code/run_pipeline.py

#!/usr/bin/env python
"""Run the full MacroLens benchmark pipeline end-to-end.

Usage (foreground):
    uv run python -m projects.agent_builder.scripts.whatif_bench.run_pipeline

Usage (background with log):
    nohup uv run python -m projects.agent_builder.scripts.whatif_bench.run_pipeline \
        > whatif_pipeline.log 2>&1 &

The script mirrors the notebook's steps across 3 layers (plus XBRL
ontology construction), running non-interactively with full logging.
Each step is resumable -- if it detects existing output files it will skip.

To force a full re-run, delete the data/ directory first.
"""

from __future__ import annotations

import asyncio
import logging
import os
import sys
import time
from pathlib import Path

# Ensure project root is importable
_ROOT = Path(__file__).resolve().parents[4]  # platform/
if str(_ROOT) not in sys.path:
    sys.path.insert(0, str(_ROOT))

# Load .env from project root (contains API keys for SEC, FRED, EIA, RentCast)
_ENV_FILE = _ROOT / ".env"
if _ENV_FILE.exists():
    from dotenv import load_dotenv
    load_dotenv(_ENV_FILE, override=False)

logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s [%(levelname)s] %(name)s: %(message)s",
    handlers=[logging.StreamHandler(sys.stdout)],
)
logger = logging.getLogger("whatif_pipeline")


def _elapsed(start: float) -> str:
    secs = time.time() - start
    if secs < 60:
        return f"{secs:.1f}s"
    mins = secs / 60
    if mins < 60:
        return f"{mins:.1f}m"
    return f"{mins / 60:.1f}h"


def main() -> None:
    pipeline_start = time.time()
    logger.info("=" * 60)
    logger.info("MacroLens Benchmark Pipeline -- STARTING")
    logger.info("=" * 60)

    # ================================================================
    # Layer 1: Raw Data Collection
    # ================================================================

    # Step 1: Universe
    logger.info("--- Step 1: Collecting ticker universe ---")
    t0 = time.time()
    from projects.agent_builder.scripts.whatif_bench import collect_universe
    universe_df = collect_universe.run()
    logger.info("Step 1 done in %s: %d tickers", _elapsed(t0), len(universe_df))

    tickers = universe_df["ticker"].tolist()

    # Step 2: Fundamentals
    logger.info("--- Step 2: Collecting fundamentals ---")
    t0 = time.time()
    from projects.agent_builder.scripts.whatif_bench import collect_fundamentals
    collect_fundamentals.run(tickers=tickers)
    logger.info("Step 2 done in %s", _elapsed(t0))

    # Step 3: Daily prices
    logger.info("--- Step 3: Collecting daily prices ---")
    t0 = time.time()
    from projects.agent_builder.scripts.whatif_bench import collect_prices
    collect_prices.run(tickers=tickers)
    logger.info("Step 3 done in %s", _elapsed(t0))

    # Step 4: SEC filings (async)
    logger.info("--- Step 4: Collecting SEC filings ---")
    t0 = time.time()
    from projects.agent_builder.scripts.whatif_bench import collect_filings
    asyncio.run(collect_filings.run_async(tickers=tickers))
    logger.info("Step 4 done in %s", _elapsed(t0))

    # Step 5: Macro data (async)
    logger.info("--- Step 5: Collecting macro data ---")
    t0 = time.time()
    from projects.agent_builder.scripts.whatif_bench import collect_macro
    asyncio.run(collect_macro.run_async())
    logger.info("Step 5 done in %s", _elapsed(t0))

    # Step 6: Real estate (async)
    logger.info("--- Step 6: Collecting real estate data ---")
    t0 = time.time()
    from projects.agent_builder.scripts.whatif_bench import collect_real_estate
    asyncio.run(collect_real_estate.run_async())
    logger.info("Step 6 done in %s", _elapsed(t0))

    # Step 4b: Build XBRL ontology (from XBRL facts collected in Step 4)
    logger.info("--- Step 4b: Building XBRL industry ontology ---")
    t0 = time.time()
    from projects.agent_builder.scripts.whatif_bench import build_ontology
    ontology_summary = build_ontology.run()
    logger.info("Step 4b done in %s: %s", _elapsed(t0), ontology_summary)

    # ================================================================
    # Layer 2: Preprocessing  (all granularities)
    # ================================================================

    from projects.agent_builder.scripts.whatif_bench import preprocess

    _GRANULARITIES = ["daily", "weekly", "monthly"]

    for gran in _GRANULARITIES:
        logger.info("--- Step 7: Preprocessing (%s) ---", gran)
        t0 = time.time()
        panel_df = preprocess.run(granularity=gran)
        logger.info(
            "Step 7 done (%s) in %s: %d rows, %d columns",
            gran, _elapsed(t0), len(panel_df), len(panel_df.columns),
        )

    # ================================================================
    # Layer 3: Benchmark Construction  (all granularities)
    # ================================================================

    from projects.agent_builder.scripts.whatif_bench import assemble_benchmark
    from projects.agent_builder.scripts.whatif_bench import generate_scenarios

    for gran in _GRANULARITIES:
        # Step 8: Assemble benchmark
        logger.info("--- Step 8: Assembling benchmark (%s) ---", gran)
        t0 = time.time()
        assemble_benchmark.run(granularity=gran)
        logger.info("Step 8 done (%s) in %s", gran, _elapsed(t0))

        # Step 9: Generate scenarios
        logger.info("--- Step 9: Generating scenarios (%s) ---", gran)
        t0 = time.time()
        generate_scenarios.run(granularity=gran)
        logger.info("Step 9 done (%s) in %s", gran, _elapsed(t0))

    # Step 10: News collection (after scenarios exist)
    logger.info("--- Step 10: Collecting news ---")
    t0 = time.time()
    from projects.agent_builder.scripts.whatif_bench import collect_news
    asyncio.run(collect_news.run_async(tickers=tickers))
    logger.info("Step 10 done in %s", _elapsed(t0))

    # Step 11: Enrich benchmark with news-derived features (all granularities)
    from projects.agent_builder.scripts.whatif_bench import enrich_benchmark
    for gran in _GRANULARITIES:
        logger.info("--- Step 11: Enriching benchmark (%s) ---", gran)
        t0 = time.time()
        enrich_benchmark.run(granularity=gran)
        logger.info("Step 11 done (%s) in %s", gran, _elapsed(t0))

    # Step 12: Build valuation benchmark ground truth (all granularities)
    logger.info("--- Step 12: Building valuation benchmark ---")
    for gran in _GRANULARITIES:
        t0 = time.time()
        try:
            from projects.agent_builder.scripts.whatif_bench.build_valuation_tasks import (
                build_valuation_benchmark,
            )
            summary = build_valuation_benchmark(granularity=gran)
            logger.info("Step 12 done (%s) in %s: %s", gran, _elapsed(t0), summary)
        except Exception:
            logger.warning("Step 12 (%s) skipped or failed:", gran, exc_info=True)

    # ================================================================
    # Done
    # ================================================================
    logger.info("=" * 60)
    logger.info("Pipeline COMPLETE in %s", _elapsed(pipeline_start))
    logger.info("=" * 60)


if __name__ == "__main__":
    main()