Minimal AgentState dict for portfolio_manager_node.

Flatten a captured prompt (str, message list, or objects) to text.

A minimal context-manager response whose read() runs ``read_fn``.

Build a MagicMock LLM whose with_structured_output binding captures the prompt and returns a real TraderProposal so render_trader_proposal works.

Return ``llm.with_structured_output(schema)`` or ``None`` if unsupported. Logs a warning when the binding fails so the user understands the agent

Return the step number of the latest checkpoint, or None if none exists.

Create the initial state for the agent graph. ``instrument_context`` is the deterministic ticker-identity string resolved once at run

Classify on the canonical symbol so e.g. BTCUSD and BTC-USDT both read as crypto (#981/#982), matching what the data path will actually fetch.

Check whether a resumable checkpoint exists for ticker+date.

Analysis dates from ``start_date``, never past today. A future date has no outcome to settle against, and the graph rejects one, so the grid

Read a portfolio JSON file, failing before the run rather than mid-graph.

Resolve user input to its canonical Yahoo symbol (single source of truth). Delegates to the data layer's ``normalize_symbol`` so the symbol the C

Return one of Buy / Overweight / Hold / Underweight / Sell, or REVIEW. An unrecognizable decision yields ``REVIEW`` rather than a fabricated

Return the default backend URL for a provider key, or None if unknown.

Render a SentimentReport to the markdown shape the rest of the system expects. The structured header (band + score + confidence) is prepended to

Render a TraderProposal to markdown. The trailing ``FINAL TRANSACTION PROPOSAL: **BUY/HOLD/SELL**`` line is preserved for backward compatibil

Alpha Vantage path; the API call is mocked so a leak would be visible.

A graph without __init__ (no LLM clients), wired to a temp log.

Return the SQLite checkpoint DB path for a ticker.

Drop annual/quarterly reports dated after curr_date to prevent look-ahead. ``_make_api_request`` returns the fundamentals payload as a JSON strin

(display_name, provider_key, base_url) for every supported provider. Shared by the interactive picker and by env-driven configuration so an e

The option value equal to ``default``, or None to leave the menu as is.

Prompt for a required value; exit cleanly if the user cancels. ``questionary.text(...).ask()`` returns None on Ctrl-C/Esc; mirror the exit-on

Call the tool through a ToolNode in a graph carrying the run's trade_date.

Build a MagicMock LLM whose with_structured_output binding captures the prompt and returns a real PortfolioDecision (so render_pm_decision works).

MagicMock LLM whose structured binding captures the prompt and returns a real SentimentReport so render_sentiment_report works.

``[start, end]`` with its end clamped to the run date. A window wholly after the run date keeps its length and moves back to end there.

Extract string content from various message formats. Returns None if no meaningful text content is found.

Drop financial statement columns (fiscal period timestamps) after curr_date. yfinance financial statements use fiscal period end dates as columns

Initialize the configuration with default values.

Fetch OHLCV data with caching, filtered to prevent look-ahead bias. Downloads 5 years of data up to today and caches per symbol. On subsequen

Resolve the backend URL with the correct precedence. An explicit env override (``env_url``, from ``TRADINGAGENTS_LLM_BACKEND_URL`` via ``DEFA

Determine if risk analysis should continue.

Save a completed run's reports to ``save_path``; return the complete-report path.

Drop oldest resolved blocks when their count exceeds max_entries. Pending blocks are always kept (they represent unprocessed work). R

Stub _bedrock_class so the constructor kwargs are testable without the optional langchain-aws extra installed.

Normalise an LLM-written optional numeric field before validation. Three shapes show up in practice: a placeholder string ("None", "N/A") in

Create tool nodes for different data sources using abstract methods.

Parse a message's ISO 8601 ``created_at``; None when missing or malformed.

Normalize the date column to ``Date``. Some yfinance builds leave the index unnamed (so ``reset_index()`` yields ``index``) or use ``Datetime

Extract article data from yfinance news format (handles nested 'content' structure).

Log the final state to a JSON file.

Point-in-time cutoff for past-context lessons (#1251). A historical/backtest run (trade date before today) filters lessons to those a

Gamma encodes ``outcomes``/``outcomePrices`` as JSON-string arrays.

A post's ``created_utc`` epoch as a UTC datetime, or None when missing.

Minimal DataFrame matching yfinance .history() output shape. Uses a DatetimeIndex like real yfinance output, so resolution-date extraction (s

Build a resolved entry tag, recording the outcome's known-by date. ``resolution_date`` (the date of the last price bar used for the return)

Execute the graph and write the resulting state to disk and memory log.

Select a model for the given provider and mode (quick/deep).

Whether the (native OpenAI) model accepts ``reasoning_effort``.

Concise prompt for reflect_on_final_decision (Phase B log entries). Produces 2-4 sentences of plain prose, compact enough to be re-injected

Ask for report output language. ``default`` is offered only when it is one of the listed languages: a custom one entered last time is free te

Apply multiple outcome updates in a single read + atomic write. Each element of updates must have keys: ticker, trade_date, raw_retur