Skip to content

Cli

cli

Flag-driven runner for the TradingAgents pipeline.

Exposes :func:run_cli as the cli subcommand of the tradingagents console script. Each parameter is a fire-friendly flag, with defaults matching the previous hard-coded values from the legacy cli.py.

Functions:

Name Description
run_cli

Run the TradingAgents pipeline for a single ticker.

DEFAULT_ANALYSTS 

DEFAULT_ANALYSTS: tuple[str, ...] = SUPPORTED_ANALYSTS

run_cli 

run_cli(
    ticker: str = "GOOG",
    date: str | None = None,
    llm_provider: LLMProvider = "google_genai",
    deep_think_llm: str = "gemini-3.1-pro-preview",
    quick_think_llm: str = "gemini-3-flash-preview",
    reasoning_effort: ReasoningEffort = "high",
    response_language: ResponseLanguage = "zh-TW",
    max_debate_rounds: int = 10,
    max_risk_discuss_rounds: int = 10,
    max_recur_limit: int = 100,
    selected_analysts: list[str] | tuple[str, ...] | None = None,
    debug: bool = True,
) -> str

Run the TradingAgents pipeline for a single ticker.

Parameters:

Name Type Description Default

ticker

str

Ticker symbol or company name to analyse. Defaults to GOOG.

 'GOOG'

date

str | None

Trade date in YYYY-MM-DD format. Defaults to today's local date when None.

 None

llm_provider

LLMProvider

LangChain init_chat_model registry key. Defaults to google_genai.

 'google_genai'

deep_think_llm

str

Model name for deep-thinking nodes (Research Manager, Risk Manager). Defaults to gemini-3.1-pro-preview.

 'gemini-3.1-pro-preview'

quick_think_llm

str

Model name for quick-thinking nodes (analysts, researchers, trader, debaters). Defaults to gemini-3-flash-preview.

 'gemini-3-flash-preview'

reasoning_effort

ReasoningEffort

Unified reasoning level mapped per provider. Defaults to high.

 'high'

response_language

ResponseLanguage

BCP 47 language tag (zh-TW, zh-CN, en-US, ja-JP, ko-KR, de-DE) appended to agent prompts. Defaults to zh-TW.

 'zh-TW'

max_debate_rounds

int

Maximum Bull/Bear investment debate rounds. Defaults to 10.

 10

max_risk_discuss_rounds

int

Maximum risk management debate rounds. Defaults to 10.

 10

max_recur_limit

int

Maximum recursion limit for LangGraph execution. Defaults to 100.

 100

selected_analysts

list[str] | tuple[str, ...] | None

Analyst types to include. Valid options are market, social, news, fundamentals. Defaults to all four when None.

 None

debug

bool

Stream agent messages live (still routed through the Rich renderer regardless of this flag — the flag only affects the underlying graph's verbosity). Defaults to True.

 True

Returns:

Name Type Description
str str

The final BUY / SELL / HOLD decision text.
Source code in src/tradingagents/interface/cli.py 
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
def run_cli(  # noqa: PLR0913
    ticker: str = "GOOG",
    date: str | None = None,
    llm_provider: LLMProvider = "google_genai",
    deep_think_llm: str = "gemini-3.1-pro-preview",
    quick_think_llm: str = "gemini-3-flash-preview",
    reasoning_effort: ReasoningEffort = "high",
    response_language: ResponseLanguage = "zh-TW",
    max_debate_rounds: int = 10,
    max_risk_discuss_rounds: int = 10,
    max_recur_limit: int = 100,
    selected_analysts: list[str] | tuple[str, ...] | None = None,
    debug: bool = True,
) -> str:
    """Run the TradingAgents pipeline for a single ticker.

    Args:
        ticker (str): Ticker symbol or company name to analyse.
            Defaults to GOOG.
        date (str | None, optional): Trade date in YYYY-MM-DD format.
            Defaults to today's local date when None.
        llm_provider (LLMProvider, optional): LangChain init_chat_model
            registry key. Defaults to google_genai.
        deep_think_llm (str, optional): Model name for deep-thinking nodes
            (Research Manager, Risk Manager). Defaults to
            gemini-3.1-pro-preview.
        quick_think_llm (str, optional): Model name for quick-thinking
            nodes (analysts, researchers, trader, debaters). Defaults to
            gemini-3-flash-preview.
        reasoning_effort (ReasoningEffort, optional): Unified reasoning
            level mapped per provider. Defaults to high.
        response_language (ResponseLanguage, optional): BCP 47 language
            tag (zh-TW, zh-CN, en-US, ja-JP, ko-KR, de-DE) appended to
            agent prompts. Defaults to zh-TW.
        max_debate_rounds (int, optional): Maximum Bull/Bear investment
            debate rounds. Defaults to 10.
        max_risk_discuss_rounds (int, optional): Maximum risk management
            debate rounds. Defaults to 10.
        max_recur_limit (int, optional): Maximum recursion limit for
            LangGraph execution. Defaults to 100.
        selected_analysts (list[str] | tuple[str, ...] | None, optional):
            Analyst types to include. Valid options are market, social,
            news, fundamentals. Defaults to all four when None.
        debug (bool, optional): Stream agent messages live (still routed
            through the Rich renderer regardless of this flag — the flag
            only affects the underlying graph's verbosity). Defaults to
            True.

    Returns:
        str: The final BUY / SELL / HOLD decision text.
    """
    date = _normalize_trade_date(date)
    analysts = _normalize_selected_analysts(selected_analysts)

    config = TradingAgentsConfig(
        llm_provider=llm_provider,
        deep_think_llm=deep_think_llm,
        quick_think_llm=quick_think_llm,
        max_debate_rounds=max_debate_rounds,
        max_risk_discuss_rounds=max_risk_discuss_rounds,
        max_recur_limit=max_recur_limit,
        reasoning_effort=reasoning_effort,
        response_language=response_language,
    )

    console = Console()
    print_run_header(console, ticker=ticker, trade_date=date, config=config)

    renderer = MessageRenderer.for_console(console=console)

    ta = TradingAgentsGraph(debug=debug, config=config, selected_analysts=analysts)
    _, decision = ta.propagate(ticker, date, on_message=renderer)

    print_final_decision(console, decision)
    return decision