fix(graph): resolve instrument identity to stop wrong-company hallucination

Agents had no ground-truth ticker→company mapping, so the market analyst
could pattern-match a price chart to the wrong company (e.g. TOTDY read as
"TotalEnergies"), and every downstream agent inherited the bad framing.

Resolve identity once at run start via a cached, fail-open yfinance lookup
and inject company/sector/exchange into the shared instrument context that
all twelve agents consume, with an explicit do-not-substitute instruction.
Resolution runs on both the propagate() and CLI entry points.

Also replaces the bare "Continue" message-clear placeholder, which some
OpenAI-compatible providers interpreted as the user task, with a
context-anchored placeholder carrying the resolved identity and date.

#814 #888

tradingagents/agents/utils/agent_states.py

@@ -46,6 +46,7 @@ class RiskDebateState(TypedDict):
 class AgentState(MessagesState):
     company_of_interest: Annotated[str, "Company that we are interested in trading"]
     asset_type: Annotated[str, "Asset type under analysis such as stock or crypto"]
+    instrument_context: Annotated[str, "Deterministic ticker identity resolved at run start"]
     trade_date: Annotated[str, "What date we are trading at"]
 
     sender: Annotated[str, "Agent that sent this message"]

tradingagents/agents/utils/agent_utils.py

@@ -1,3 +1,8 @@
+import functools
+import logging
+from typing import Any, Mapping, Optional
+
+import yfinance as yf
 from langchain_core.messages import HumanMessage, RemoveMessage
 
 # Import tools from separate utility files
@@ -19,6 +24,8 @@ from tradingagents.agents.utils.news_data_tools import (
     get_global_news
 )
 
+logger = logging.getLogger(__name__)
+
 
 def get_language_instruction() -> str:
     """Return a prompt instruction for the configured output language.
@@ -36,32 +43,145 @@ def get_language_instruction() -> str:
     return f" Write your entire response in {lang}."
 
 
-def build_instrument_context(ticker: str, asset_type: str = "stock") -> str:
-    """Describe the exact instrument so agents preserve exchange-qualified tickers."""
-    instrument_label = "asset" if asset_type == "crypto" else "instrument"
-    extra_hint = (
-        " Treat it as a crypto asset rather than a company, and do not assume company fundamentals are available."
-        if asset_type == "crypto"
-        else ""
+def _clean_identity_value(value: Any) -> Optional[str]:
+    """Return a trimmed string, or None for empty / placeholder-ish values."""
+    if not isinstance(value, str):
+        return None
+    cleaned = value.strip()
+    if not cleaned or cleaned.lower() in {"none", "n/a", "nan", "null"}:
+        return None
+    return cleaned
+
+
+@functools.lru_cache(maxsize=256)
+def resolve_instrument_identity(ticker: str) -> dict:
+    """Resolve deterministic identity metadata (company name, sector, …) for a ticker.
+
+    This exists to stop the pipeline from hallucinating a *different* company
+    when a chart pattern suggests a different industry than the real one
+    (#814): without a ground-truth name, the market analyst would pattern-match
+    the price action to a narrative and invent an identity that then cascaded
+    through every downstream agent.
+
+    Best-effort by design: if yfinance is unavailable, rate-limited, or doesn't
+    recognise the ticker, we return ``{}`` and the caller falls back to
+    ticker-only context rather than failing before analysis starts. Cached so
+    the lookup happens at most once per ticker per process.
+    """
+    try:
+        info = yf.Ticker(ticker.upper()).info or {}
+    except Exception as exc:  # noqa: BLE001 — fail open, never block the run
+        logger.debug("Could not resolve instrument identity for %s: %s", ticker, exc)
+        return {}
+
+    identity: dict[str, str] = {}
+    company_name = _clean_identity_value(info.get("longName")) or _clean_identity_value(
+        info.get("shortName")
     )
-    return (
+    if company_name:
+        identity["company_name"] = company_name
+    for source_key, target_key in (
+        ("sector", "sector"),
+        ("industry", "industry"),
+        ("exchange", "exchange"),
+        ("quoteType", "quote_type"),
+    ):
+        value = _clean_identity_value(info.get(source_key))
+        if value:
+            identity[target_key] = value
+    return identity
+
+
+def build_instrument_context(
+    ticker: str,
+    asset_type: str = "stock",
+    identity: Optional[Mapping[str, str]] = None,
+) -> str:
+    """Describe the exact instrument so agents preserve identity and ticker.
+
+    When ``identity`` is provided (resolved deterministically via
+    :func:`resolve_instrument_identity`), the company name and business
+    classification are injected so agents anchor to the real company rather
+    than pattern-matching the price chart to a wrong one (#814).
+    """
+    is_crypto = asset_type == "crypto"
+    instrument_label = "asset" if is_crypto else "instrument"
+    context = (
         f"The {instrument_label} to analyze is `{ticker}`. "
         "Use this exact ticker in every tool call, report, and recommendation, "
         "preserving any exchange suffix (e.g. `.TO`, `.L`, `.HK`, `.T`, `-USD`)."
-        + extra_hint
     )
 
+    details = []
+    if identity:
+        name = identity.get("company_name") or identity.get("name")
+        if name:
+            details.append(f"{'Name' if is_crypto else 'Company'}: {name}")
+        sector, industry = identity.get("sector"), identity.get("industry")
+        if sector and industry:
+            details.append(f"Business classification: {sector} / {industry}")
+        elif sector:
+            details.append(f"Sector: {sector}")
+        elif industry:
+            details.append(f"Industry: {industry}")
+        if identity.get("exchange"):
+            details.append(f"Exchange: {identity['exchange']}")
+
+    if details:
+        context += (
+            f" Resolved identity: {'; '.join(details)}. "
+            "Do not substitute a different company or ticker unless a tool "
+            "result explicitly disproves this resolved identity."
+        )
+
+    if is_crypto:
+        context += (
+            " Treat it as a crypto asset rather than a company, and do not "
+            "assume company fundamentals are available."
+        )
+    return context
+
+
+def get_instrument_context_from_state(state: Mapping[str, Any]) -> str:
+    """Return the instrument context for the current run.
+
+    Prefers the identity-resolved context computed once at run start and
+    stored on the state (see ``TradingAgentsGraph.resolve_instrument_context``).
+    Falls back to a ticker-only context — with no network lookup — when the
+    state was constructed without it (bare programmatic states, tests), so a
+    consumer is never forced to make a yfinance call mid-graph.
+    """
+    context = state.get("instrument_context")
+    if isinstance(context, str) and context.strip():
+        return context
+    return build_instrument_context(
+        str(state["company_of_interest"]),
+        state.get("asset_type", "stock"),
+    )
+
+
 def create_msg_delete():
     def delete_messages(state):
-        """Clear messages and add placeholder for Anthropic compatibility"""
-        messages = state["messages"]
+        """Clear messages and add a context-anchored placeholder.
 
-        # Remove all messages
+        The placeholder must not be a bare ``"Continue"``: some
+        OpenAI-compatible providers interpret that literally as the user task
+        and produce output about the word "continue" instead of analysing the
+        instrument (#888). Anchoring it to the resolved instrument context and
+        date keeps the next analyst on-task even if the provider treats the
+        placeholder as a standalone request.
+        """
+        messages = state["messages"]
         removal_operations = [RemoveMessage(id=m.id) for m in messages]
 
-        # Add a minimal placeholder message
-        placeholder = HumanMessage(content="Continue")
-
+        instrument_context = get_instrument_context_from_state(state)
+        trade_date = state.get("trade_date", "the requested date")
+        placeholder = HumanMessage(
+            content=(
+                f"Proceed with your assigned analysis for this workflow. "
+                f"{instrument_context} The analysis date is {trade_date}."
+            )
+        )
         return {"messages": removal_operations + [placeholder]}
 
     return delete_messages