aEquity — Autonomous Equity Analyst

🔗 Live Demo

aEquity analyzes S&P 500 stocks through four guru-driven lenses: business quality, competitive moat, financial health, and governance. It combines quantitative metrics from yfinance, qualitative signals from SEC 10-K filings, and Claude LLM scoring into a 0–100 scorecard.

Prerequisites

• Python 3.11+
• Anthropic API key
• A real email address for SEC EDGAR access (required by EDGAR terms of service)

Install

pip install -e ".[dev]"
cp .env.example .env
# Edit .env and fill in ANTHROPIC_API_KEY and SEC_USER_AGENT_EMAIL

Quick Start

Single stock (CLI):

python main.py AAPL

Streamlit dashboard:

streamlit run app.py

Batch analysis:

python batch.py --limit 20          # analyze 20 S&P 500 companies
python batch.py --tickers AAPL MSFT GOOG

Architecture

main.py          CLI entry point
app.py           Streamlit dashboard (Analyze / Screener / Macro Radar tabs)
batch.py         Batch runner — populates the SQLite DB
pipeline.py      Core analysis: fetch → calculate → LLM → score → assemble
config.py        Settings loaded from .env via pydantic-settings
models.py        Pydantic schemas (CompanyAnalysis, PillarAnalysis, GuruScorecard)
scoring_config.py  All tunable constants (weights, thresholds, model IDs)

tools/
  calculator_tools.py  Financial ratio calculations (yfinance)
  sec_tools.py         SEC EDGAR 10-K downloader and section extractor
  validator.py         Ticker symbol validation

db/
  init.py        SQLite helpers (open_db, save_analysis, get_all_latest)

Four Pillars

Pillar	What it measures
The Engine	Business quality — ROIC, gross margin
The Moat	Competitive defensibility — LLM analysis of 10-K text
The Fortress	Financial health — FCF conversion, net debt/EBITDA
Alignment	Governance — insider ownership, shareholder yield

Virtual Investment Committee

Each stock is scored from the perspective of four legendary investors:

• Warren Buffett — ROIC, FCF conversion, moat, leverage
• Peter Lynch — PEG ratio, earnings growth, business understandability
• Ben Graham — P/B ratio, current ratio, earnings stability
• Aswath Damodaran — ROIC, PEG, leverage (proxy for reverse-DCF)

Configuration

All tunable constants (weights, normalization ranges, model IDs, verdict boundaries) live in scoring_config.py. See .env.example for environment variables.

Deployment (Railway)

The app is deployed on Railway. Each analysis triggers real Anthropic API calls (~$0.05–0.20 each). To prevent unbounded spend on the public URL, set ANALYZE_ACCESS_TOKEN in Railway → Variables. When set, the Analyze tab requires the token before any analysis can run. The Screener and Watchlist tabs remain public (they read from the pre-populated DB, no API calls).

ANALYZE_ACCESS_TOKEN=your-secret-token   # set in Railway Variables, not in .env

Development

ruff check .                          # lint
mypy --ignore-missing-imports .       # type check
pytest tests/ -v                      # run tests
pytest tests/ --cov=. --cov-report=term-missing  # with coverage

Further Reading

See docs/Plan.md for the full design rationale and roadmap.