Overview

Previous: #5 — Backend Stabilization and Data Pipeline Improvements

This sprint (#6) ran three major work streams across 35 commits. First, a sixth expert (news/macro analyst) was added to the signal pipeline and DART-based analysis was significantly expanded. Second, advanced analysis features — DCF valuation, portfolio risk, signal history — were implemented. Third, the frontend received a large-scale expansion: ScheduleManager, SignalDetailModal, and investment memo export.

Signal Pipeline Expansion

News/Macro Analyst — The Sixth Expert

Added a news/macro analyst to the existing five-expert lineup (technical, fundamental, sentiment, flow, risk). Uses Google News RSS as a fallback to improve news collection reliability.

graph TD
    A["Signal Pipeline"] --> B["Technical Analyst"]
    A --> C["Fundamental Analyst"]
    A --> D["Sentiment Analyst"]
    A --> E["Flow Analyst"]
    A --> F["Risk Analyst"]
    A --> G["News/Macro Analyst"]
    G --> H["Google News RSS"]
    G --> I["DART Disclosures"]
    B --> J["Combined Signal"]
    C --> J
    D --> J
    E --> J
    F --> J
    G --> J

Major DART Data Expansion

Significantly expanded the data pulled from the DART electronic disclosure system:

• Insider trading data (feat: add DART insider trading data) — executive buy/sell trends
• Foreign/institutional investor trends (feat: add foreign/institutional investor trend) — fund flow analysis
• Catalyst calendar (feat: add catalyst calendar with DART disclosures) — earnings announcements and disclosure schedules visualized as a timeline UI
• Peer comparison (feat: add peer comparison with sector-based DART valuation) — sector-level valuation benchmarking

8 New Database Tables

Added 8 tables, an ANALYST role, and metadata initialization in a single migration to support the new analysis features.

Advanced Analysis Features

DCF Valuation

Implemented a Discounted Cash Flow valuation module with sensitivity tables and heatmaps that visualize the fair value range across WACC and growth rate combinations.

# DCF sensitivity heatmap core logic
for wacc in wacc_range:
    for growth in growth_range:
        intrinsic_value = calculate_dcf(fcf, wacc, growth, terminal_growth)
        heatmap[wacc][growth] = intrinsic_value

Portfolio Risk Analysis

Calculates VaR (Value at Risk), beta, and sector concentration from real portfolio data. Renders a correlation matrix heatmap for cross-position correlation analysis.

• VaR: Historical simulation approach, 95%/99% confidence intervals for maximum loss estimation
• Beta: Portfolio beta relative to KOSPI200
• Sector concentration: Collects KOSPI200 sector data from Naver Finance for sector distribution analysis

Signal History Snapshots

Added point-in-time signal storage and a timeline feature for comparing against historical signals.

Frontend Expansion

ScheduleManager

Implemented a schedule management component with cron editing and a run-now button. Includes agent name display, friendly task labels, and sorting by cron time (hour:minute).

SignalDetailModal

Added a detail modal that lets users drill down from a signal into associated order history. Includes expert opinion expansion, risk_notes display, and compact/expanded view toggle.

Investment Memo Export

Added investment memo generation in HTML and DOCX formats based on signal data. Uses python-docx for Word document export.

Other UI Improvements

MCP Middleware Fix

Discovered that ctx.set_state() and ctx.get_state() are async methods but were being called without await in Session 1, causing repeated “MCP tool call failed” errors in server logs.

# Before
ctx.set_state(factory.CONTEXT_STARTED_AT, started_dt.strftime("%Y-%m-%d %H:%M:%S"))

# After
await ctx.set_state(factory.CONTEXT_STARTED_AT, started_dt.strftime("%Y-%m-%d %H:%M:%S"))

Also added auto-reconnect logic so MCP connection failures recover automatically.

Unit Tests

Added unit tests for the DCF valuation and portfolio risk services.

Commit Log

Key Takeaways

This sprint was a large-scale expansion that simultaneously deepened the analysis layer and raised the frontend polish. Adding the sixth expert rounds out the signal pipeline for more balanced decision-making. DCF valuation, VaR, and beta give the system institutional-grade analytical tools — it’s evolving from a signal generator into a comprehensive investment analysis platform. Expanding DART data coverage to insider trading, fund flows, and disclosure calendars sharpens the differentiation of this agent for the Korean equity market.