portfolio-state

portfolio-state

Load + validate the full portfolio state and enrich with live quotes and FX.

Steps

1. Check staleness. Read portfolio/holdings.yaml. If as_of is older than 7 days, emit a warning:

   "holdings.yaml is N days old — run holdings-import before relying on this snapshot."

2. Load and validate by running the state loader:

python -c "
import json
from dataclasses import asdict
from utils.state import load_snapshot
s = load_snapshot('portfolio')
print(json.dumps({
  'as_of': s.as_of, 'fx_usdinr': s.fx_usdinr,
  'accounts': [asdict(a) for a in s.accounts],
  'lots': [asdict(l) for l in s.lots],
  'targets': s.targets, 'constraints': s.constraints,
  'universes': s.universes,
}, default=str))
"

3. Fetch live quotes. For every unique symbol in lots, call mcp__market-data__quote. Batch where possible.

4. Fetch FX. Call mcp__market-data__fx_usdinr. If the returned rate differs from holdings.yaml.fx.USDINR by >1%, note the drift in the output.

5. For MF lots (instrument == equity_mf), fetch AMFI NAV:

curl -sSL https://www.amfiindia.com/spages/NAVAll.txt > /tmp/amfi-nav.txt
python -c "
import json
from utils.amfi_nav import parse_amfi_nav_file
print(json.dumps(parse_amfi_nav_file('/tmp/amfi-nav.txt')))
" > /tmp/amfi-nav.json

Lookup MF lots by ISIN (stored in symbol field for MF).

6. Compute derived fields:

   • market_value_native = qty × price
   • market_value_inr = native * FX (for USD) or native (for INR)
   • current_weight = market_value_inr / total
   • drift_vs_target per asset class per targets.yaml

7. Validate invariants. Fail loudly if any lot has no entry_date, any symbol has no price, or any account referenced in a lot is not in accounts.yaml.

Output

Return a PortfolioSnapshot JSON blob to the caller.

Rules

• Every number traces to a source: holdings.yaml, a tool call, or a rules file.
• Stale holdings (>7 days) produces a warning, not a silent pass.