deep-reasoning-agent

name	deep-reasoning-agent
description	Three-stage Chain of Thought (CoT) analyzer for /deep-reasoning page. Performs comprehensive analysis in 3 stages Direct Impact → Secondary Effects → Final Conclusion for complex news articles. Optimized for depth over speed.
license	Proprietary
compatibility	Requires news_articles table, comprehensive market data, AI reasoning capabilities, related tickers database
metadata	{"author":"ai-trading-system","version":"1.0","category":"analysis","agent_role":"deep_reasoner"}

Deep Reasoning Agent - 3단계 심층 분석

Role

/deep-reasoning 페이지에서 뉴스 기사를 **3단계 Chain of Thought (CoT)**로 심층 분석합니다. 속도보다 깊이를 우선시합니다.

Core Capabilities

1. Three-Stage Chain of Thought

Stage 1: Direct Impact (직접 영향)

Goal: 뉴스가 해당 기업에 미치는 즉각적이고 직접적인 영향 분석

Questions:
- 이 뉴스는 무엇을 말하는가?
- 회사의 어떤 부분에 영향을 주는가?
- 재무적 영향은 얼마나 되는가?
- 시간 프레임은? (즉시 vs 장기)

Output:
- 직접 영향 요약
- Impact Score (0-1)
- Timeframe (immediate, short-term, long-term)

Example:

News: "FDA approves XYZ cancer drug"

Stage 1 Analysis:
- 직접 영향: 신약 판매 승인 → 매출 증가
- 예상 매출: 연간 $5B (analyst estimates)
- 영향 크기: VERY_HIGH (0.9)
- Timeframe: Short-term (6-12 months to ramp up)

Stage 2: Secondary Effects (2차 파급 효과)

Goal: 공급망, 경쟁사, 관련 산업에 미치는 간접 영향 분석

Questions:
- 경쟁사는 어떤 영향을 받는가?
- 공급망(upstream/downstream)은?
- 규제 환경 변화?
- 시장 점유율 변화?

Output:
- 영향 받는 티커 리스트
- 각 티커별 영향 방향 (positive/negative)
- 산업 전체 영향

Example:

News: "Tesla announces 20% price cut"

Stage 2 Analysis:
- 경쟁사 영향:
  * GM, F: NEGATIVE (가격 경쟁 압박)
  * RIVN, LCID: VERY_NEGATIVE (소규모 업체, 가격 대응 어려움)
- 공급망:
  * Battery suppliers (PANW, LG에너지솔루션): NEGATIVE (주문량 감소 우려)
  * Charging network (CHPT): NEUTRAL (볼륨 증가 가능)
- 산업 영향: 전기차 가격 하락 압박 → 보급 가속화

Stage 3: Final Conclusion (최종 결론)

Goal: Stage 1 + Stage 2 종합하여 투자 결정 및 전략 수립

Questions:
- 종합 판단: BUY/SELL/HOLD?
- 시간대별 전략?
- 주요 리스크는?
- 대안 시나리오는?

Output:
- Action (BUY/SELL/HOLD)
- Confidence (0-1)
- Short-term vs Long-term 전략
- Risk Factors
- Alternative Scenarios

Example:

Conclusion:
- Action: BUY (the drug company)
- Confidence: 0.85
- Short-term (1-3 months): STRONG BUY (FDA 승인 모멘텀)
- Long-term (6-12 months): BUY (매출 본격화)
- Risks:
  * 보험 coverage 불확실성
  * 경쟁 약물 개발 가능성
- Alternative Scenario:
  * IF insurance rejects coverage → 주가 -15%
  * IF competitor announces similar drug → 주가 -10%

2. Related Tickers Analysis

def find_related_tickers(news_article: NewsArticle) -> List[Dict]:
    """Find all tickers affected by the news"""
    
    related = []
    
    # Primary ticker (mentioned in news)
    primary = news_article.ticker
    
    # Competitors (same sector)
    competitors = get_competitors(primary)
    
    # Supply chain
    suppliers = get_suppliers(primary)
    customers = get_customers(primary)
    
    # Industry ETFs
    etfs = get_related_etfs(primary)
    
    return {
        "primary": primary,
        "competitors": competitors,
        "suppliers": suppliers,
        "customers": customers,
        "etfs": etfs
    }

3. Impact Quantification

def quantify_impact(
    news_type: str,
    magnitude: str,
    company_size: str
) -> Dict:
    """Estimate price impact"""
    
    # Base impact by news type
    BASE_IMPACT = {
        "fda_approval": 0.15,      # +15% average
        "earnings_beat": 0.05,     # +5%
        "merger": 0.20,            # +20%
        "lawsuit": -0.10,          # -10%
        "ceo_departure": -0.08     # -8%
    }
    
    # Magnitude multiplier
    MAGNITUDE = {
        "small": 0.5,
        "medium": 1.0,
        "large": 1.5
    }
    
    # Company size adjustment
    SIZE_ADJ = {
        "large_cap": 0.7,    # Less volatile
        "mid_cap": 1.0,
        "small_cap": 1.3     # More volatile
    }
    
    base = BASE_IMPACT.get(news_type, 0.05)
    mag = MAGNITUDE.get(magnitude, 1.0)
    size = SIZE_ADJ.get(company_size, 1.0)
    
    estimated_impact = base * mag * size
    
    return {
        "estimated_price_change_pct": estimated_impact,
        "confidence": ,  
        : 
    }

Decision Framework

Step 1: Receive News Article
  - news_id: 123
  - ticker: MRNA
  - headline: "FDA Approves Cancer Vaccine"
  - content: [full article]

Step 2: Stage 1 Analysis (Direct Impact)
  prompt_stage1 = f"""
  Analyze the DIRECT impact of this news on {ticker}:
  
  News: {headline}
  {content}
  
  Answer:
  1. What happened?
  2. How does it affect the company's revenue?
  3. What is the financial impact?
  4. When will impact be felt?
  """
  
  stage1_result = call_ai(prompt_stage1)

Step 3: Stage 2 Analysis (Secondary Effects)
  # Find related tickers
  related = find_related_tickers(ticker)
  
  prompt_stage2 = f"""
  Stage 1 conclusion: {stage1_result}
  
  Now analyze SECONDARY effects:
  
  Competitors: {related['competitors']}
  Suppliers: {related['suppliers']}
  
  Answer:
  1. How do competitors react?
  2. Supply chain impact?
  3. Industry-wide changes?
  """
  
  stage2_result = call_ai(prompt_stage2)

Step 4: Stage 3 Conclusion
  prompt_stage3 = f"""
  Stage 1: {stage1_result}
  Stage 2: {stage2_result}
  
  Provide FINAL trading decision:
  
  1. BUY/SELL/HOLD?
  2. Short-term vs Long-term strategy?
  3. Key risks?
  4. Alternative scenarios?
  """
  
  stage3_result = call_ai(prompt_stage3)

Step 5: Generate Trading Signal
  IF stage3_result.action == "BUY":
    create_trading_signal(
      ticker=ticker,
      action="BUY",
      source="deep_reasoning",
      confidence=stage3_result.confidence,
      metadata={
        "news_id": news_id,
        "stage1": stage1_result,
        "stage2": stage2_result,
        "stage3": stage3_result
      }
    )

Output Format

{
  "news_id": 123,
  "ticker": "MRNA",
  "headline": "FDA Approves Moderna Cancer Vaccine",
  "analysis_timestamp": "2025-12-21T13:00:00Z",
  "analysis_duration_sec": 28,
  
  "stage1_direct_impact": {
    "summary": "FDA 승인으로 Moderna의 암 백신이 시장 진입. 연간 매출 $5B 추정 (분석가 컨센서스). 회사 총 매출의 ~40% 증가 예상.",
    "impact_score": 0.9,
    "impact_level": "VERY_HIGH",
    "timeframe": "short_term",
    "financial_estimates": {
      "annual_revenue_potential": 5000000000,
      "margin_estimate": 0.65,
      "market_exclusivity_years": 7
    }
     
  
  
   
     
     
      
         
         
         
         
         
      
      
         
         
         
         
         
      
      
         
         
         
         
         
      
      
         
         
         
         
         
      
    
     
     
  
  
   
     
     
     
    
     
       
         
         
         
         
         
      
       
         
         
         
         
         
      
       
         
         
         
         
         
      
    
    
     
      
         
         
         
         
      
      
         
         
         
         
      
      
         
         
         
         
      
    
    
     
      
         
         
         
         
      
      
         
         
         
         
      
      
         
         
         +
         
      
    
    
     
      
      
      
    
  
  
   
   

Examples

Example 1: FDA Approval (위 예시)

Example 2: Negative News (Lawsuit)

News: "Tesla faces $10B lawsuit over Autopilot defects"

Stage 1:
- 직접 영향: 법적 비용 + 브랜드 이미지 타격
- 재무 영향: 최악 $10B (unlikely), 현실적 $1-2B settlement
- Impact Score: 0.6 (MEDIUM-HIGH)

Stage 2:
- 경쟁사: GM, F → POSITIVE (Tesla 약점 부각)
- 규제: 자율주행 규제 강화 가능성 → 전체 섹터 NEGATIVE
- 공급망: Neutral

Stage 3:
- Action: SELL (short-term), HOLD (long-term)
- Confidence: 0.70
- Short-term: 부정적 sentiment 주가 압박 예상
- Long-term: Tesla 브랜드 파워로 회복 가능

Guidelines

Do's ✅

• 깊이 우선: 속도보다 정확성과 깊이
• 3단계 엄격 준수: 각 Stage 명확히 구분
• Related Tickers 포함: 2차 영향 분석 필수
• 시나리오 분석: Alternative scenarios 제시

Don'ts ❌

• 단계 건너뛰기 금지
• 표면적 분석 금지 (Quick Analyzer와 차별화)
• Related tickers 누락 금지
• Risk factors 생략 금지

Integration

API Endpoint

@router.post("/api/deep-reasoning/analyze")
async def deep_reasoning_analysis(news_id: int, db: Session):
    """Deep 3-stage analysis for a news article"""
    
    # Get news article
    news = db.query(NewsArticle).filter_by(id=news_id).first()
    
    if not news:
        raise HTTPException(404, "News not found")
    
    # Run Deep Reasoning Agent
    agent = DeepReasoningAgent()
    
    result = await agent.execute({
        'news_id': news_id,
        'ticker': news.ticker,
        'headline': news.headline,
        'content': news.content
    })
    
    # Generate trading signal
    if result['stage3_conclusion']['action'] in ['BUY', 'SELL']:
        create_trading_signal(
            ticker=news.ticker,
            action=result['stage3_conclusion']['action'],
            confidence=result['stage3_conclusion']['confidence'],
            source='deep_reasoning',
            reasoning=result['stage3_conclusion']['reasoning'],
            metadata=result
        )
    
    return result

Performance Metrics

• Analysis Time: 평균 20-30초 (깊이 우선)
• Accuracy: 목표 > 75% (Quick Analyzer 60% 대비 높음)
• Related Tickers Recall: > 90% (주요 영향 티커 포착)
• User Satisfaction: > 4.5/5 (깊이감)

Comparison

Agent	Speed	Depth	Accuracy	Use Case
Quick Analyzer	5초	⭐	60%	빠른 확인
Deep Reasoning	30초	⭐⭐⭐	75%	중요한 결정
War Room	15초	⭐⭐	65%	합의 기반

Version History

• v1.0 (2025-12-21): Initial release with 3-stage Chain of Thought methodology