sentiment: LLM buy/hold/avoid + full analysis, and search-budget scoping

Richer LLM output (same grounded call, ~no extra cost):
- All providers now also return a recommendation (buy/hold/avoid) and a thorough
  reasoning paragraph; Gemini now actually captures reasoning + grounding
  citations (it was dropping them). Stored on sentiment_scores (migration 008),
  exposed in the API; display-only — NOT fed into the composite/EV.
- Ticker Sentiment panel shows an "LLM view" badge and a "Full analysis & sources"
  expander with the complete reasoning + citations.

Search-budget scoping (Gemini grounding free tier = 5000/mo):
- collect_sentiment now targets only watchlist + open paper trades + top-N by
  composite, skips tickers refreshed within sentiment_fresh_hours (72h), and caps
  per run (sentiment_max_per_run). Once the relevant set is fresh, runs spend 0
  searches until it ages out — bounding monthly usage well under the free tier.
- Widened sentiment lookback to 7d (scoring + display) so sparser collection
  still feeds the dimension score.

Deploy: alembic upgrade (sentiment_scores.recommendation). Switch provider to
Gemini Flash in Admin for the cost win (grounded, cheapest).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

app/providers/openai_compatible_sentiment.py

@@ -28,18 +28,26 @@ _CA_BUNDLE = os.environ.get("SSL_CERT_FILE", "")
 
 _SENTIMENT_PROMPT = """\
 Assess the CURRENT market sentiment for the stock ticker {ticker} based on your \
-knowledge of the company, its sector, and recent developments you are aware of.
+knowledge of the company, its sector, and recent developments you are aware of, \
+and whether BUYING here looks advisable.
 
-Respond ONLY with a JSON object in this exact format (no markdown, no extra text):
-{{"classification": "<bullish|bearish|neutral>", "confidence": <0-100>, "reasoning": "<brief explanation>"}}
+Respond ONLY with a JSON object (no markdown, no extra text):
+{{"classification": "<bullish|bearish|neutral>", "confidence": <0-100>, "recommendation": "<buy|hold|avoid>", "reasoning": "<a thorough explanation of the drivers>"}}
 
 Rules:
 - classification must be exactly one of: bullish, bearish, neutral
+- recommendation must be exactly one of: buy, hold, avoid
 - confidence must be an integer from 0 to 100
-- reasoning should be a brief one-sentence explanation
+- reasoning should be several sentences
 """
 
 VALID_CLASSIFICATIONS = {"bullish", "bearish", "neutral"}
+VALID_RECOMMENDATIONS = {"buy", "hold", "avoid"}
+
+
+def _parse_recommendation(value: object) -> str | None:
+    v = str(value or "").strip().lower()
+    return v if v in VALID_RECOMMENDATIONS else None
 
 
 def _clean_json_text(raw: str) -> str:
@@ -116,6 +124,7 @@ class OpenAICompatibleSentimentProvider:
                 source=self._source,
                 timestamp=datetime.now(timezone.utc),
                 reasoning=reasoning,
+                recommendation=_parse_recommendation(parsed.get("recommendation")),
             )
 
         except json.JSONDecodeError as exc: