: 1 2;
+    }
+
+    #chat-log {
+        width: 100%;
+        height: 1fr;
+        background: $bg;
+        border: none;
+        margin-bottom: 1;
+    }
+
+    Input {
+        width: 100%;
+        height: 3;
+        background: $surface;
+        border: solid $dim;
+        color: $text;
+        padding: 0 1; 
+    }
+    
+    Input:focus {
+        border: solid $accent;
+        background: $surface;
+    }
+    
+    Input .suggestion {
+        color: $dim;
+    }
+
+    /* --- SIDEBAR --- */
+    #sidebar {
+        width: 40;
+        dock: right;
+        background: $surface;
+        border-left: solid $dim 20%;
+        display: none;
+    }
+
+    #sidebar.visible {
+        display: block;
+    }
+
+    .sidebar-header {
+        background: $surface;
+        color: $dim;
+        text-align: center;
+        text-style: bold;
+        padding: 1;
+        border-bottom: solid $dim 20%;
+    }
+    
+    #trace-log {
+        background: $surface;
+        padding: 1;
+    }
+    """
+
+    SCREENS = {
+        "splash": SplashScreen,
+        "main": MainScreen
+    }
+
     def on_mount(self):
-        chat = self.query_one("#chat-log", ChatLog)
-        chat.write_welcome()
+        # Initialize Core Logic
+        self.router = CommandRouter()
+        try:
+            self.llm_router = get_router(get_settings())
+        except Exception:
+            self.llm_router = None
+            
+        self.push_screen("splash")
+
+    @on(Input.Submitted)
+    async def on_input(self, event: Input.Submitted):
+        query = event.value
+        if not query.strip(): return
         
-        provider = getattr(self.settings.default_provider, 'value', str(self.settings.default_provider))
-        chat.write_system(f"{SIGMA} Provider: [bold]{provider}[/bold] | Model: [bold]{self.settings.default_model}[/bold]")
-        chat.write_system(f"{SIGMA} Type [cyan]/help[/cyan] for commands • [cyan]/keys[/cyan] to set up API keys")
-        chat.write_system("")
+        # Clear input
+        event.input.value = ""
         
-        self._init_llm()
-        self.query_one("#prompt-input", Input).focus()
-    
-    def _init_llm(self):
-        """Initialize the LLM client with proper error handling."""
+        # Get chat log - relative to the screen containing the input
+        # This fixes the issue where self.query_one fails on App or wrong screen
         try:
-            self.llm = get_llm(self.settings.default_provider, self.settings.default_model)
-        except SigmaError as e:
-            chat = self.query_one("#chat-log", ChatLog)
-            chat.write_error(f"[E{e.code}] {e.message}")
-            if e.details.get("hint"):
-                chat.write_system(f"[dim]Hint: {e.details['hint']}[/dim]")
-            self.llm = None
+            chat_log = event.control.screen.query_one("#chat-log", RichLog)
         except Exception as e:
-            chat = self.query_one("#chat-log", ChatLog)
-            chat.write_error(f"[E{ErrorCode.PROVIDER_ERROR}] Failed to initialize: {str(e)[:100]}")
-            chat.write_system("[dim]Use /keys to configure API keys[/dim]")
-            self.llm = None
-    
-    @on(Input.Changed)
-    def on_input_change(self, event: Input.Changed):
-        """Update ticker highlight as user types."""
-        ticker_display = self.query_one("#ticker-highlight", TickerHighlight)
-        ticker_display.update_tickers(event.value)
-    
-    @on(Input.Submitted)
-    def handle_input(self, event: Input.Submitted):
-        if self.is_processing:
+            self.notify(f"UI Error: Could not find chat log on active screen. {e}", severity="error")
             return
-        
-        text = event.value.strip()
-        if not text:
-            return
-        
-        self.query_one("#prompt-input", Input).value = ""
-        self.history.append(text)
-        self.history_idx = len(self.history)
-        
-        chat = self.query_one("#chat-log", ChatLog)
-        
-        if text.startswith("/"):
-            self._handle_command(text, chat)
-        else:
-            chat.write_user(text)
-            self._process_query(text, chat)
-    
-    def _handle_command(self, cmd: str, chat: ChatLog):
-        parts = cmd.lower().split()
-        command = parts[0]
-        args = parts[1:] if len(parts) > 1 else []
-        
-        if command == "/help":
-            self._show_comprehensive_help(chat)
-        elif command == "/clear":
-            chat.clear()
-            self.conversation = []
-            chat.write_system("Chat cleared")
-        elif command == "/keys":
-            self._show_keys(chat)
-        elif command == "/models":
-            self._show_models(chat)
-        elif command == "/status":
-            self._show_status(chat)
-        elif command == "/backtest":
-            self._show_strategies(chat)
-        elif command == "/provider" and args:
-            self._switch_provider(args[0], chat)
-        elif command == "/model" and args:
-            self._switch_model(args[0], chat)
-        elif command.startswith("/setkey") and len(parts) >= 3:
-            self._set_key(parts[1], parts[2], chat)
-        elif command == "/tickers":
-            self._show_popular_tickers(chat)
-        else:
-            chat.write_error(f"Unknown command: {command}. Type /help for available commands.")
-    
-    def _show_comprehensive_help(self, chat: ChatLog):
-        """Show comprehensive help with examples."""
-        help_text = f"""
-[bold cyan]═══════════════════════════════════════════════════════════════[/bold cyan]
-[bold]                    {SIGMA} SIGMA HELP CENTER                      [/bold]
-[bold cyan]═══════════════════════════════════════════════════════════════[/bold cyan]
-
-[bold yellow]QUICK START[/bold yellow]
-  Just type naturally! Examples:
-  • "analyze AAPL"           - Full analysis of Apple
-  • "compare NVDA AMD INTC"  - Compare multiple stocks
-  • "is TSLA overvalued?"    - Get AI insights
-  • "market overview"        - See major indices
-
-[bold yellow]COMMANDS[/bold yellow]
-  [cyan]/help[/cyan]              This help screen
-  [cyan]/clear[/cyan]             Clear chat history
-  [cyan]/keys[/cyan]              Configure API keys
-  [cyan]/models[/cyan]            Show available models
-  [cyan]/status[/cyan]            Current configuration
-  [cyan]/backtest[/cyan]          Show backtest strategies
-  [cyan]/provider <name>[/cyan]   Switch AI provider
-  [cyan]/model <name>[/cyan]      Switch model
-  [cyan]/setkey <p> <k>[/cyan]    Set API key
-  [cyan]/tickers[/cyan]           Popular tickers list
-
-[bold yellow]ANALYSIS EXAMPLES[/bold yellow]
-  • "technical analysis of SPY"
-  • "fundamentals of MSFT"
-  • "insider trading for AAPL"
-  • "analyst recommendations for NVDA"
-  • "sector performance"
-
-[bold yellow]BACKTESTING[/bold yellow]
-  • "backtest SMA crossover on AAPL"
-  • "backtest RSI strategy on SPY"
-  • "backtest MACD on NVDA"
-  Strategies: sma_crossover, rsi, macd, bollinger, momentum, breakout
-
-[bold yellow]KEYBOARD SHORTCUTS[/bold yellow]
-  [bold]Tab[/bold]      Autocomplete suggestion
-  [bold]Ctrl+L[/bold]   Clear chat
-  [bold]Ctrl+M[/bold]   Show models
-  [bold]Ctrl+H[/bold]   Toggle quick help
-  [bold]Ctrl+P[/bold]   Command palette
-  [bold]Esc[/bold]      Cancel operation
-
-[bold yellow]TIPS[/bold yellow]
-  • Type [cyan]$AAPL[/cyan] or [cyan]AAPL[/cyan] - tickers auto-detected
-  • Use Tab for smart autocomplete
-  • Detected tickers shown next to input
-"""
-        chat.write(Panel(
-            Text.from_markup(help_text),
-            title=f"[bold cyan]{SIGMA} Help[/bold cyan]",
-            border_style="cyan",
-            padding=(0, 1),
-        ))
-    
-    def _show_popular_tickers(self, chat: ChatLog):
-        """Show popular tickers organized by category."""
-        tickers_text = """
-[bold]Tech Giants[/bold]: AAPL, MSFT, GOOGL, AMZN, META, NVDA
-[bold]Semiconductors[/bold]: NVDA, AMD, INTC, AVGO, QCOM, TSM
-[bold]EVs & Auto[/bold]: TSLA, RIVN, LCID, F, GM
-[bold]Finance[/bold]: JPM, BAC, GS, MS, V, MA
-[bold]Healthcare[/bold]: JNJ, PFE, UNH, MRK, ABBV
-[bold]ETFs[/bold]: SPY, QQQ, IWM, DIA, VTI, VOO
-[bold]Sector ETFs[/bold]: XLK, XLF, XLE, XLV, XLI
-"""
-        chat.write(Panel(
-            Text.from_markup(tickers_text),
-            title=f"[cyan]{SIGMA} Popular Tickers[/cyan]",
-            border_style="dim",
-        ))
-    
-    def _show_keys(self, chat: ChatLog):
-        """Show comprehensive API key management interface."""
-        keys_help = f"""
-[bold cyan]━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[/bold cyan]
-[bold]                    {SIGMA} API KEY MANAGER                     [/bold]
-[bold cyan]━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━[/bold cyan]
-
-[bold yellow]QUICK SETUP[/bold yellow]
-  Set a key:  [cyan]/setkey <provider> <your-api-key>[/cyan]
-  
-[bold yellow]LLM PROVIDERS[/bold yellow]
-  [bold]google[/bold]      → https://aistudio.google.com/apikey
-  [bold]openai[/bold]      → https://platform.openai.com/api-keys
-  [bold]anthropic[/bold]   → https://console.anthropic.com/settings/keys
-  [bold]groq[/bold]        → https://console.groq.com/keys [dim](free!)[/dim]
-  [bold]xai[/bold]         → https://console.x.ai
-
-[bold yellow]DATA PROVIDERS[/bold yellow] [dim](optional - enhances data quality)[/dim]
-  [bold]polygon[/bold]     → https://polygon.io/dashboard/api-keys
 
-[bold yellow]EXAMPLES[/bold yellow]
-  /setkey google AIzaSyB...
-  /setkey openai sk-proj-...
-  /setkey polygon abc123...
-  /provider groq          [dim]← switch to groq[/dim]
-
-[bold yellow]TIPS[/bold yellow]
-  • [green]Groq[/green] is free and fast - great for starting out
-  • [green]Ollama[/green] runs locally - no key needed (/provider ollama)
-  • Keys are stored in [dim]~/.sigma/config.env[/dim]
-"""
-        chat.write(Panel(
-            Text.from_markup(keys_help),
-            title=f"[bold cyan]{SIGMA} API Keys[/bold cyan]",
-            border_style="cyan",
-            padding=(0, 1),
-        ))
-        self._show_status(chat)
-    
-    def _show_status(self, chat: ChatLog):
-        """Show current configuration status."""
-        table = Table(show_header=True, box=None, padding=(0, 2), title=f"{SIGMA} Current Status")
-        table.add_column("Setting", style="bold")
-        table.add_column("Value")
-        table.add_column("Status")
-        
-        provider = getattr(self.settings.default_provider, 'value', str(self.settings.default_provider))
-        table.add_row("Provider", provider, "[green][*][/green] Active")
-        table.add_row("Model", self.settings.default_model, "")
-        table.add_row("", "", "")
-        
-        # LLM Keys
-        llm_keys = [
-            ("Google", self.settings.google_api_key, "google"),
-            ("OpenAI", self.settings.openai_api_key, "openai"),
-            ("Anthropic", self.settings.anthropic_api_key, "anthropic"),
-            ("Groq", self.settings.groq_api_key, "groq"),
-            ("xAI", self.settings.xai_api_key, "xai"),
-        ]
-        for name, key, prov in llm_keys:
-            if key:
-                masked = key[:8] + "..." + key[-4:] if len(key) > 12 else "***"
-                status = "[green][ok][/green]"
-            else:
-                masked = "[dim]not set[/dim]"
-                status = "[dim]--[/dim]"
-            
-            # Highlight active provider
-            if prov == provider:
-                name = f"[bold cyan]{name}[/bold cyan]"
-            table.add_row(f"  {name}", Text.from_markup(masked), Text.from_markup(status))
-        
-        table.add_row("", "", "")
-        
-        # Data Keys
-        polygon_key = getattr(self.settings, 'polygon_api_key', None)
-        if polygon_key:
-            table.add_row("  Polygon", polygon_key[:8] + "...", Text.from_markup("[green][ok][/green]"))
-        else:
-            table.add_row("  Polygon", Text.from_markup("[dim]not set[/dim]"), Text.from_markup("[dim]optional[/dim]"))
+        # Prepare display text
+        display_query = escape(query)
         
-        chat.write(Panel(table, border_style="dim"))
-    
-    def _show_models(self, chat: ChatLog):
-        table = Table(title=f"{SIGMA} Models", show_header=True, border_style="dim")
-        table.add_column("Provider", style="cyan")
-        table.add_column("Models")
-        for p, m in AVAILABLE_MODELS.items():
-            table.add_row(p, ", ".join(m))
-        chat.write(table)
-    
-    def _show_strategies(self, chat: ChatLog):
-        strategies = get_available_strategies()
-        table = Table(title=f"{SIGMA} Strategies", show_header=True, border_style="dim")
-        table.add_column("Name", style="cyan")
-        table.add_column("Description")
-        for k, v in strategies.items():
-            table.add_row(k, v.get('description', ''))
-        chat.write(table)
-    
-    def _switch_provider(self, provider: str, chat: ChatLog):
-        valid = ["google", "openai", "anthropic", "groq", "xai", "ollama"]
-        if provider not in valid:
-            chat.write_error(f"Invalid. Use: {', '.join(valid)}")
-            return
+        # Highlight tickers
         try:
-            self.settings.default_provider = LLMProvider(provider)
-            if provider in AVAILABLE_MODELS:
-                self.settings.default_model = AVAILABLE_MODELS[provider][0]
-            self._init_llm()
-            chat.write_system(f"Switched to {provider}")
+            req = self.router.parse(query)
+            for ticker in req.tickers:
+                escaped_ticker = re.escape(ticker)
+                # Apply green highlight
+                display_query = re.sub(
+                    f"(?i)\\b{escaped_ticker}\\b", 
+                    f"[bold #22c55e]{ticker.upper()}[/]", 
+                    display_query
+                )
+        except Exception:
+            pass
+
+        # Write to log
+        try:
+            chat_log.write(
+                Text.assemble(
+                    ("❯ ", "bold #d97757"),
+                    Text.from_markup(display_query)
+                )
+            )
+            chat_log.write("") # Add spacing
         except Exception as e:
-            chat.write_error(str(e))
-    
-    def _switch_model(self, model: str, chat: ChatLog):
-        self.settings.default_model = model
-        self._init_llm()
-        chat.write_system(f"Model: {model}")
-    
-    def _set_key(self, provider: str, key: str, chat: ChatLog):
-        """Save an API key for a provider."""
-        # Normalize provider name
-        provider = provider.lower().strip()
-        valid_providers = ["google", "openai", "anthropic", "groq", "xai", "polygon", "alphavantage", "exa"]
+            self.notify(f"Display Error: {e}", severity="error")
+            chat_log.write(f"❯ {query}")
         
-        if provider not in valid_providers:
-            chat.write_error(f"[E{ErrorCode.INVALID_INPUT}] Unknown provider: {provider}")
-            chat.write_system(f"Valid providers: {', '.join(valid_providers)}")
-            return
+        # Pass the specific chat_log instance to the worker to avoid re-querying failure
+        self.run_query(query, chat_log)
+
+    @work
+    async def run_query(self, query: str, chat_log: RichLog):
+        # Trace log logic simplified
+        trace_log = None
+        try:
+             # Try to find trace log on the same screen as chat_log
+             trace_log = chat_log.screen.query_one("#trace-log", RichLog)
+        except:
+             pass
         
-        # Basic key validation
-        key = key.strip()
-        if len(key) < 10:
-            chat.write_error(f"[E{ErrorCode.INVALID_INPUT}] API key seems too short")
-            return
+        # UI Animation: Thinking
+        try:
+            thinker = chat_log.screen.query_one(ThinkingStatus)
+            thinker.start()
+        except:
+            thinker = None
         
         try:
-            success = save_api_key(provider, key)
-            if success:
-                # Reload settings
-                self.settings = get_settings()
-                
-                # Show success with masked key
-                masked = key[:6] + "..." + key[-4:] if len(key) > 10 else "***"
-                chat.write_system(f"[green][ok][/green] {SIGMA} Key saved for [bold]{provider}[/bold]: {masked}")
-                
-                # Auto-switch to this provider if it's an LLM provider and we don't have an LLM
-                llm_providers = ["google", "openai", "anthropic", "groq", "xai"]
-                if provider in llm_providers:
-                    if not self.llm or provider == getattr(self.settings.default_provider, 'value', ''):
-                        self._switch_provider(provider, chat)
+            req = self.router.parse(query)
+            if trace_log:
+                trace_log.write(f"[dim]Action:[/dim] {req.action}")
+
+            if req.is_command:
+                await self.handle_command(req, chat_log, trace_log)
             else:
-                chat.write_error(f"[E{ErrorCode.UNKNOWN_ERROR}] Failed to save key")
+                await self.handle_chat(req, chat_log, trace_log)
+
         except Exception as e:
-            chat.write_error(f"[E{ErrorCode.UNKNOWN_ERROR}] {str(e)}")
-    
-    @work(exclusive=True)
-    async def _process_query(self, query: str, chat: ChatLog):
-        if not self.llm:
-            chat.write_error(f"[E{ErrorCode.API_KEY_MISSING}] No LLM configured. Use /keys to set up.")
+            chat_log.write(f"[red]Error: {e}[/red]")
+            
+        finally:
+            if thinker:
+                thinker.stop()
+
+    async def handle_command(self, req: Request, chat_log: RichLog, trace_log: Optional[RichLog]):
+        if req.action == "backtest":
+            chat_log.write("[dim]Running backtest...[/dim]")
+            try:
+                 symbol = req.tickers[0] if req.tickers else "SPY"
+                 strategy = req.details.get("strategy") or "momentum"
+                 from .backtest import run_backtest
+                 
+                 result = await asyncio.to_thread(run_backtest, symbol, strategy, "1y")
+                 
+                 if "error" in result:
+                     chat_log.write(f"[red]Failed: {result['error']}[/red]")
+                 else:
+                     perf = result.get("performance", {})
+                     # Minimalist Table
+                     table = Table(box=None, show_header=False, padding=(0, 2))
+                     table.add_row("[bold]Return[/]", f"[green]{perf.get('total_return')}[/]")
+                     table.add_row("[bold]Sharpe[/]", f"{result.get('risk', {}).get('sharpe_ratio')}")
+                     table.add_row("[bold]Equity[/]", f"{perf.get('final_equity')}")
+                     
+                     chat_log.write(Panel(table, title=f"Backtest: {symbol}", border_style="#d97757"))
+            except Exception as e:
+                chat_log.write(f"[red]{e}[/red]")
+
+    async def handle_chat(self, req: Request, chat_log: RichLog, trace_log: Optional[RichLog]):
+        if not self.llm_router:
+            chat_log.write("[red]Setup required.[/red]")
             return
+
+        chat_log.write("") # break
+        
+        # System prompt to enforce thought signatures for Hack Club/Gemini compatibility
+        system_prompt = (
+            "You are Sigma, an advanced financial research assistant. "
+            "You are powered by various LLMs and have access to real-time financial tools. "
+            "You MUST use available tools for any question regarding stock prices, financials, market news, or analysis. "
+            "Never hallucinate financial data. Always fetch it using the provided tools. "
+            "CRITICAL INSTRUCTION: When calling any tool, you MUST provide the 'thought_signature' parameter. "
+            "This parameter should contain your internal reasoning for why you are calling that specific tool. "
+            "Do not output the thought signature in your final response, only use it in the tool call. "
+            "Format your final response in clean Markdown."
+        )
         
-        self.is_processing = True
-        indicator = self.query_one("#sigma-indicator", SigmaIndicator)
-        tool_display = self.query_one("#tool-calls-display", ToolCallDisplay)
-        ticker_highlight = self.query_one("#ticker-highlight", TickerHighlight)
-        
-        # Clear ticker highlight and start sigma animation
-        ticker_highlight.update("")
-        indicator.set_active(True, mode="thinking")
+        messages = [
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": req.original_query}
+        ]
         
         try:
-            messages = [{"role": "system", "content": SYSTEM_PROMPT}]
-            messages.extend(self.conversation)
-            messages.append({"role": "user", "content": query})
-            
-            all_tools = TOOLS + [BACKTEST_TOOL]
-            
-            async def on_tool(name: str, args: dict):
-                tool_display.add_tool_call(name)
-                if name == "run_backtest":
-                    result = run_backtest(**args)
-                else:
-                    result = execute_tool(name, args)
-                tool_display.complete_tool_call(name)
-                return result
-            
-            response = await self.llm.generate(messages, tools=all_tools, on_tool_call=on_tool)
-            
-            # Clear tool display after getting response
-            await asyncio.sleep(0.5)  # Brief pause to show completion
-            tool_display.clear()
+            stream = await self.llm_router.chat(
+                messages,
+                tools=TOOL_REGISTRY.to_llm_format(),
+                on_tool_call=self.mk_tool_callback(trace_log)
+            )
             
-            if response:
-                chat.write_assistant(response)
-                self.conversation.append({"role": "user", "content": query})
-                self.conversation.append({"role": "assistant", "content": response})
-                if len(self.conversation) > 20:
-                    self.conversation = self.conversation[-20:]
+            if isinstance(stream, str):
+                chat_log.write(Markdown(stream))
             else:
-                chat.write_error(f"[E{ErrorCode.RESPONSE_INVALID}] No response received")
+                acc = ""
+                async for chunk in stream:
+                    acc += chunk
+                chat_log.write(Markdown(acc))
                 
-        except SigmaError as e:
-            tool_display.clear()
-            # Format SigmaError nicely
-            chat.write_error(f"[E{e.code}] {e.message}")
-            if e.details.get("hint"):
-                chat.write_system(f"[dim]Hint: {e.details['hint']}[/dim]")
         except Exception as e:
-            tool_display.clear()
-            # Try to parse common API errors
-            error_str = str(e)
-            if "401" in error_str or "invalid" in error_str.lower() and "key" in error_str.lower():
-                chat.write_error(f"[E{ErrorCode.API_KEY_INVALID}] API key is invalid. Use /keys to update.")
-            elif "429" in error_str or "rate" in error_str.lower():
-                chat.write_error(f"[E{ErrorCode.API_KEY_RATE_LIMITED}] Rate limit hit. Wait a moment and try again.")
-            else:
-                chat.write_error(f"[E{ErrorCode.PROVIDER_ERROR}] {error_str[:200]}")
-        finally:
-            indicator.set_active(False)
-            self.is_processing = False
-            self.query_one("#prompt-input", Input).focus()
-    
-    def action_clear(self):
-        chat = self.query_one("#chat-log", ChatLog)
-        chat.clear()
-        self.conversation = []
-        chat.write_system("Cleared")
-    
-    def action_models(self):
-        self._show_models(self.query_one("#chat-log", ChatLog))
-    
-    def action_help_toggle(self):
-        """Toggle quick help panel."""
-        help_panel = self.query_one("#help-panel", Static)
-        if self.show_help:
-            help_panel.remove_class("visible")
-            help_panel.update("")
-        else:
-            help_panel.add_class("visible")
-            help_panel.update(Text.from_markup(
-                "[bold]Quick Commands:[/bold] /help /clear /keys /models /status /backtest  "
-                "[bold]Shortcuts:[/bold] Tab=autocomplete Ctrl+L=clear Ctrl+M=models"
-            ))
-        self.show_help = not self.show_help
-    
-    def action_cancel(self):
-        if self.is_processing:
-            self.is_processing = False
-            tool_display = self.query_one("#tool-calls-display", ToolCallDisplay)
-            tool_display.clear()
-
+            chat_log.write(f"[red]LLM Error: {e}[/red]")
+
+    def mk_tool_callback(self, trace_log: Optional[RichLog]):
+        async def callback(name: str, args: dict):
+            if trace_log:
+                trace_log.write(f"[yellow]➜ {name}[/yellow]")
+            try:
+                res = await TOOL_REGISTRY.execute(name, args)
+                if trace_log:
+                    trace_log.write(f"[green]✔[/green]")
+                return res
+            except Exception as e:
+                if trace_log:
+                    trace_log.write(f"[red]✖ {e}[/red]")
+                raise
+        return callback
+
+    def action_toggle_sidebar(self):
+        try:
+            sidebar = self.query_one("#sidebar")
+            sidebar.set_class(not sidebar.has_class("visible"), "visible")
+        except:
+            pass
 
 def launch():
-    """Launch Sigma."""
-    SigmaApp().run()
-
-
-if __name__ == "__main__":
-    launch()
+    app = SigmaApp()
+    app.run()