AlphaEar Logic Visualizer Skill

This is an institutional-quality research and investment skill.

Overview

This skill creates visual representations of investment logic flows and financial data. It supports two rendering paths:

1. Inline rendering (preferred in claude.ai) — use the native show_widget / Visualizer tool to render SVG or HTML directly in chat
2. File rendering — use scripts/visualizer.py to save an HTML file for download

Use inline rendering for simple diagrams and quick visualizations. Use file rendering for complex multi-panel outputs or when the user explicitly wants a downloadable file.

Chart Types & When to Use Each

1. Logic Transmission Graph (generate_transmission_graph)

Best for: Investment signal chains, macro transmission mechanisms, thesis visualizations, cause-and-effect maps.

Renders an interactive force-directed graph via pyecharts. Nodes are colour-coded by impact type:

• 🟢 Green — bullish / positive impact ("Positive")
• 🔴 Red — bearish / negative impact ("Negative")
• ⚫ Grey — neutral ("Neutral")

Input — nodes_data (list of dicts):

[
  {
    "node_name": "Fed Rate Cut",
    "impact_type": "Positive",
    "logic": "Lower real yields → risk-on",
    "id": "1"
  },
  {
    "node_name": "USD Weakens",
    "impact_type": "Positive",
    "logic": "Dollar index falls as rate differential narrows",
    "id": "2",
    "parent_id": "1"
  },
  {
    "node_name": "Gold Rallies",
    "impact_type": "Positive",
    "logic": "Inverse USD relationship + safe haven demand",
    "id": "3",
    "parent_id": "2"
  }
]

Tool call:

from scripts.visualizer import VisualizerTools
graph = VisualizerTools.generate_transmission_graph(nodes_data, title="Fed Cut → Gold Transmission")
VisualizerTools.render_chart_to_file(graph, filename="output/chain.html")

2. Draw.io Diagram (render_drawio_to_html)

Best for: Structured flowcharts, decision trees, process diagrams, architecture-style layouts that need to be shared as a diagrams.net-compatible HTML file.

Workflow:

1. Use the Draw.io XML generation prompt in references/PROMPTS.md to produce valid <mxGraphModel> XML
2. Pass the XML to render_drawio_to_html(xml_content, filename)

Tool call:

VisualizerTools.render_drawio_to_html(
    xml_content="<mxGraphModel>...</mxGraphModel>",
    filename="output/diagram.html",
    title="My Logic Diagram"
)

The output HTML embeds the diagrams.net viewer JS — the user can open it in any browser.

Node colour convention:

• Positive → fillColor=#d5e8d4 (green)
• Negative → fillColor=#f8cecc (red)
• Neutral → fillColor=#f5f5f5 (grey)

See references/PROMPTS.md for the full XML generation prompt and node/edge template.

3. Sentiment Trend Chart (generate_sentiment_trend_chart)

Best for: Rendering time-series output from the alphaear-sentiment skill — showing sentiment score evolution over time.

Input:

sentiment_history = [
    {"date": "2025-01-01", "score": 0.72},
    {"date": "2025-01-02", "score": -0.15},
]

Tool call:

chart = VisualizerTools.generate_sentiment_trend_chart(sentiment_history)
VisualizerTools.render_chart_to_file(chart, filename="output/sentiment.html")

4. ISQ Radar Chart (generate_isq_radar_chart)

Best for: Visualizing signal quality across five dimensions — sentiment strength, confidence, intensity, expectation gap, timeliness.

Tool call:

chart = VisualizerTools.generate_isq_radar_chart(
    sentiment=0.8,
    confidence=0.9,
    intensity=4,
    expectation_gap=0.6,
    timeliness=0.85,
    title="Signal Quality: NVDA Earnings"
)
VisualizerTools.render_chart_to_file(chart, filename="output/isq_radar.html")

5. Stock K-Line Chart (generate_stock_chart)

Best for: Rendering OHLCV candlestick charts with optional prediction overlays or ground truth comparison lines.

Input: A pandas.DataFrame with columns date, open, close, low, high, volume.

Tool call:

chart = VisualizerTools.generate_stock_chart(df, ticker="NVDA", title="NVDA Price Action")
VisualizerTools.render_chart_to_file(chart, filename="output/kline.html")

Decision Guide: Which Path to Use?

Agentic Workflow (file rendering path)

1. Parse user's logic description → extract nodes, links, impact types
2. Choose chart type using Decision Guide above
3. Structure input data (nodes_data dict list or DataFrame)
4. Call the appropriate VisualizerTools method
5. Call render_chart_to_file() or render_drawio_to_html() to save HTML
6. Present the file to the user via present_files tool

Dependencies

pandas
loguru
pyecharts

Install with:

pip install pandas loguru pyecharts --break-system-packages

Reference Files

• references/PROMPTS.md — Draw.io XML generation prompt and node/edge template (read this when using the Draw.io path)
• scripts/visualizer.py — All chart generation and rendering methods (VisualizerTools class)
• scripts/visualizer_prompt.py — Programmatic version of the Draw.io prompt for API-based generation