News - Unified Daily Market & Portfolio Report

One command pulls live Schwab + IBKR holdings, fetches ~800 ETF prices from Yahoo Finance (no Bloomberg terminal), runs unit-tested analytics, and has Claude Opus write a single unified market + portfolio report as a PDF:

python3 report/main.py

Full documentation: report/README.md

python3 report/main.py --no-llm           # data + analytics only (free, fast)
python3 report/main.py --non-interactive  # cron mode (stale fallback, no prompts)
python3 -m pytest tests/ -v               # 25 tests on the financial math

Repository layout

News/
├── report/                          # THE daily report system (start here)
│   ├── main.py                      #   one-command pipeline
│   ├── prompts/system.md            #   LLM strategist prompt
│   └── README.md                    #   full docs
├── tests/                           # pytest suite for the financial math
├── data/                            # universe.xlsx, report.db, holdings.xlsx
├── outputs/unified/                 # generated reports (PDF/MD + data packages)
│
├── Step 1 Data Collection/          # Universe construction (run rarely)
├── Step 2 Data Processing - Final1000/  #   classification & selection
├── Step 3 Data Analysis/            #   analytics on the universe
├── fine tuning/                     #   ML classifier training
│
├── archive/                         # Legacy reporting chain (replaced by report/)
├── AGENTS.md                        # AI agent instructions
└── README.md                        # this file

The two pipelines

1. Daily report (report/) - run every day

See report/README.md. Replaces the old Phase 0 -> Step 4 -> Phase 2 chain (now in archive/). Universe: 763 unique ETFs in data/universe.xlsx, priced via Yahoo Finance. History: single SQLite db at data/report.db.

2. Universe construction (Steps 1-3 + fine tuning) - run rarely

Builds the Final 1000 Asset Master List from Bloomberg indices, ETFs, and Goldman baskets via LLM classification. Only needed to rebuild/refresh the universe; after changing it, regenerate the report universe with:

python3 report/build_universe.py

Stage	Purpose	Output
Step 1 Data Collection	Raw data acquisition	Filtered datasets from 3 sources
Step 2 Data Processing	Classification & selection	Final 1000 Asset Master List
Step 3 Data Analysis	Performance analytics	Factor profiles & deduplication
Fine Tuning	ML model training	Fine-tuned Llama classifier

Requirements

• Python 3.14 (yfinance, pandas, anthropic>=0.109, schwabdev, python-dotenv, markdown, pytest)
• .venv-ibkr312/ - Python 3.12 venv with ib_insync (IBKR API requirement)
• PrinceXML (brew install prince) for PDF rendering
• .env at repo root: ANTHROPIC_API_KEY, SCHWAB_APP_KEY, SCHWAB_APP_SECRET
• TWS / IB Gateway logged in (auto-launched if not running)

Documentation

File	Purpose
report/README.md	Daily report system - full guide
AGENTS.md	AI coding agent instructions
archive/README.md	What the legacy code was and what replaced it
Step 2 .../README.md	Classification workflow
fine tuning/README.md	ML training guide

---

Last Updated: 2026-06-10 Version: 3.0.0 (unified report rearchitecture)