feat: add optional Adanos sentiment prompt context

Experiments/multi_model_ipo/prompt_orchestration/get_prompt_data.py

@@ -72,8 +72,12 @@ def get_macro_news(n: int = 5, summary_limit: int = 200):
 # =========================================================
 
 MASSIVE_API_KEY = os.getenv("MASSIVE_API_KEY")
+ADANOS_API_KEY = os.getenv("ADANOS_API_KEY")
 BASE_URL = "https://api.polygon.io"
+ADANOS_API_BASE_URL = (os.getenv("ADANOS_API_BASE_URL") or "https://api.adanos.org").rstrip("/")
 REQUEST_TIMEOUT = 20
+ADANOS_REQUEST_TIMEOUT = 8
+ADANOS_PLATFORMS = ("reddit", "x", "news", "polymarket")
 
 session = requests.Session()
 
@@ -113,10 +117,82 @@ def _request_json(url: str, params: dict) -> dict | list | None:
         resp.raise_for_status()
         return resp.json()
 
+    except (requests.RequestException, ValueError):
+        return None
+
+
+def _request_adanos_json(platform: str, ticker: str) -> dict | None:
+    if not ADANOS_API_KEY:
+        return None
+
+    try:
+        resp = session.get(
+            f"{ADANOS_API_BASE_URL}/{platform}/stocks/v1/stock/{ticker}",
+            headers={"X-API-Key": ADANOS_API_KEY},
+            timeout=ADANOS_REQUEST_TIMEOUT,
+        )
+        if resp.status_code == 404:
+            return None
+        resp.raise_for_status()
+        payload = resp.json()
+        return payload if isinstance(payload, dict) else None
     except requests.RequestException:
         return None
 
 
+def _format_adanos_sentiment_row(ticker: str, platform: str, payload: dict) -> str | None:
+    if payload.get("found") is False:
+        return None
+
+    fields = [
+        f"TICKER={ticker}",
+        f"SOURCE={platform}",
+        f"SENTIMENT={payload.get('sentiment_score', 'UNKNOWN')}",
+        f"BUZZ={payload.get('buzz_score', 'UNKNOWN')}",
+        f"MENTIONS={payload.get('mentions', 'UNKNOWN')}",
+        f"BULLISH_PCT={payload.get('bullish_pct', 'UNKNOWN')}",
+        f"BEARISH_PCT={payload.get('bearish_pct', 'UNKNOWN')}",
+        f"TREND={payload.get('trend', 'UNKNOWN')}",
+    ]
+    return " | ".join(fields)
+
+
+def get_adanos_sentiment_context(tickers, max_tickers: int = 12) -> str:
+    """
+    Return optional Adanos Market Sentiment context for already-allowed tickers.
+
+    This helper never expands the model's trading universe. Callers should pass
+    only portfolio tickers or the explicit IPO whitelist already included in the
+    prompt.
+    """
+    normalized: list[str] = []
+    for ticker in tickers:
+        if not isinstance(ticker, str):
+            continue
+        value = ticker.strip().upper()
+        if value and value not in normalized:
+            normalized.append(value)
+        if len(normalized) >= max_tickers:
+            break
+
+    lines = ["ADANOS_SENTIMENT_START"]
+    if not normalized:
+        lines.append("No eligible tickers supplied.")
+    elif not ADANOS_API_KEY:
+        lines.append("Optional Adanos Market Sentiment API disabled: ADANOS_API_KEY is not configured.")
+    else:
+        for ticker in normalized:
+            for platform in ADANOS_PLATFORMS:
+                payload = _request_adanos_json(platform, ticker)
+                if not payload:
+                    continue
+                row = _format_adanos_sentiment_row(ticker, platform, payload)
+                if row:
+                    lines.append(row)
+    lines.append("ADANOS_SENTIMENT_END")
+    return "\n".join(lines)
+
+
 # =========================================================
 # FEATURE HELPERS
 # =========================================================

Experiments/multi_model_ipo/prompts/daily_prompt.py

@@ -1,6 +1,10 @@
 import pandas as pd
 
-from ..prompt_orchestration.get_prompt_data import get_macro_news, build_eligibility_series
+from ..prompt_orchestration.get_prompt_data import (
+    build_eligibility_series,
+    get_adanos_sentiment_context,
+    get_macro_news,
+)
 from libb.model import LIBBmodel
 
 # -------------------------------------------------------------------
@@ -115,6 +119,7 @@
 You receive ONLY:
 
 - MACRO_NEWS (rates, inflation, liquidity, sector rotation)
+- ADANOS_SENTIMENT_CONTEXT (optional sentiment for PORTFOLIO_STATE tickers only)
 - PORTFOLIO_STATE (authoritative holdings, cost basis, stops, conviction)
 - TRADE_EXECUTION_LOG (CSV: timestamp | ticker | action | status | shares | price)
 
@@ -137,6 +142,9 @@
 MACRO_NEWS:
 {MACRO_NEWS}
 
+ADANOS_SENTIMENT_CONTEXT:
+{ADANOS_SENTIMENT_CONTEXT}
+
 PORTFOLIO_STATE:
 {PORTFOLIO_STATE}
 
@@ -295,6 +303,7 @@ def create_daily_prompt(libb: LIBBmodel):
 
     macro_news = get_macro_news()
     portfolio_state = libb.portfolio
+    adanos_sentiment_context = get_adanos_sentiment_context(portfolio_state["ticker"])
     portfolio_eligibility = build_eligibility_series(portfolio_state["ticker"])
     execution_log = libb.recent_execution_logs()
 
@@ -308,6 +317,7 @@ def create_daily_prompt(libb: LIBBmodel):
         + INPUT_BLOCK
         + GIVEN_DATA.format(
             MACRO_NEWS=macro_news,
+            ADANOS_SENTIMENT_CONTEXT=adanos_sentiment_context,
             PORTFOLIO_STATE=portfolio_state,
             TRADE_EXECUTION_LOG=execution_log,
             PORTFOLIO_TICKER_ELIGIBILITY=portfolio_eligibility,
@@ -319,4 +329,4 @@ def create_daily_prompt(libb: LIBBmodel):
         + OUTPUT_TEMPLATE
     )
 
-    return prompt
+    return prompt

Experiments/multi_model_ipo/prompts/deep_research_prompt.py

@@ -98,6 +98,7 @@
 You receive ONLY:
 
 - MACRO_NEWS (rates, inflation, liquidity, sector rotation)
+- ADANOS_SENTIMENT_CONTEXT (optional sentiment for IPO_UNIVERSE and current portfolio tickers only)
 - IPO_UNIVERSE (hard BUY whitelist)
 - PORTFOLIO_STATE (authoritative holdings, cost basis, stops, conviction)
 - TRADE_EXECUTION_LOG (CSV: timestamp | ticker | action | status | shares | price)
@@ -121,6 +122,9 @@
 MACRO_NEWS:
 {MACRO_NEWS}
 
+ADANOS_SENTIMENT_CONTEXT:
+{ADANOS_SENTIMENT_CONTEXT}
+
 IPO_UNIVERSE:
 {IPO_UNIVERSE}
 
@@ -285,10 +289,13 @@ def create_deep_research_prompt(libb: LIBBmodel):
 
     ipo_universe = get_ipo_universe()
     formatted_ipo_universe = format_universe_for_prompt(ipo_universe)
+    sentiment_tickers = [item.get("ticker") for item in ipo_universe if isinstance(item, dict)]
 
     ipo_universe_eligibility = build_eligibility_series_from_universe(ipo_universe)
 
     portfolio_state = libb.portfolio
+    sentiment_tickers.extend(portfolio_state["ticker"])
+    adanos_sentiment_context = get_adanos_sentiment_context(sentiment_tickers)
     portfolio_tickers_eligibility = build_eligibility_series(portfolio_state["ticker"])
     execution_log = libb.recent_execution_logs()
 
@@ -308,6 +315,7 @@ def create_deep_research_prompt(libb: LIBBmodel):
         + INPUT_BLOCK
         + GIVEN_DATA.format(
             MACRO_NEWS=macro_news,
+            ADANOS_SENTIMENT_CONTEXT=adanos_sentiment_context,
             IPO_UNIVERSE=formatted_ipo_universe,
             IPO_TICKER_ELIGIBILITY=ipo_universe_eligibility,
             PORTFOLIO_STATE=portfolio_state,
@@ -323,4 +331,4 @@ def create_deep_research_prompt(libb: LIBBmodel):
         + OUTPUT_TEMPLATE
     )
 
-    return prompt
+    return prompt

README.md

@@ -138,6 +138,20 @@ Here are the artifacts links for the Micro-Cap Experiment:
 - yfinance (primary data source)  
 - Stooq (fallback data source)  
 - Matplotlib  
+- Optional Adanos Market Sentiment API context for IPO experiment prompts (`ADANOS_API_KEY`)
+
+### Optional Sentiment Context
+
+The multi-model IPO experiment can enrich prompts with Adanos Market Sentiment
+context for tickers that are already present in the allowed portfolio or IPO
+universe. This does not expand the model's trading universe; it only adds
+Reddit, X / FinTwit, News, and Polymarket sentiment context for eligible U.S.
+equities.
+
+```bash
+export ADANOS_API_KEY="sk_live_..."
+export ADANOS_API_BASE_URL="https://api.adanos.org"
+```
 
 ---
 

tests/test_adanos_sentiment_context.py

@@ -0,0 +1,47 @@
+import sys
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[1]
+if str(ROOT) not in sys.path:
+    sys.path.insert(0, str(ROOT))
+
+from Experiments.multi_model_ipo.prompt_orchestration import get_prompt_data
+
+
+def test_adanos_sentiment_context_is_disabled_without_key(monkeypatch):
+    monkeypatch.setattr(get_prompt_data, "ADANOS_API_KEY", None)
+
+    output = get_prompt_data.get_adanos_sentiment_context(["AAPL"])
+
+    assert "ADANOS_SENTIMENT_START" in output
+    assert "ADANOS_API_KEY is not configured" in output
+    assert "ADANOS_SENTIMENT_END" in output
+
+
+def test_adanos_sentiment_context_formats_available_sources(monkeypatch):
+    calls = []
+
+    def fake_request(platform, ticker):
+        calls.append((platform, ticker))
+        if platform == "reddit":
+            return {
+                "found": True,
+                "sentiment_score": 0.31,
+                "buzz_score": 64.2,
+                "mentions": 42,
+                "bullish_pct": 52,
+                "bearish_pct": 17,
+                "trend": "rising",
+            }
+        return {"found": False}
+
+    monkeypatch.setattr(get_prompt_data, "ADANOS_API_KEY", "sk_live_test")
+    monkeypatch.setattr(get_prompt_data, "_request_adanos_json", fake_request)
+
+    output = get_prompt_data.get_adanos_sentiment_context(["aapl", "AAPL", "msft"])
+
+    assert "TICKER=AAPL | SOURCE=reddit | SENTIMENT=0.31" in output
+    assert "TICKER=MSFT | SOURCE=reddit | SENTIMENT=0.31" in output
+    assert ("reddit", "AAPL") in calls
+    assert calls.count(("reddit", "AAPL")) == 1