Market Intelligence Agent

This project generates a weekday morning market-intelligence report for a configurable watchlist. Stage 1 supports watchlist validation, basic market-data collection, Treasury yield collection, SQLite persistence, Markdown reports, JSON output, and tests.

Installation

python -m venv .venv
. .venv/bin/activate
python -m pip install -e ".[dev]"

Configuration

• Edit config/watchlist.yaml to add or remove assets.
• Edit config/report_settings.yaml to change report time, timezone, score thresholds, or output paths.
• Edit config/sources.yaml to adjust provider cache and timeout settings.
• Copy .env.example to .env only when optional email or webhook delivery is needed.

TCIEX is included as configurable and marked requires_identity_verification: true. Stage 1 validates the shape of the ticker and flags the identity-verification requirement; it does not assert fund identity until a reliable fund source is added.

Data Sources

Stage 1 uses public data sources where available:

• Yahoo Finance via yfinance for equities, ETFs, and supported mutual funds.
• U.S. Treasury daily yield curve CSV data for Treasury yields.

Limitations: quote data may be delayed, mutual fund NAVs may be prior-day only, and some symbols may still be unavailable from free market-data coverage. Missing fields are preserved and labeled in reports instead of failing the whole run.

CLI Usage

python -m src.main validate-watchlist
python -m src.main collect
python -m src.main analyze
python -m src.main report
python -m src.main run-daily
python -m src.main backfill --days 30

Scheduled Automation

GitHub Actions cannot express one timezone-aware cron that automatically follows daylight saving time. The included workflow runs at both possible UTC times for 7:30 a.m. Eastern:

• 11:30 UTC, used during Eastern Daylight Time
• 12:30 UTC, used during Eastern Standard Time

The workflow then checks the current America/New_York local time and only runs the report when the local hour/minute equals 07:30.

Scheduled runs upload the reports/ directory as a workflow artifact and also commit the generated Markdown, JSON, and YTD chart assets back into the repository so the latest dated run output is visible in GitHub.

Testing

pytest
ruff check .

Known Limitations

• Stage 1 does not yet collect SEC filings, news, catalysts, or source-linked factual claims.
• Stage 1 reports use deterministic templates rather than a language model.
• Historical duplicate suppression and event scoring are scaffolded for Stage 2 but not yet populated by real news events.
• Free market-data coverage varies by symbol and may not include all mutual funds.

Adding Assets or Providers

Add assets to config/watchlist.yaml. Each entry should include symbol, name, asset_type, themes, and priority. Add new provider implementations under src/collectors/ by implementing the provider protocols in src/collectors/interfaces.py.