feat(news): read sentiment_score with v1 fallback

dashboard/backend/integrations/news_sentiment.py

@@ -236,7 +236,9 @@ def _story_fields(sig: dict) -> dict:
 def _project_entry(sig: dict, reference_ts: float) -> dict:
     return {
         "sentiment": sig["sentiment"],
-        "score": sig["score"],
+        # v2 sentiment_score with transitional v1 fallback — PR-2 of the
+        # FinSearch score-field disambiguation deletes the fallback.
+        "score": sig["sentiment_score"] if "sentiment_score" in sig else sig["score"],
         **_story_fields(sig),
         "n_articles": sig["n_articles"],
         "age_hours": max(0.0, (reference_ts - float(sig["published"])) / 3600.0),

dashboard/backend/tests/test_news_sentiment_fixture.py

@@ -102,7 +102,7 @@ def test_items_fixture_is_the_batch_the_signals_fixture_came_from():
 
 def test_fixture_matches_contract_essentials():
     body = load_signals_fixture()
-    assert body["schema_version"] == 1
+    assert body["schema_version"] in (1, 2)  # transitional; PR-2 pins == 2
     assert isinstance(body["signals"], dict) and body["signals"]
     sample = next(iter(body["signals"].values()))
     for field in ("sentiment", "score", "rationale", "headline", "source",

dashboard/backend/tests/test_sentiment_score_fallback.py

@@ -0,0 +1,36 @@
+"""PR-1 of the FinSearch score-field disambiguation (see FinSearch spec
+2026-07-14-score-field-disambiguation-design.md): _project_entry must read
+sentiment_score (signals v2) and fall back to score (v1) until PR-2 deletes
+the fallback."""
+import pytest
+
+from dashboard.backend.integrations.news_sentiment import _project_entry
+
+BASE = {"sentiment": "bullish", "rationale": "r", "headline": "h",
+        "source": "Reuters", "url": "https://example.com/a",
+        "published": 1783330000.0, "guid": "g1", "n_articles": 2}
+
+
+def test_project_entry_reads_sentiment_score_v2():
+    entry = _project_entry({**BASE, "sentiment_score": 0.5},
+                           reference_ts=1783333600.0)
+    assert entry["score"] == 0.5
+
+
+def test_project_entry_falls_back_to_v1_score():
+    entry = _project_entry({**BASE, "score": -0.3},
+                           reference_ts=1783333600.0)
+    assert entry["score"] == -0.3
+
+
+def test_project_entry_prefers_sentiment_score_when_both_present():
+    entry = _project_entry({**BASE, "sentiment_score": 0.5, "score": -0.9},
+                           reference_ts=1783333600.0)
+    assert entry["score"] == 0.5
+
+
+def test_project_entry_raises_when_both_score_keys_absent():
+    # Fail loud like every other required field: a payload with neither key is
+    # broken upstream, and score=None would flow silently into the panel.
+    with pytest.raises(KeyError):
+        _project_entry(dict(BASE), reference_ts=1783333600.0)

docs/integrations/finsearch-news-sentiment.md

@@ -36,6 +36,11 @@ No ATL-side contract change is needed — the `/api/v2` refactor shipped the who
   | `age_hours` | float | ≥ 0.0 |
   | `n_articles` | int | ≥ 0 |
   | `rationale` | str \| None | optional — see design note |
+
+> **Transitional:** until FinSearch's schema-v2 deploy lands, the producer
+> still sends `score` (v1); the adapter reads `sentiment_score` with a `score`
+> fallback. PR-2 removes the fallback and pins `schema_version == 2`.
+
 - **Universe:** the loader is called with `list(DJIA_30)` (the canonical current Dow-30 constant, `infrastructure/llm/validator.py`, reconciled in #91/#94 — the old `AMEX` typo is gone). Pass that straight through as the `?tickers=` filter.
 
 ---
@@ -82,7 +87,7 @@ Top level — the fields the **public, bearer-gated** response carries (consume
 ```json
 "MSFT": {
   "sentiment": "bullish",
-  "score": 0.5,
+  "sentiment_score": 0.5,
   "rationale": "Two distinct outlets report upbeat Azure guidance.",
   "headline": "Microsoft raises Azure guidance after record quarter",
   "source": "Reuters",
@@ -104,7 +109,7 @@ The producer artifact and the consumer type were designed independently and **do
 | `NewsSentimentEntry` | ← source | transform |
 |----------------------|----------|-----------|
 | `sentiment` | `sentiment` | passthrough — the enums are identical (`bullish`/`bearish`/`neutral`) |
-| `score` | `score` | passthrough (already −1…1) |
+| `score` | `sentiment_score` | passthrough (already −1…1) |
 | `headline` | `headline` | passthrough |
 | `source` | `source` | passthrough |
 | `url` | `url` | passthrough |
@@ -191,7 +196,7 @@ def get_news_sentiment(universe, timestamp):
     for ticker, s in body.get("signals", {}).items():
         out[ticker] = {
             "sentiment": s["sentiment"],
-            "score": s["score"],
+            "score": s["sentiment_score"],
             "headline": s["headline"],
             "source": s["source"],
             "url": s["url"],