Features CSV contains rows with two date formats:
- Old rows: '2003-02-01 00:00:00' (with time)
- New rows: '2026-04-01' (date only)
pd.to_datetime() infers format from first rows then rejects the
date-only values with ValueError. Fix: format='mixed' across all
pd.to_datetime() calls in live_data_fetcher, portfolio_engine,
and explainability_engine.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Comprehensive date/error handling fixes across services
- Replace parse_dates with explicit pd.to_datetime() for consistent datetime64 type
- Normalize FRED series index to tz-naive to avoid offset-aware/naive comparison error
- Convert dt to tz-naive Timestamp in loop to prevent TypeError on > comparison
- Use pd.Timestamp for new row date so pd.concat keeps uniform datetime64 dtype
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Fix FRED missing dep and pandas 3.x date merge error
- Add fredapi to requirements.txt (was missing, caused monthly FRED fetch to fail)
- Fix pandas 3.x str/datetime64 merge error in portfolio_engine: replace
parse_dates with explicit pd.to_datetime() to guarantee consistent type
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
FastAPI + ML backend with regime detection, SHAP explainability,
stress testing, and live market data scheduler.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
