Merge b950e3a018 into 13b826a31d

.env.example

@@ -1,2 +1,17 @@
+# Data vendor API keys
 ALPHA_VANTAGE_API_KEY=alpha_vantage_api_key_placeholder
-OPENAI_API_KEY=openai_api_key_placeholder
+
+# LLM Provider API keys (set the ones you want to use)
+OPENAI_API_KEY=openai_api_key_placeholder
+ANTHROPIC_API_KEY=anthropic_api_key_placeholder
+GEMINI_API_KEY=gemini_api_key_placeholder
+OPENROUTER_API_KEY=openrouter_api_key_placeholder
+
+# Local LLM provider URLs (optional, defaults shown)
+# OLLAMA_URL=http://localhost:11434
+# LM_STUDIO_URL=http://localhost:1234
+
+# Feature flags
+# Set to "true" to fetch latest models from APIs and use latest web_search tool
+# Set to "false" or leave unset for static model lists and web_search_preview (legacy)
+FETCH_LATEST=true

cli/api_keys.py

@@ -0,0 +1,100 @@
+"""API key and endpoint validation for LLM providers."""
+
+import os
+from typing import Optional, Tuple
+import httpx
+
+
+# Map cloud providers to their required environment variables
+PROVIDER_API_KEYS = {
+    "openai": "OPENAI_API_KEY",
+    "anthropic": "ANTHROPIC_API_KEY",
+    "google": "GEMINI_API_KEY",
+    "openrouter": "OPENROUTER_API_KEY",
+}
+
+# Default endpoints for local providers
+LOCAL_PROVIDER_DEFAULTS = {
+    "ollama": ("OLLAMA_URL", "http://localhost:11434"),
+    "lm studio": ("LM_STUDIO_URL", "http://localhost:1234"),
+}
+
+
+def get_api_key(provider: str) -> Optional[str]:
+    """Get API key for a cloud provider, returns None if not set."""
+    provider_lower = provider.lower()
+
+    # Special case: OpenRouter can use OPENROUTER_API_KEY or OPENAI_API_KEY with sk-or- prefix
+    if provider_lower == "openrouter":
+        openrouter_key = os.getenv("OPENROUTER_API_KEY")
+        if openrouter_key:
+            return openrouter_key
+        # Check if OPENAI_API_KEY is actually an OpenRouter key
+        openai_key = os.getenv("OPENAI_API_KEY", "")
+        if openai_key.startswith("sk-or-"):
+            return openai_key
+        return None
+
+    env_var = PROVIDER_API_KEYS.get(provider_lower)
+    if env_var is None:
+        return None
+    return os.getenv(env_var)
+
+
+def get_local_endpoint(provider: str) -> Optional[str]:
+    """Get the endpoint URL for a local provider."""
+    provider_lower = provider.lower()
+    if provider_lower not in LOCAL_PROVIDER_DEFAULTS:
+        return None
+
+    env_var, default_url = LOCAL_PROVIDER_DEFAULTS[provider_lower]
+    return os.getenv(env_var, default_url)
+
+
+def is_local_provider_running(provider: str) -> bool:
+    """Check if a local provider (Ollama/LM Studio) is running by probing its endpoint."""
+    endpoint = get_local_endpoint(provider)
+    if not endpoint:
+        return False
+
+    try:
+        # Probe the models endpoint with a short timeout
+        response = httpx.get(
+            f"{endpoint}/v1/models",
+            timeout=1.0
+        )
+        return response.status_code == 200
+    except (httpx.RequestError, httpx.TimeoutException):
+        return False
+
+
+def is_provider_available(provider: str) -> Tuple[bool, str]:
+    """
+    Check if a provider is available.
+
+    Returns:
+        Tuple of (is_available, reason_if_unavailable)
+    """
+    provider_lower = provider.lower()
+
+    # Local providers: check if endpoint is reachable
+    if provider_lower in LOCAL_PROVIDER_DEFAULTS:
+        if is_local_provider_running(provider):
+            return (True, "")
+        return (False, "Not running")
+
+    # Cloud providers: check for API key
+    if get_api_key(provider) is not None:
+        return (True, "")
+    return (False, "No API key")
+
+
+def get_all_provider_availability() -> dict:
+    """
+    Get availability status for all providers.
+
+    Returns:
+        Dict mapping provider name to (is_available, reason) tuple
+    """
+    all_providers = list(PROVIDER_API_KEYS.keys()) + list(LOCAL_PROVIDER_DEFAULTS.keys())
+    return {provider: is_provider_available(provider) for provider in all_providers}

cli/compile_reports.py

@@ -0,0 +1,501 @@
+#!/usr/bin/env python3
+"""
+Compile all trading agent reports into a single consolidated PDF.
+
+Creates a PDF with:
+1. Summary table showing all symbols, their decisions, and analysis dates
+2. Detailed reports for each symbol (in order specified by REPORT_ORDER)
+
+Usage:
+    python cli/compile_reports.py                         # Compile all results into single PDF
+    python cli/compile_reports.py --output report.pdf     # Custom output filename
+    python cli/compile_reports.py --date 2026-01-18       # Filter to specific date (auto-names output)
+"""
+
+import argparse
+import re
+import sys
+from datetime import datetime
+from pathlib import Path
+
+import markdown2
+from playwright.sync_api import sync_playwright
+
+# Report order (top to bottom for each symbol's section)
+REPORT_ORDER = [
+    ("final_trade_decision.md", "Final Trade Decision"),
+    ("trader_investment_plan.md", "Trader Investment Plan"),
+    ("investment_plan.md", "Investment Plan"),
+    ("fundamentals_report.md", "Fundamentals Analysis"),
+    ("news_report.md", "News Analysis"),
+    ("sentiment_report.md", "Sentiment Analysis"),
+    ("market_report.md", "Market Analysis"),
+]
+
+# Clean GitHub-style markdown CSS
+CSS_STYLES = """
+@page {
+    size: A4;
+    margin: 0.75in;
+}
+
+body {
+    font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', 'Noto Sans', Helvetica, Arial, sans-serif;
+    font-size: 14px;
+    line-height: 1.6;
+    color: #24292f;
+    max-width: 100%;
+    margin: 0;
+    padding: 0;
+}
+
+h1 {
+    font-size: 2em;
+    font-weight: 600;
+    border-bottom: 1px solid #d0d7de;
+    padding-bottom: 0.3em;
+    margin-top: 24px;
+    margin-bottom: 16px;
+}
+
+h2 {
+    font-size: 1.5em;
+    font-weight: 600;
+    border-bottom: 1px solid #d0d7de;
+    padding-bottom: 0.3em;
+    margin-top: 24px;
+    margin-bottom: 16px;
+}
+
+h3 {
+    font-size: 1.25em;
+    font-weight: 600;
+    margin-top: 24px;
+    margin-bottom: 16px;
+}
+
+h4 {
+    font-size: 1em;
+    font-weight: 600;
+    margin-top: 24px;
+    margin-bottom: 16px;
+}
+
+p {
+    margin-top: 0;
+    margin-bottom: 16px;
+}
+
+ul, ol {
+    padding-left: 2em;
+    margin-top: 0;
+    margin-bottom: 16px;
+}
+
+li {
+    margin-bottom: 4px;
+}
+
+li + li {
+    margin-top: 4px;
+}
+
+table {
+    border-collapse: collapse;
+    width: 100%;
+    margin-top: 0;
+    margin-bottom: 16px;
+}
+
+th, td {
+    padding: 6px 13px;
+    border: 1px solid #d0d7de;
+}
+
+th {
+    background-color: #f6f8fa;
+    font-weight: 600;
+}
+
+tr:nth-child(2n) {
+    background-color: #f6f8fa;
+}
+
+hr {
+    border: 0;
+    border-top: 1px solid #d0d7de;
+    margin: 24px 0;
+}
+
+code {
+    background-color: rgba(175, 184, 193, 0.2);
+    padding: 0.2em 0.4em;
+    border-radius: 6px;
+    font-family: ui-monospace, SFMono-Regular, 'SF Mono', Menlo, Consolas, monospace;
+    font-size: 85%;
+}
+
+pre {
+    background-color: #f6f8fa;
+    padding: 16px;
+    border-radius: 6px;
+    overflow-x: auto;
+    margin-bottom: 16px;
+    font-size: 85%;
+    line-height: 1.45;
+}
+
+pre code {
+    padding: 0;
+    background: none;
+    font-size: 100%;
+}
+
+blockquote {
+    border-left: 0.25em solid #d0d7de;
+    padding: 0 1em;
+    margin: 0 0 16px 0;
+    color: #57606a;
+}
+
+strong {
+    font-weight: 600;
+}
+
+/* Decision color styling */
+.decision-buy { color: #1a7f37; font-weight: 700; }
+.decision-sell { color: #cf222e; font-weight: 700; }
+.decision-hold { color: #9a6700; font-weight: 700; }
+
+/* Symbol section - page break before each new symbol */
+.symbol-section {
+    page-break-before: always;
+}
+
+.symbol-section:first-of-type {
+    page-break-before: avoid;
+}
+
+/* Report title styling */
+.report-title {
+    color: #0969da;
+    font-size: 1.3em;
+    font-weight: 600;
+    margin-top: 32px;
+    margin-bottom: 16px;
+    padding-bottom: 8px;
+    border-bottom: 2px solid #0969da;
+}
+
+.report-title:first-of-type {
+    margin-top: 16px;
+}
+"""
+
+
+def extract_decision(content: str) -> str:
+    """Extract BUY/SELL/HOLD decision from final trade decision content."""
+    content_lower = content.lower()
+
+    patterns = [
+        r"recommendation[:\s]*\*{0,2}(buy|sell|hold)\*{0,2}",
+        r"\*{0,2}(buy|sell|hold)\*{0,2}[:\s]*recommendation",
+        r"final.*?decision[:\s]*\*{0,2}(buy|sell|hold)\*{0,2}",
+        r"recommend.*?(buy|sell|hold)",
+        r"action[:\s]*\*{0,2}(buy|sell|hold)\*{0,2}",
+    ]
+
+    for pattern in patterns:
+        match = re.search(pattern, content_lower)
+        if match:
+            return match.group(1).upper()
+
+    buy_count = len(re.findall(r"\bbuy\b", content_lower))
+    sell_count = len(re.findall(r"\bsell\b", content_lower))
+    hold_count = len(re.findall(r"\bhold\b", content_lower))
+
+    max_count = max(buy_count, sell_count, hold_count)
+    if max_count > 0:
+        if sell_count == max_count:
+            return "SELL"
+        if buy_count == max_count:
+            return "BUY"
+        return "HOLD"
+
+    return "N/A"
+
+
+def markdown_to_html(md_content: str) -> str:
+    """Convert markdown to HTML with extras."""
+    return markdown2.markdown(
+        md_content,
+        extras=[
+            "tables",
+            "fenced-code-blocks",
+            "strike",
+            "task_list",
+            "cuddled-lists",
+        ],
+    )
+
+
+def find_all_reports(results_dir: Path, date_filter: str | None = None) -> list[dict]:
+    """Find all report directories and extract their data.
+
+    Args:
+        results_dir: Path to the results directory
+        date_filter: Optional date string (YYYY-MM-DD) to filter reports
+    """
+    all_reports = []
+
+    if not results_dir.exists():
+        return all_reports
+
+    for symbol_dir in sorted(results_dir.iterdir()):
+        if not symbol_dir.is_dir():
+            continue
+
+        symbol = symbol_dir.name
+        if symbol.startswith(".") or " " in symbol:
+            continue
+
+        for date_dir in sorted(symbol_dir.iterdir(), reverse=True):
+            if not date_dir.is_dir():
+                continue
+
+            date = date_dir.name
+
+            # Skip if date doesn't match filter
+            if date_filter and date != date_filter:
+                continue
+
+            reports_dir = date_dir / "reports"
+
+            if not reports_dir.exists():
+                continue
+
+            report_files = []
+            decision = "N/A"
+
+            for filename, title in REPORT_ORDER:
+                file_path = reports_dir / filename
+                if file_path.exists():
+                    content = file_path.read_text(encoding="utf-8")
+                    html_content = markdown_to_html(content)
+                    report_files.append((filename, title, html_content))
+
+                    if filename == "final_trade_decision.md":
+                        decision = extract_decision(content)
+
+            if report_files:
+                all_reports.append({
+                    "symbol": symbol,
+                    "date": date,
+                    "decision": decision,
+                    "reports_dir": reports_dir,
+                    "reports": report_files,
+                })
+
+    return all_reports
+
+
+def build_html_document(all_reports: list[dict]) -> str:
+    """Build complete HTML document with summary table and all reports."""
+
+    # Build summary table rows
+    summary_rows = []
+    for report_data in all_reports:
+        decision = report_data["decision"]
+        decision_class = f"decision-{decision.lower()}" if decision in ["BUY", "SELL", "HOLD"] else ""
+        summary_rows.append(f'''<tr>
+            <td><strong>{report_data["symbol"]}</strong></td>
+            <td>{report_data["date"]}</td>
+            <td class="{decision_class}">{decision}</td>
+            <td>{len(report_data["reports"])} reports</td>
+        </tr>''')
+
+    summary_table = "\n".join(summary_rows)
+
+    # Build symbol sections
+    symbol_sections = []
+    for report_data in all_reports:
+        symbol = report_data["symbol"]
+        date = report_data["date"]
+        decision = report_data["decision"]
+        decision_class = f"decision-{decision.lower()}" if decision in ["BUY", "SELL", "HOLD"] else ""
+
+        # Build report content - simple flowing structure
+        reports_html_parts = []
+        for _, title, html_content in report_data["reports"]:
+            reports_html_parts.append(f'''<div class="report-title">{title}</div>
+{html_content}''')
+
+        reports_html = "\n".join(reports_html_parts)
+
+        symbol_sections.append(f'''<div class="symbol-section">
+<h1>{symbol} Trading Analysis Report</h1>
+<p><strong>Date:</strong> {date} &nbsp;|&nbsp; <strong>Recommendation:</strong> <span class="{decision_class}">{decision}</span></p>
+<hr>
+{reports_html}
+</div>''')
+
+    all_symbols_html = "\n".join(symbol_sections)
+
+    generated_date = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+
+    html = f'''<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>Trading Analysis Report</title>
+    <style>
+{CSS_STYLES}
+    </style>
+</head>
+<body>
+<h1>Trading Analysis Report</h1>
+<p><em>Generated: {generated_date}</em></p>
+
+<h2>Summary of Recommendations</h2>
+<table>
+    <thead>
+        <tr>
+            <th>Symbol</th>
+            <th>Analysis Date</th>
+            <th>Decision</th>
+            <th>Reports</th>
+        </tr>
+    </thead>
+    <tbody>
+        {summary_table}
+    </tbody>
+</table>
+
+<hr>
+
+{all_symbols_html}
+
+<hr>
+<p><em>Report generated by TradingAgents</em></p>
+</body>
+</html>'''
+    return html
+
+
+def compile_to_pdf(html_content: str, output_path: Path) -> bool:
+    """Generate PDF from HTML using Playwright."""
+    try:
+        with sync_playwright() as p:
+            browser = p.chromium.launch()
+            page = browser.new_page()
+            page.set_content(html_content, wait_until="networkidle")
+
+            page.pdf(
+                path=str(output_path),
+                format="A4",
+                margin={
+                    "top": "0.5in",
+                    "bottom": "0.5in",
+                    "left": "0.5in",
+                    "right": "0.5in",
+                },
+                print_background=True,
+            )
+            browser.close()
+        return True
+    except Exception as e:
+        print(f"Error generating PDF: {e}")
+        return False
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Compile all trading agent reports into a single consolidated PDF",
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+        epilog="""
+Examples:
+    python cli/compile_reports.py
+    python cli/compile_reports.py --output my_report.pdf
+    python cli/compile_reports.py --date 2026-01-18
+    python cli/compile_reports.py --date 2026-01-18 --output custom.pdf
+        """,
+    )
+    parser.add_argument(
+        "--output", "-o",
+        default="./results/trading_analysis_report.pdf",
+        help="Output PDF filename (default: ./results/trading_analysis_report.pdf)",
+    )
+    parser.add_argument(
+        "--results-dir", "-r",
+        default="./results",
+        help="Results directory (default: ./results)",
+    )
+    parser.add_argument(
+        "--date", "-d",
+        help="Filter reports to a specific date (format: YYYY-MM-DD)",
+    )
+
+    args = parser.parse_args()
+
+    # Validate date format if provided
+    if args.date:
+        import re as re_module
+        if not re_module.match(r"^\d{4}-\d{2}-\d{2}$", args.date):
+            print(f"Error: Invalid date format '{args.date}'. Expected YYYY-MM-DD")
+            sys.exit(1)
+
+    results_dir = Path(args.results_dir)
+    default_output = "./results/trading_analysis_report.pdf"
+
+    if not results_dir.exists():
+        print(f"Error: Results directory not found: {results_dir}")
+        sys.exit(1)
+
+    if args.date:
+        print(f"Scanning {results_dir} for reports on {args.date}...")
+    else:
+        print(f"Scanning {results_dir} for reports...")
+
+    all_reports = find_all_reports(results_dir, date_filter=args.date)
+
+    if not all_reports:
+        if args.date:
+            print(f"No reports found for date {args.date}")
+        else:
+            print("No reports found")
+        sys.exit(1)
+
+    print(f"Found {len(all_reports)} symbol analysis report(s):\n")
+
+    for report_data in all_reports:
+        decision_indicator = {
+            "BUY": "[BUY]",
+            "SELL": "[SELL]",
+            "HOLD": "[HOLD]",
+        }.get(report_data["decision"], "[N/A]")
+
+        print(f"  {report_data['symbol']:6} | {report_data['date']} | {decision_indicator:6} | {len(report_data['reports'])} reports")
+
+    # Determine output path
+    if args.date and args.output == default_output:
+        # Generate dynamic filename from date + symbols (up to 5)
+        symbols = [r["symbol"] for r in all_reports[:5]]
+        symbols_str = "_".join(symbols)
+        output_path = Path(f"./results/trading_report_{args.date}_{symbols_str}.pdf")
+    else:
+        output_path = Path(args.output)
+
+    print("\nGenerating PDF...")
+
+    html_document = build_html_document(all_reports)
+
+    if compile_to_pdf(html_document, output_path):
+        print(f"\n+ PDF created: {output_path}")
+    else:
+        print("\n- Failed to create PDF")
+        sys.exit(1)
+
+
+if __name__ == "__main__":
+    main()

cli/main.py

@@ -429,10 +429,12 @@ def get_user_selections():
             box_content += f"\n[dim]Default: {default}[/dim]"
         return Panel(box_content, border_style="blue", padding=(1, 2))
 
-    # Step 1: Ticker symbol
+    # Step 1: Ticker symbol(s)
     console.print(
         create_question_box(
-            "Step 1: Ticker Symbol", "Enter the ticker symbol to analyze", "SPY"
+            "Step 1: Ticker Symbol(s)",
+            "Enter ticker symbol(s) to analyze (comma-separated for multiple)",
+            "SPY",
         )
     )
     selected_ticker = get_ticker()
@@ -475,13 +477,20 @@ def get_user_selections():
     )
     selected_llm_provider, backend_url = select_llm_provider()
     
-    # Step 6: Thinking agents
+    # Step 6: Quick-Thinking LLM Engine
     console.print(
         create_question_box(
-            "Step 6: Thinking Agents", "Select your thinking agents for analysis"
+            "Step 6: Quick-Thinking LLM Engine", "Select your quick-thinking model for fast operations"
         )
     )
     selected_shallow_thinker = select_shallow_thinking_agent(selected_llm_provider)
+
+    # Step 7: Deep-Thinking LLM Engine
+    console.print(
+        create_question_box(
+            "Step 7: Deep-Thinking LLM Engine", "Select your deep-thinking model for complex reasoning"
+        )
+    )
     selected_deep_thinker = select_deep_thinking_agent(selected_llm_provider)
 
     return {
@@ -497,8 +506,11 @@ def get_user_selections():
 
 
 def get_ticker():
-    """Get ticker symbol from user input."""
-    return typer.prompt("", default="SPY")
+    """Get ticker symbol(s) from user input. Supports comma-separated symbols."""
+    raw_input = typer.prompt("", default="SPY")
+    # Split by comma, strip whitespace, convert to uppercase
+    symbols = [s.strip().upper() for s in raw_input.split(",") if s.strip()]
+    return symbols if len(symbols) > 1 else symbols[0]
 
 
 def get_analysis_date():
@@ -736,6 +748,7 @@ def extract_content_string(content):
         return str(content)
 
 def run_analysis():
+    """Run analysis for one or more ticker symbols."""
     # First get all user selections
     selections = get_user_selections()
 
@@ -748,13 +761,33 @@ def run_analysis():
     config["backend_url"] = selections["backend_url"]
     config["llm_provider"] = selections["llm_provider"].lower()
 
-    # Initialize the graph
+    # Normalize ticker(s) to list
+    tickers = selections["ticker"] if isinstance(selections["ticker"], list) else [selections["ticker"]]
+
+    # Initialize the graph once and reuse for all symbols
     graph = TradingAgentsGraph(
         [analyst.value for analyst in selections["analysts"]], config=config, debug=True
     )
 
+    for i, ticker in enumerate(tickers, 1):
+        if len(tickers) > 1:
+            console.print(f"\n[bold cyan]{'═' * 50}[/bold cyan]")
+            console.print(f"[bold cyan]  Analyzing {ticker} ({i}/{len(tickers)})[/bold cyan]")
+            console.print(f"[bold cyan]{'═' * 50}[/bold cyan]\n")
+
+        run_single_analysis(ticker, selections, config, graph)
+
+        if i < len(tickers):
+            console.print(f"\n[dim]Moving to next symbol...[/dim]\n")
+
+    if len(tickers) > 1:
+        console.print(f"\n[bold green]Completed analysis for all {len(tickers)} symbols: {', '.join(tickers)}[/bold green]")
+
+
+def run_single_analysis(ticker: str, selections: dict, config: dict, graph: TradingAgentsGraph):
+    """Run analysis for a single ticker symbol."""
     # Create result directory
-    results_dir = Path(config["results_dir"]) / selections["ticker"] / selections["analysis_date"]
+    results_dir = Path(config["results_dir"]) / ticker / selections["analysis_date"]
     results_dir.mkdir(parents=True, exist_ok=True)
     report_dir = results_dir / "reports"
     report_dir.mkdir(parents=True, exist_ok=True)
@@ -808,7 +841,7 @@ def run_analysis():
         update_display(layout)
 
         # Add initial messages
-        message_buffer.add_message("System", f"Selected ticker: {selections['ticker']}")
+        message_buffer.add_message("System", f"Selected ticker: {ticker}")
         message_buffer.add_message(
             "System", f"Analysis date: {selections['analysis_date']}"
         )
@@ -835,13 +868,13 @@ def run_analysis():
 
         # Create spinner text
         spinner_text = (
-            f"Analyzing {selections['ticker']} on {selections['analysis_date']}..."
+            f"Analyzing {ticker} on {selections['analysis_date']}..."
         )
         update_display(layout, spinner_text)
 
         # Initialize state and get graph args
         init_agent_state = graph.propagator.create_initial_state(
-            selections["ticker"], selections["analysis_date"]
+            ticker, selections["analysis_date"]
         )
         args = graph.propagator.get_graph_args()
 

cli/model_fetcher.py

@@ -0,0 +1,225 @@
+"""Dynamic model fetching from LLM provider APIs with caching."""
+
+import os
+from typing import List, Tuple, Optional
+import httpx
+
+# Cache for fetched models (provider -> list of models)
+_model_cache: dict = {}
+
+# Maximum number of models to display (None = no limit, show all)
+MAX_MODELS = None
+
+
+def is_fetch_latest() -> bool:
+    """Check if FETCH_LATEST is enabled in environment.
+
+    When enabled, fetches models dynamically from provider APIs.
+    When disabled, falls back to static hardcoded model lists.
+    """
+    return os.getenv("FETCH_LATEST", "false").lower() in ("true", "1", "yes")
+
+
+def fetch_openai_models() -> Optional[List[Tuple[str, str]]]:
+    """
+    Fetch available models from OpenAI API, sorted by creation date (newest first).
+
+    Returns:
+        List of (display_name, model_id) tuples, or None on failure
+    """
+    if "openai" in _model_cache:
+        return _model_cache["openai"]
+
+    api_key = os.getenv("OPENAI_API_KEY")
+    if not api_key or api_key.startswith("sk-or-"):
+        return None
+
+    try:
+        response = httpx.get(
+            "https://api.openai.com/v1/models",
+            headers={"Authorization": f"Bearer {api_key}"},
+            timeout=10.0
+        )
+        response.raise_for_status()
+        models_data = response.json().get("data", [])
+
+        # Filter to chat/reasoning models and keep metadata for sorting
+        chat_models = []
+        for model in models_data:
+            model_id = model.get("id", "")
+            created = model.get("created", 0)
+
+            # Include GPT models and reasoning models (o-series)
+            if (model_id.startswith("gpt-") or
+                model_id.startswith("o1") or
+                model_id.startswith("o3") or
+                model_id.startswith("o4") or
+                model_id.startswith("o5") or
+                model_id.startswith("gpt-5")):
+                # Skip snapshot/dated versions to keep list clean
+                if "-20" not in model_id and "-preview" not in model_id.lower():
+                    chat_models.append((model_id, created))
+
+        # Remove duplicates (keep highest created timestamp for each model_id)
+        model_dict = {}
+        for model_id, created in chat_models:
+            if model_id not in model_dict or created > model_dict[model_id]:
+                model_dict[model_id] = created
+
+        # Sort by created timestamp (newest first) and limit
+        sorted_models = sorted(model_dict.items(), key=lambda x: -x[1])[:MAX_MODELS]
+        result = [(model_id, model_id) for model_id, _ in sorted_models]
+
+        _model_cache["openai"] = result
+        return result
+    except (httpx.RequestError, httpx.HTTPStatusError, ValueError, KeyError):
+        return None
+
+
+def fetch_anthropic_models() -> Optional[List[Tuple[str, str]]]:
+    """
+    Fetch available models from Anthropic API, sorted by creation date (newest first).
+
+    Returns:
+        List of (display_name, model_id) tuples, or None on failure
+    """
+    if "anthropic" in _model_cache:
+        return _model_cache["anthropic"]
+
+    api_key = os.getenv("ANTHROPIC_API_KEY")
+    if not api_key:
+        return None
+
+    try:
+        response = httpx.get(
+            "https://api.anthropic.com/v1/models",
+            headers={
+                "x-api-key": api_key,
+                "anthropic-version": "2023-06-01"
+            },
+            timeout=10.0
+        )
+        response.raise_for_status()
+        models_data = response.json().get("data", [])
+
+        # Filter to Claude models and keep metadata for sorting
+        claude_models = []
+        for model in models_data:
+            model_id = model.get("id", "")
+            # Anthropic API returns created_at as ISO string (RFC 3339)
+            created_at = model.get("created_at", "")
+            display_name = model.get("display_name", "")
+
+            if model_id.startswith("claude-"):
+                # Skip dated versions (e.g., claude-3-sonnet-20240229)
+                if "-20" not in model_id:
+                    # Use display_name if available, otherwise model_id
+                    label = display_name if display_name else model_id
+                    claude_models.append((model_id, label, created_at))
+
+        # Remove duplicates (keep latest for each model_id)
+        model_dict = {}
+        for model_id, label, created_at in claude_models:
+            if model_id not in model_dict or created_at > model_dict[model_id][1]:
+                model_dict[model_id] = (label, created_at)
+
+        # Sort by created_at (newest first) and limit
+        sorted_models = sorted(model_dict.items(), key=lambda x: x[1][1], reverse=True)[:MAX_MODELS]
+        result = [(label, model_id) for model_id, (label, _) in sorted_models]
+
+        _model_cache["anthropic"] = result
+        return result
+    except (httpx.RequestError, httpx.HTTPStatusError, ValueError, KeyError):
+        return None
+
+
+def fetch_google_models() -> Optional[List[Tuple[str, str]]]:
+    """
+    Fetch available models from Google Generative AI API.
+    Uses displayName for user-friendly labels, sorted as returned by API (typically newest first).
+
+    Returns:
+        List of (display_name, model_id) tuples, or None on failure
+    """
+    if "google" in _model_cache:
+        return _model_cache["google"]
+
+    api_key = os.getenv("GEMINI_API_KEY")
+    if not api_key:
+        return None
+
+    try:
+        response = httpx.get(
+            f"https://generativelanguage.googleapis.com/v1/models?key={api_key}",
+            timeout=10.0
+        )
+        response.raise_for_status()
+        models_data = response.json().get("models", [])
+
+        # Filter to Gemini models that support generateContent
+        gemini_models = []
+        for model in models_data:
+            model_name = model.get("name", "")
+            display_name = model.get("displayName", "")
+            supported_methods = model.get("supportedGenerationMethods", [])
+
+            # Extract model ID from "models/gemini-..." format
+            if model_name.startswith("models/"):
+                model_id = model_name.replace("models/", "")
+            else:
+                model_id = model_name
+
+            # Only include Gemini models that support content generation
+            if model_id.startswith("gemini") and "generateContent" in supported_methods:
+                # Use displayName if available, otherwise model_id
+                label = display_name if display_name else model_id
+                gemini_models.append((label, model_id))
+
+        # API returns in a reasonable order, just dedupe and limit
+        seen = set()
+        unique_models = []
+        for label, model_id in gemini_models:
+            if model_id not in seen:
+                seen.add(model_id)
+                unique_models.append((label, model_id))
+
+        result = unique_models[:MAX_MODELS]
+
+        _model_cache["google"] = result
+        return result
+    except (httpx.RequestError, httpx.HTTPStatusError, ValueError, KeyError):
+        return None
+
+
+def fetch_models_for_provider(provider: str) -> Optional[List[Tuple[str, str]]]:
+    """
+    Fetch models for a given provider.
+
+    Only fetches dynamically if FETCH_LATEST is enabled. Otherwise returns None
+    to trigger fallback to static model lists.
+
+    Args:
+        provider: Provider name (openai, anthropic, google)
+
+    Returns:
+        List of (display_name, model_id) tuples, or None if not supported/failed
+    """
+    # Return None if FETCH_LATEST is not enabled - will use static lists
+    if not is_fetch_latest():
+        return None
+
+    provider_lower = provider.lower()
+
+    if provider_lower == "openai":
+        return fetch_openai_models()
+    elif provider_lower == "anthropic":
+        return fetch_anthropic_models()
+    elif provider_lower == "google":
+        return fetch_google_models()
+
+    return None
+
+
+def clear_cache():
+    """Clear the model cache."""
+    _model_cache.clear()

cli/utils.py

@@ -1,7 +1,12 @@
 import questionary
 from typing import List, Optional, Tuple, Dict
+from rich.console import Console
 
 from cli.models import AnalystType
+from cli.api_keys import is_provider_available
+from cli.model_fetcher import fetch_models_for_provider
+
+console = Console()
 
 ANALYST_ORDER = [
     ("Market Analyst", AnalystType.MARKET),
@@ -125,7 +130,7 @@ def select_research_depth() -> int:
 def select_shallow_thinking_agent(provider) -> str:
     """Select shallow thinking llm engine using an interactive selection."""
 
-    # Define shallow thinking llm engine options with their corresponding model names
+    # Static fallback options for each provider
     SHALLOW_AGENT_OPTIONS = {
         "openai": [
             ("GPT-4o-mini - Fast and efficient for quick tasks", "gpt-4o-mini"),
@@ -142,24 +147,43 @@ def select_shallow_thinking_agent(provider) -> str:
         "google": [
             ("Gemini 2.0 Flash-Lite - Cost efficiency and low latency", "gemini-2.0-flash-lite"),
             ("Gemini 2.0 Flash - Next generation features, speed, and thinking", "gemini-2.0-flash"),
-            ("Gemini 2.5 Flash - Adaptive thinking, cost efficiency", "gemini-2.5-flash-preview-05-20"),
+            ("Gemini 2.5 Flash-Lite - Lightweight and cost efficient", "gemini-2.5-flash-lite"),
+            ("Gemini 2.5 Flash - Adaptive thinking, cost efficiency", "gemini-2.5-flash"),
+            ("Gemini 3 Flash Preview - Latest generation flash model", "gemini-3-flash-preview"),
         ],
         "openrouter": [
+            ("Xiaomi MiMo V2 Flash - Fast and efficient multimodal model", "xiaomi/mimo-v2-flash:free"),
             ("Meta: Llama 4 Scout", "meta-llama/llama-4-scout:free"),
             ("Meta: Llama 3.3 8B Instruct - A lightweight and ultra-fast variant of Llama 3.3 70B", "meta-llama/llama-3.3-8b-instruct:free"),
             ("google/gemini-2.0-flash-exp:free - Gemini Flash 2.0 offers a significantly faster time to first token", "google/gemini-2.0-flash-exp:free"),
         ],
         "ollama": [
-            ("llama3.1 local", "llama3.1"),
-            ("llama3.2 local", "llama3.2"),
+            ("llama3.2:3b local", "llama3.2:3b"),
+            ("phi3.5 local", "phi3.5:latest"),
+        ],
+        "lm studio": [
+            ("Local Model (default)", "local-model"),
         ]
     }
 
+    provider_lower = provider.lower()
+
+    # Try dynamic fetch for supported providers (OpenAI, Anthropic, Google)
+    model_options = None
+    if provider_lower in ["openai", "anthropic", "google"]:
+        dynamic_models = fetch_models_for_provider(provider_lower)
+        if dynamic_models:
+            model_options = dynamic_models
+
+    # Fall back to static list if dynamic fetch failed or not supported
+    if model_options is None:
+        model_options = SHALLOW_AGENT_OPTIONS.get(provider_lower, [])
+
     choice = questionary.select(
         "Select Your [Quick-Thinking LLM Engine]:",
         choices=[
             questionary.Choice(display, value=value)
-            for display, value in SHALLOW_AGENT_OPTIONS[provider.lower()]
+            for display, value in model_options
         ],
         instruction="\n- Use arrow keys to navigate\n- Press Enter to select",
         style=questionary.Style(
@@ -183,7 +207,7 @@ def select_shallow_thinking_agent(provider) -> str:
 def select_deep_thinking_agent(provider) -> str:
     """Select deep thinking llm engine using an interactive selection."""
 
-    # Define deep thinking llm engine options with their corresponding model names
+    # Static fallback options for each provider
     DEEP_AGENT_OPTIONS = {
         "openai": [
             ("GPT-4.1-nano - Ultra-lightweight model for basic operations", "gpt-4.1-nano"),
@@ -199,29 +223,46 @@ def select_deep_thinking_agent(provider) -> str:
             ("Claude Sonnet 3.5 - Highly capable standard model", "claude-3-5-sonnet-latest"),
             ("Claude Sonnet 3.7 - Exceptional hybrid reasoning and agentic capabilities", "claude-3-7-sonnet-latest"),
             ("Claude Sonnet 4 - High performance and excellent reasoning", "claude-sonnet-4-0"),
-            ("Claude Opus 4 - Most powerful Anthropic model", "	claude-opus-4-0"),
+            ("Claude Opus 4 - Most powerful Anthropic model", "claude-opus-4-0"),
         ],
         "google": [
             ("Gemini 2.0 Flash-Lite - Cost efficiency and low latency", "gemini-2.0-flash-lite"),
             ("Gemini 2.0 Flash - Next generation features, speed, and thinking", "gemini-2.0-flash"),
-            ("Gemini 2.5 Flash - Adaptive thinking, cost efficiency", "gemini-2.5-flash-preview-05-20"),
-            ("Gemini 2.5 Pro", "gemini-2.5-pro-preview-06-05"),
+            ("Gemini 2.5 Flash-Lite - Lightweight and cost efficient", "gemini-2.5-flash-lite"),
+            ("Gemini 2.5 Flash - Adaptive thinking, cost efficiency", "gemini-2.5-flash"),
+            ("Gemini 3 Flash Preview - Latest generation flash model", "gemini-3-flash-preview"),
         ],
         "openrouter": [
+            ("Xiaomi MiMo V2 Flash - Fast and efficient multimodal model", "xiaomi/mimo-v2-flash:free"),
             ("DeepSeek V3 - a 685B-parameter, mixture-of-experts model", "deepseek/deepseek-chat-v3-0324:free"),
-            ("Deepseek - latest iteration of the flagship chat model family from the DeepSeek team.", "deepseek/deepseek-chat-v3-0324:free"),
         ],
         "ollama": [
-            ("llama3.1 local", "llama3.1"),
-            ("qwen3", "qwen3"),
+            ("llama3.2:3b local", "llama3.2:3b"),
+            ("phi3.5 local", "phi3.5:latest"),
+        ],
+        "lm studio": [
+            ("Local Model (default)", "local-model"),
         ]
     }
-    
+
+    provider_lower = provider.lower()
+
+    # Try dynamic fetch for supported providers (OpenAI, Anthropic, Google)
+    model_options = None
+    if provider_lower in ["openai", "anthropic", "google"]:
+        dynamic_models = fetch_models_for_provider(provider_lower)
+        if dynamic_models:
+            model_options = dynamic_models
+
+    # Fall back to static list if dynamic fetch failed or not supported
+    if model_options is None:
+        model_options = DEEP_AGENT_OPTIONS.get(provider_lower, [])
+
     choice = questionary.select(
         "Select Your [Deep-Thinking LLM Engine]:",
         choices=[
             questionary.Choice(display, value=value)
-            for display, value in DEEP_AGENT_OPTIONS[provider.lower()]
+            for display, value in model_options
         ],
         instruction="\n- Use arrow keys to navigate\n- Press Enter to select",
         style=questionary.Style(
@@ -240,22 +281,35 @@ def select_deep_thinking_agent(provider) -> str:
     return choice
 
 def select_llm_provider() -> tuple[str, str]:
-    """Select the OpenAI api url using interactive selection."""
-    # Define OpenAI api options with their corresponding endpoints
+    """Select the LLM provider using interactive selection with availability checks."""
+    # Define provider options with their corresponding endpoints
     BASE_URLS = [
         ("OpenAI", "https://api.openai.com/v1"),
         ("Anthropic", "https://api.anthropic.com/"),
         ("Google", "https://generativelanguage.googleapis.com/v1"),
         ("Openrouter", "https://openrouter.ai/api/v1"),
-        ("Ollama", "http://localhost:11434/v1"),        
+        ("Ollama", "http://localhost:11434/v1"),
+        ("LM Studio", "http://localhost:1234/v1"),
     ]
-    
+
+    # Build choices with availability status
+    choices = []
+    for display, url in BASE_URLS:
+        available, reason = is_provider_available(display)
+        if available:
+            choices.append(questionary.Choice(display, value=(display, url)))
+        else:
+            # Show disabled option with reason
+            disabled_label = f"{display} ({reason})"
+            choices.append(questionary.Choice(
+                disabled_label,
+                value=(display, url),
+                disabled=reason
+            ))
+
     choice = questionary.select(
         "Select your LLM Provider:",
-        choices=[
-            questionary.Choice(display, value=(display, value))
-            for display, value in BASE_URLS
-        ],
+        choices=choices,
         instruction="\n- Use arrow keys to navigate\n- Press Enter to select",
         style=questionary.Style(
             [
@@ -265,12 +319,12 @@ def select_llm_provider() -> tuple[str, str]:
             ]
         ),
     ).ask()
-    
+
     if choice is None:
-        console.print("\n[red]no OpenAI backend selected. Exiting...[/red]")
+        console.print("\n[red]No LLM provider selected. Exiting...[/red]")
         exit(1)
-    
+
     display_name, url = choice
     print(f"You selected: {display_name}\tURL: {url}")
-    
+
     return display_name, url

requirements.txt

@@ -24,3 +24,5 @@ rich
 questionary
 langchain_anthropic
 langchain-google-genai
+playwright
+markdown2

tradingagents/agents/analysts/fundamentals_analyst.py

@@ -1,7 +1,7 @@
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 import time
 import json
-from tradingagents.agents.utils.agent_utils import get_fundamentals, get_balance_sheet, get_cashflow, get_income_statement, get_insider_sentiment, get_insider_transactions
+from tradingagents.agents.utils.agent_utils import get_fundamentals, get_balance_sheet, get_cashflow, get_income_statement, get_insider_sentiment, get_insider_transactions, normalize_content
 from tradingagents.dataflows.config import get_config
 
 
@@ -53,7 +53,7 @@ def create_fundamentals_analyst(llm):
         report = ""
 
         if len(result.tool_calls) == 0:
-            report = result.content
+            report = normalize_content(result.content)
 
         return {
             "messages": [result],

tradingagents/agents/analysts/market_analyst.py

@@ -1,7 +1,7 @@
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 import time
 import json
-from tradingagents.agents.utils.agent_utils import get_stock_data, get_indicators
+from tradingagents.agents.utils.agent_utils import get_stock_data, get_indicators, normalize_content
 from tradingagents.dataflows.config import get_config
 
 
@@ -75,7 +75,7 @@ Volume-Based Indicators:
         report = ""
 
         if len(result.tool_calls) == 0:
-            report = result.content
+            report = normalize_content(result.content)
        
         return {
             "messages": [result],

tradingagents/agents/analysts/news_analyst.py

@@ -1,7 +1,7 @@
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 import time
 import json
-from tradingagents.agents.utils.agent_utils import get_news, get_global_news
+from tradingagents.agents.utils.agent_utils import get_news, get_global_news, normalize_content
 from tradingagents.dataflows.config import get_config
 
 
@@ -48,7 +48,7 @@ def create_news_analyst(llm):
         report = ""
 
         if len(result.tool_calls) == 0:
-            report = result.content
+            report = normalize_content(result.content)
 
         return {
             "messages": [result],

tradingagents/agents/analysts/social_media_analyst.py

@@ -1,7 +1,7 @@
 from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
 import time
 import json
-from tradingagents.agents.utils.agent_utils import get_news
+from tradingagents.agents.utils.agent_utils import get_news, normalize_content
 from tradingagents.dataflows.config import get_config
 
 
@@ -49,7 +49,7 @@ def create_social_media_analyst(llm):
         report = ""
 
         if len(result.tool_calls) == 0:
-            report = result.content
+            report = normalize_content(result.content)
 
         return {
             "messages": [result],

tradingagents/agents/utils/agent_utils.py

@@ -1,5 +1,20 @@
 from langchain_core.messages import HumanMessage, RemoveMessage
 
+
+def normalize_content(content):
+    """Normalize LLM response content to string.
+
+    Gemini returns content as a list of dicts with 'text' keys,
+    while OpenAI/Anthropic return a simple string.
+    """
+    if isinstance(content, list):
+        return "".join(
+            block.get("text", "") if isinstance(block, dict) else str(block)
+            for block in content
+        )
+    return content
+
+
 # Import tools from separate utility files
 from tradingagents.agents.utils.core_stock_tools import (
     get_stock_data

tradingagents/dataflows/config.py

@@ -5,6 +5,38 @@ from typing import Dict, Optional
 _config: Optional[Dict] = None
 DATA_DIR: Optional[str] = None
 
+# Local LLM providers that don't support OpenAI's web_search_preview
+LOCAL_LLM_PROVIDERS = ["ollama", "lm studio"]
+# Methods that require OpenAI's web_search_preview tool
+OPENAI_ONLY_METHODS = ["get_news", "get_global_news", "get_fundamentals"]
+
+
+def validate_config(config: Dict):
+    """Validate configuration and warn about incompatible settings."""
+    llm_provider = config.get("llm_provider", "").lower()
+
+    if llm_provider in LOCAL_LLM_PROVIDERS:
+        # Check data vendors
+        data_vendors = config.get("data_vendors", {})
+        tool_vendors = config.get("tool_vendors", {})
+
+        warnings = []
+        if data_vendors.get("news_data") == "openai":
+            warnings.append("data_vendors.news_data")
+        if data_vendors.get("fundamental_data") == "openai":
+            warnings.append("data_vendors.fundamental_data")
+
+        for method in OPENAI_ONLY_METHODS:
+            if tool_vendors.get(method) == "openai":
+                warnings.append(f"tool_vendors.{method}")
+
+        if warnings:
+            print(f"WARNING: Using local LLM provider '{llm_provider}' with 'openai' data vendors.")
+            print(f"  The following settings use OpenAI's web_search_preview which is not available locally:")
+            for w in warnings:
+                print(f"    - {w}")
+            print(f"  Recommendation: Change these to 'alpha_vantage', 'google', or 'local'.")
+
 
 def initialize_config():
     """Initialize the configuration with default values."""
@@ -21,6 +53,7 @@ def set_config(config: Dict):
         _config = default_config.DEFAULT_CONFIG.copy()
     _config.update(config)
     DATA_DIR = _config["data_dir"]
+    validate_config(_config)
 
 
 def get_config() -> Dict:

tradingagents/dataflows/interface.py

@@ -1,4 +1,5 @@
 from typing import Annotated
+import time
 
 # Import from vendor-specific modules
 from .local import get_YFin_data, get_finnhub_news, get_finnhub_company_insider_sentiment, get_finnhub_company_insider_transactions, get_simfin_balance_sheet, get_simfin_cashflow, get_simfin_income_statements, get_reddit_global_news, get_reddit_company_news
@@ -16,6 +17,7 @@ from .alpha_vantage import (
     get_news as get_alpha_vantage_news
 )
 from .alpha_vantage_common import AlphaVantageRateLimitError
+from openai import APIConnectionError, APITimeoutError, RateLimitError
 
 # Configuration and routing logic
 from .config import get_config
@@ -194,25 +196,58 @@ def route_to_vendor(method: str, *args, **kwargs):
         else:
             vendor_methods = [(vendor_impl, vendor)]
 
-        # Run methods for this vendor
+        # Run methods for this vendor with retry logic
         vendor_results = []
         for impl_func, vendor_name in vendor_methods:
-            try:
-                print(f"DEBUG: Calling {impl_func.__name__} from vendor '{vendor_name}'...")
-                result = impl_func(*args, **kwargs)
-                vendor_results.append(result)
-                print(f"SUCCESS: {impl_func.__name__} from vendor '{vendor_name}' completed successfully")
-                    
-            except AlphaVantageRateLimitError as e:
-                if vendor == "alpha_vantage":
-                    print(f"RATE_LIMIT: Alpha Vantage rate limit exceeded, falling back to next available vendor")
+            max_retries = 3
+            base_delay = 1.0
+            last_error = None
+
+            for retry_attempt in range(max_retries):
+                try:
+                    if retry_attempt > 0:
+                        print(f"RETRY: Attempt {retry_attempt + 1}/{max_retries} "
+                              f"for {impl_func.__name__}")
+                    else:
+                        print(f"DEBUG: Calling {impl_func.__name__} "
+                              f"from vendor '{vendor_name}'...")
+
+                    result = impl_func(*args, **kwargs)
+                    vendor_results.append(result)
+                    print(f"SUCCESS: {impl_func.__name__} from vendor "
+                          f"'{vendor_name}' completed successfully")
+                    last_error = None
+                    break  # Success, exit retry loop
+
+                except (AlphaVantageRateLimitError, RateLimitError) as e:
+                    print(f"RATE_LIMIT: {type(e).__name__} exceeded, falling back to next vendor.")
                     print(f"DEBUG: Rate limit details: {e}")
-                # Continue to next vendor for fallback
-                continue
-            except Exception as e:
-                # Log error but continue with other implementations
-                print(f"FAILED: {impl_func.__name__} from vendor '{vendor_name}' failed: {e}")
-                continue
+                    last_error = e
+                    break  # Don't retry rate limits, move to next vendor
+
+                except (ConnectionError, TimeoutError, OSError,
+                        APIConnectionError, APITimeoutError) as e:
+                    # Transient errors - retry with backoff
+                    last_error = e
+                    if retry_attempt < max_retries - 1:
+                        delay = base_delay * (2 ** retry_attempt)
+                        print(f"TRANSIENT_ERROR: {type(e).__name__} - {e}")
+                        print(f"RETRY: Waiting {delay}s before retry...")
+                        time.sleep(delay)
+                    else:
+                        print(f"FAILED: {impl_func.__name__} from vendor "
+                              f"'{vendor_name}' failed after {max_retries} "
+                              f"attempts: {e}")
+
+                except Exception as e:
+                    # Non-transient errors - don't retry
+                    last_error = e
+                    print(f"FAILED: {impl_func.__name__} from vendor "
+                          f"'{vendor_name}' failed: {type(e).__name__}: {e}")
+                    break
+
+            if last_error is not None:
+                continue  # Move to next implementation
 
         # Add this vendor's results
         if vendor_results:

tradingagents/dataflows/openai.py

@@ -1,7 +1,46 @@
+import os
+
 from openai import OpenAI
 from .config import get_config
 
 
+def _get_web_search_tool_type() -> str:
+    """Return the appropriate web search tool type based on FETCH_LATEST setting.
+
+    - FETCH_LATEST=true: Use 'web_search' (GA version, supports GPT-5)
+    - FETCH_LATEST=false/unset: Use 'web_search_preview' (legacy, wider compatibility)
+    """
+    fetch_latest = os.getenv("FETCH_LATEST", "false").lower() in ("true", "1", "yes")
+    return "web_search" if fetch_latest else "web_search_preview"
+
+
+def _extract_text_from_response(response):
+    """Safely extract text content from OpenAI Responses API output.
+
+    The response.output array typically contains:
+    - output[0]: ResponseFunctionWebSearch (the web search call)
+    - output[1]: ResponseOutputMessage (the text response)
+
+    This function handles edge cases where the structure may differ.
+    """
+    if not response.output:
+        raise RuntimeError("OpenAI response has empty output")
+
+    # Look for a message with text content
+    for item in response.output:
+        if hasattr(item, 'content') and item.content:
+            for content_block in item.content:
+                if hasattr(content_block, 'text') and content_block.text:
+                    return content_block.text
+
+    # If we get here, no text was found
+    output_types = [type(item).__name__ for item in response.output]
+    raise RuntimeError(
+        f"No text content found in OpenAI response. "
+        f"Output types: {output_types}"
+    )
+
+
 def get_stock_news_openai(query, start_date, end_date):
     config = get_config()
     client = OpenAI(base_url=config["backend_url"])
@@ -23,7 +62,7 @@ def get_stock_news_openai(query, start_date, end_date):
         reasoning={},
         tools=[
             {
-                "type": "web_search_preview",
+                "type": _get_web_search_tool_type(),
                 "user_location": {"type": "approximate"},
                 "search_context_size": "low",
             }
@@ -34,7 +73,7 @@ def get_stock_news_openai(query, start_date, end_date):
         store=True,
     )
 
-    return response.output[1].content[0].text
+    return _extract_text_from_response(response)
 
 
 def get_global_news_openai(curr_date, look_back_days=7, limit=5):
@@ -58,7 +97,7 @@ def get_global_news_openai(curr_date, look_back_days=7, limit=5):
         reasoning={},
         tools=[
             {
-                "type": "web_search_preview",
+                "type": _get_web_search_tool_type(),
                 "user_location": {"type": "approximate"},
                 "search_context_size": "low",
             }
@@ -69,7 +108,7 @@ def get_global_news_openai(curr_date, look_back_days=7, limit=5):
         store=True,
     )
 
-    return response.output[1].content[0].text
+    return _extract_text_from_response(response)
 
 
 def get_fundamentals_openai(ticker, curr_date):
@@ -93,7 +132,7 @@ def get_fundamentals_openai(ticker, curr_date):
         reasoning={},
         tools=[
             {
-                "type": "web_search_preview",
+                "type": _get_web_search_tool_type(),
                 "user_location": {"type": "approximate"},
                 "search_context_size": "low",
             }
@@ -104,4 +143,4 @@ def get_fundamentals_openai(ticker, curr_date):
         store=True,
     )
 
-    return response.output[1].content[0].text
+    return _extract_text_from_response(response)

tradingagents/graph/trading_graph.py

@@ -12,6 +12,8 @@ from langchain_google_genai import ChatGoogleGenerativeAI
 
 from langgraph.prebuilt import ToolNode
 
+from tradingagents.llm import requires_responses_api, ChatOpenAIResponses
+
 from tradingagents.agents import *
 from tradingagents.default_config import DEFAULT_CONFIG
 from tradingagents.agents.utils.memory import FinancialSituationMemory
@@ -72,9 +74,20 @@ class TradingAgentsGraph:
         )
 
         # Initialize LLMs
-        if self.config["llm_provider"].lower() == "openai" or self.config["llm_provider"] == "ollama" or self.config["llm_provider"] == "openrouter":
-            self.deep_thinking_llm = ChatOpenAI(model=self.config["deep_think_llm"], base_url=self.config["backend_url"])
-            self.quick_thinking_llm = ChatOpenAI(model=self.config["quick_think_llm"], base_url=self.config["backend_url"])
+        if self.config["llm_provider"].lower() in ["openai", "ollama", "openrouter", "lm studio"]:
+            # Select LLM class based on model - newer models require Responses API
+            deep_model = self.config["deep_think_llm"]
+            quick_model = self.config["quick_think_llm"]
+
+            if requires_responses_api(deep_model):
+                self.deep_thinking_llm = ChatOpenAIResponses(model=deep_model, base_url=self.config["backend_url"])
+            else:
+                self.deep_thinking_llm = ChatOpenAI(model=deep_model, base_url=self.config["backend_url"])
+
+            if requires_responses_api(quick_model):
+                self.quick_thinking_llm = ChatOpenAIResponses(model=quick_model, base_url=self.config["backend_url"])
+            else:
+                self.quick_thinking_llm = ChatOpenAI(model=quick_model, base_url=self.config["backend_url"])
         elif self.config["llm_provider"].lower() == "anthropic":
             self.deep_thinking_llm = ChatAnthropic(model=self.config["deep_think_llm"], base_url=self.config["backend_url"])
             self.quick_thinking_llm = ChatAnthropic(model=self.config["quick_think_llm"], base_url=self.config["backend_url"])

tradingagents/llm/__init__.py

@@ -0,0 +1,9 @@
+"""LLM wrapper utilities for TradingAgents.
+
+This module provides custom LLM wrappers for different API endpoints.
+"""
+
+from tradingagents.llm.model_utils import requires_responses_api
+from tradingagents.llm.openai_responses import ChatOpenAIResponses
+
+__all__ = ["requires_responses_api", "ChatOpenAIResponses"]

tradingagents/llm/model_utils.py

@@ -0,0 +1,24 @@
+"""Utility functions for model detection and selection."""
+
+# Model prefixes that require the OpenAI Responses API (/v1/responses)
+# instead of the Chat Completions API (/v1/chat/completions)
+RESPONSES_API_PREFIXES = [
+    "gpt-5",      # All GPT-5 variants (gpt-5, gpt-5.1, gpt-5.1-codex-mini, etc.)
+    "codex",      # Codex models that use Responses API
+]
+
+
+def requires_responses_api(model_name: str) -> bool:
+    """Check if a model requires the Responses API instead of Chat Completions.
+
+    Some newer OpenAI models only support the /v1/responses endpoint and will
+    return a 404 error if called via /v1/chat/completions.
+
+    Args:
+        model_name: The model identifier (e.g., "gpt-5.1-codex-mini", "gpt-4o")
+
+    Returns:
+        True if the model requires the Responses API, False otherwise.
+    """
+    model_lower = model_name.lower()
+    return any(prefix in model_lower for prefix in RESPONSES_API_PREFIXES)

tradingagents/llm/openai_responses.py

@@ -0,0 +1,307 @@
+"""LangChain-compatible wrapper for OpenAI's Responses API.
+
+This module provides ChatOpenAIResponses, a drop-in replacement for ChatOpenAI
+that uses the /v1/responses endpoint instead of /v1/chat/completions.
+
+This is required for newer models like gpt-5.1-codex-mini that only support
+the Responses API.
+"""
+
+import json
+import os
+import uuid
+from typing import Any, Dict, Iterator, List, Optional, Sequence, Union
+
+from langchain_core.callbacks import CallbackManagerForLLMRun
+from langchain_core.language_models.chat_models import BaseChatModel
+from langchain_core.messages import (
+    AIMessage,
+    BaseMessage,
+    HumanMessage,
+    SystemMessage,
+    ToolMessage,
+)
+from langchain_core.outputs import ChatGeneration, ChatResult
+from langchain_core.tools import BaseTool
+from openai import OpenAI
+from pydantic import Field
+
+
+class ChatOpenAIResponses(BaseChatModel):
+    """LangChain-compatible chat model using OpenAI's Responses API.
+
+    This class provides the same interface as ChatOpenAI but uses the
+    /v1/responses endpoint, which is required for certain newer models.
+
+    Example:
+        >>> llm = ChatOpenAIResponses(model="gpt-5.1-codex-mini")
+        >>> llm_with_tools = llm.bind_tools([my_tool])
+        >>> result = llm_with_tools.invoke([HumanMessage(content="Hello")])
+    """
+
+    model: str = Field(default="gpt-5.1-codex-mini")
+    base_url: Optional[str] = Field(default=None)
+    api_key: Optional[str] = Field(default=None)
+    temperature: float = Field(default=1.0)
+    max_output_tokens: int = Field(default=4096)
+    top_p: float = Field(default=1.0)
+
+    # Internal state for tool binding
+    _bound_tools: List[Dict[str, Any]] = []
+    _client: Optional[OpenAI] = None
+
+    class Config:
+        arbitrary_types_allowed = True
+
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+        self._bound_tools = []
+        self._client = None
+
+    @property
+    def _llm_type(self) -> str:
+        return "openai-responses"
+
+    @property
+    def client(self) -> OpenAI:
+        """Lazily initialize the OpenAI client."""
+        if self._client is None:
+            api_key = self.api_key or os.getenv("OPENAI_API_KEY")
+            if self.base_url:
+                self._client = OpenAI(api_key=api_key, base_url=self.base_url)
+            else:
+                self._client = OpenAI(api_key=api_key)
+        return self._client
+
+    def bind_tools(
+        self,
+        tools: Sequence[Union[Dict[str, Any], BaseTool]],
+        **kwargs: Any,
+    ) -> "ChatOpenAIResponses":
+        """Bind tools to this model instance.
+
+        Args:
+            tools: A sequence of tools to bind. Can be LangChain tools or dicts.
+
+        Returns:
+            A new ChatOpenAIResponses instance with the tools bound.
+        """
+        new_instance = ChatOpenAIResponses(
+            model=self.model,
+            base_url=self.base_url,
+            api_key=self.api_key,
+            temperature=self.temperature,
+            max_output_tokens=self.max_output_tokens,
+            top_p=self.top_p,
+        )
+        new_instance._bound_tools = self._convert_tools(tools)
+        return new_instance
+
+    def _convert_tools(
+        self, tools: Sequence[Union[Dict[str, Any], BaseTool]]
+    ) -> List[Dict[str, Any]]:
+        """Convert LangChain tools to OpenAI Responses API function format.
+
+        The Responses API uses a flat structure for function tools:
+        {
+            "type": "function",
+            "name": "function_name",
+            "description": "...",
+            "parameters": {...}
+        }
+
+        This differs from Chat Completions which nests under "function" key.
+        """
+        converted = []
+        for tool in tools:
+            if isinstance(tool, BaseTool):
+                # Get the JSON schema for parameters
+                if tool.args_schema:
+                    params = tool.args_schema.model_json_schema()
+                    # Remove title field that OpenAI doesn't expect at schema level
+                    params.pop("title", None)
+                else:
+                    params = {"type": "object", "properties": {}}
+
+                # Responses API uses flat structure - name at top level, not nested
+                tool_schema = {
+                    "type": "function",
+                    "name": tool.name,
+                    "description": tool.description or "",
+                    "parameters": params,
+                }
+                converted.append(tool_schema)
+            elif isinstance(tool, dict):
+                # Handle dict format - convert from Chat Completions format if needed
+                if "function" in tool:
+                    # Chat Completions format - flatten it
+                    func = tool["function"]
+                    tool_schema = {
+                        "type": "function",
+                        "name": func.get("name", ""),
+                        "description": func.get("description", ""),
+                        "parameters": func.get("parameters", {"type": "object", "properties": {}}),
+                    }
+                    converted.append(tool_schema)
+                elif "name" in tool:
+                    # Already in Responses API format
+                    converted.append(tool)
+                else:
+                    # Unknown format, try to use as-is
+                    converted.append(tool)
+        return converted
+
+    def _convert_messages(
+        self, messages: List[BaseMessage]
+    ) -> List[Dict[str, Any]]:
+        """Convert LangChain messages to OpenAI Responses API format.
+
+        The Responses API uses a different message format than Chat Completions:
+        - System/user messages use 'input_text' content type
+        - Assistant messages use 'output_text' content type (no function_call in content)
+        - Tool calls from assistant are represented as separate 'function_call' items
+        - Tool results use 'function_call_output' content type
+        """
+        converted = []
+        for msg in messages:
+            if isinstance(msg, SystemMessage):
+                content = msg.content if isinstance(msg.content, str) else str(msg.content)
+                converted.append({
+                    "role": "system",
+                    "content": [{"type": "input_text", "text": content}],
+                })
+            elif isinstance(msg, HumanMessage):
+                content = msg.content if isinstance(msg.content, str) else str(msg.content)
+                converted.append({
+                    "role": "user",
+                    "content": [{"type": "input_text", "text": content}],
+                })
+            elif isinstance(msg, AIMessage):
+                # Handle AI messages (assistant responses)
+                # First add text content if present
+                if msg.content:
+                    content = msg.content if isinstance(msg.content, str) else str(msg.content)
+                    converted.append({
+                        "role": "assistant",
+                        "content": [{"type": "output_text", "text": content}],
+                    })
+
+                # Tool calls need to be added as separate items in the Responses API
+                if hasattr(msg, 'tool_calls') and msg.tool_calls:
+                    for tc in msg.tool_calls:
+                        # Convert args to JSON string for the API
+                        args = tc.get("args", {})
+                        if isinstance(args, dict):
+                            args_str = json.dumps(args)
+                        else:
+                            args_str = str(args)
+
+                        # Add tool call as a separate item (not inside assistant content)
+                        converted.append({
+                            "type": "function_call",
+                            "call_id": tc.get("id", str(uuid.uuid4())),
+                            "name": tc["name"],
+                            "arguments": args_str,
+                        })
+                elif not msg.content:
+                    # Empty assistant message - add placeholder
+                    converted.append({
+                        "role": "assistant",
+                        "content": [{"type": "output_text", "text": ""}],
+                    })
+            elif isinstance(msg, ToolMessage):
+                # Tool results need to be formatted as function call outputs
+                content = msg.content if isinstance(msg.content, str) else str(msg.content)
+                converted.append({
+                    "type": "function_call_output",
+                    "call_id": msg.tool_call_id,
+                    "output": content,
+                })
+        return converted
+
+    def _parse_response(self, response: Any) -> AIMessage:
+        """Parse OpenAI Responses API response into LangChain AIMessage."""
+        text_content = ""
+        tool_calls = []
+
+        if not response.output:
+            return AIMessage(content="")
+
+        for item in response.output:
+            # Handle text output
+            if hasattr(item, 'content') and item.content:
+                for block in item.content:
+                    if hasattr(block, 'text') and block.text:
+                        text_content += block.text
+
+            # Handle function/tool calls
+            if hasattr(item, 'type') and item.type == 'function_call':
+                args = item.arguments
+                if isinstance(args, str):
+                    try:
+                        args = json.loads(args)
+                    except json.JSONDecodeError:
+                        args = {"raw": args}
+
+                tool_calls.append({
+                    "id": getattr(item, 'id', None) or getattr(item, 'call_id', None) or str(uuid.uuid4()),
+                    "name": item.name,
+                    "args": args,
+                })
+
+        if tool_calls:
+            return AIMessage(content=text_content, tool_calls=tool_calls)
+        return AIMessage(content=text_content)
+
+    def _generate(
+        self,
+        messages: List[BaseMessage],
+        stop: Optional[List[str]] = None,
+        run_manager: Optional[CallbackManagerForLLMRun] = None,
+        **kwargs: Any,
+    ) -> ChatResult:
+        """Generate a response using the OpenAI Responses API.
+
+        Args:
+            messages: List of LangChain messages to send.
+            stop: Optional stop sequences (not used by Responses API).
+            run_manager: Optional callback manager.
+
+        Returns:
+            ChatResult containing the model's response.
+        """
+        # Convert messages to Responses API format
+        converted_messages = self._convert_messages(messages)
+
+        # Build request parameters
+        request_params = {
+            "model": self.model,
+            "input": converted_messages,
+            "temperature": self.temperature,
+            "max_output_tokens": self.max_output_tokens,
+            "top_p": self.top_p,
+        }
+
+        # Add tools if bound
+        if self._bound_tools:
+            request_params["tools"] = self._bound_tools
+
+        # Make the API call
+        response = self.client.responses.create(**request_params)
+
+        # Parse the response
+        ai_message = self._parse_response(response)
+
+        return ChatResult(
+            generations=[ChatGeneration(message=ai_message)],
+        )
+
+    @property
+    def _identifying_params(self) -> Dict[str, Any]:
+        """Return identifying parameters for this LLM."""
+        return {
+            "model": self.model,
+            "base_url": self.base_url,
+            "temperature": self.temperature,
+            "max_output_tokens": self.max_output_tokens,
+        }