Eda Visual Analysis

• MA10 — 10-day moving average of closing prices
• MA20 — 20-day moving average of closing prices

Step 3 — Build Chart Configurations

Chart A: Price & Moving Averages (Line chart, full-width)

Title: "{TICKER} — 30-Day Price & Moving Averages"
Datasets:
  - Price line (cyan #00d4ff, filled, tension 0.3, no point markers)
  - MA10 dashed line (amber #f59e0b, dash [4,4])
  - MA20 dashed line (green #10b981, dash [8,4])

Chart B: Volume Analysis (Bar + Line combo)

Title: "{TICKER} — Volume Analysis"
Datasets:
  - Volume bars — red (rgba(239,68,68,0.7)) if volume > avgVolume × 1.3, else cyan (rgba(0,212,255,0.4))
  - Avg Volume reference line (amber dashed)

Chart C: Analyst Consensus (Doughnut, 65% cutout)

Title: "{TICKER} — Analyst Consensus"
Labels: ['Strong Buy', 'Buy', 'Hold', 'Sell', 'Strong Sell']
Colors: ['#10b981', '#6ee7b7', '#f59e0b', '#f87171', '#dc2626']
Data: [strongBuy, buy, hold, sell, strongSell]

Chart D: News Sentiment (Bar chart)

Title: "{TICKER} — News Sentiment"
Labels: news source names
Dataset: sentiment scores per source
Colors: green (rgba(16,185,129,0.7)) if score > 0, red (rgba(239,68,68,0.7)) if score ≤ 0
Y-axis range: −1 to +1

Step 4 — LLM Analysis (EDA Insights)

Send a prompt to the LLM containing the following market data summary:

• Current price, daily change %, RSI, Trend, Sentiment score/label, MA20, MA50, MA200
• Macro risk level, macro sentiment tone, and a short macro market-context string
• A compact list of top recent headlines with their precomputed sentiment scores
• Technical indicator summary (MACD, Bollinger Bands, KDJ, OBV, VWAP, ATR/VaR when available)

Request the LLM to return a JSON object:

• insights — array of 4 key EDA observations (plain English, quantitative where possible)
• riskFlags — array of risk flags (e.g., "Overbought — potential pullback risk")
• technicalSummary — 1–2 sentences synthesising the technical picture
• momentumSignal — one of: POSITIVE, NEGATIVE, NEUTRAL

The LLM response must be JSON only. If the provider returns extra prose or fenced code, normalize and parse the structured payload before using it.

Fallback (if LLM unavailable)

Compute insights deterministically:

1. RSI zone analysis (> 70 → overbought, < 30 → oversold, else healthy range)
2. Price deviation from MA50 (%)
3. Sentiment label + score
4. Analyst upside potential (%)
5. Preserve locally computed edaFactors derived from breakout, volume regime, volatility regime, and trend-strength calculations

Step 5 — Return Response

{
  "charts": {
    "priceChart": { "type": "line", "title": "...", "data": { "labels": [], "datasets": [] } },
    "volumeChart": { "type": "bar",  "title": "...", "data": { "labels": [], "datasets": [] } },
    "analystChart": { "type": "doughnut", "title": "...", "data": { "labels": [], "datasets": [] } },
    "sentimentChart": { "type": "bar", "title": "...", "data": { "labels": [], "datasets": [] } }
  },
  "edaInsights": {
    "insights": ["...", "...", "...", "..."],
    "riskFlags": [],
    "technicalSummary": "...",
    "momentumSignal": "POSITIVE"
  },
  "skillUsed": "eda-visual-analysis"
}

References

• See references/chart-types.md for Chart.js configuration patterns and theming guidelines.