Sourced financial inputs and order readiness
build_entry_signals.py --financials-json financials.json supplements FMP data
with user-asserted financial metrics. It still fetches quotes, profiles, GAAP
EPS, FCF and dividends through FMP; this flag does not enable offline screening.
Only the listed fields are accepted. No document is fetched or verified.
{
"schema_version": 1,
"tickers": {
"PG": {
"adjusted_eps": {
"value": 6.0,
"unit": "USD/share",
"source": "Fictional issuer annual report",
"as_of": "2026-05-01",
"document_ref": "Fictional report, p. 10"
}
},
"DUK": {
"utility_metrics": {
"ffo_to_debt": {
"value": 0.18,
"unit": "fraction",
"source": "Fictional issuer credit presentation",
"as_of": "2026-05-01",
"document_ref": "Fictional presentation, p. 5"
}
}
}
}
}Each metric requires exactly value, unit, nonblank source, as_of
(YYYY-MM-DD, no later than the run's --as-of) and nonblank document_ref.
The source and reference are user assertions: document existence, contents,
publication availability and freshness are not verified. Confirm those manually.
EPS and AFFO are annual per-share values in the same USD basis as the dividend;
foreign-currency values and currency conversion are unsupported. Ratios use
fractions (0.12 means 12%), never percentage points. A fraction greater than 1
is accepted except for CET1, so an adverse combined ratio is still assessed.
| Field | Unit | Valid value |
|---|---|---|
adjusted_eps |
USD/share |
positive finite number |
reit_metrics.affo_per_share |
USD/share |
positive finite number |
bank_metrics.cet1 |
fraction |
finite 0 through 1 |
bank_metrics.deposit_beta |
fraction |
finite, nonnegative |
utility_metrics.ffo_to_debt |
fraction |
finite, nonnegative |
insurer_metrics.combined_ratio |
fraction |
finite, nonnegative |
bank_metrics.cre_concentration |
category |
low, moderate, medium, high |
bank_metrics.npl_trend, bank_metrics.nco_trend |
category |
improving, stable, deteriorating |
utility_metrics.allowed_roe_trend |
category |
rising, stable, falling, adverse |
utility_metrics.rate_case_status |
category |
constructive, neutral, pending, adverse, pending_adverse |
utility_metrics.equity_issuance_risk |
category |
low, moderate, high, dilutive |
insurer_metrics.reserve_development |
category |
favorable, neutral, adverse |
insurer_metrics.statutory_capital |
category |
strong, adequate, weak |
JSON numbers must be finite and cannot be booleans or strings. Valid inputs
merge into the provider financials; unrelated values are retained. Output
provenance.manual_financials records each accepted value and its metadata with
input_method: MANUAL. Adjusted EPS also sets adjusted_eps_source: MANUAL.
Missing ticker entries retain existing missing-data gates. Invalid metric
records, unknown fields and empty ticker records produce explicit
manual_financials:* blockers and provenance.manual_financials_errors.
They cap payout safety at HOLD-REVIEW; an existing FAIL stays FAIL. An invalid
manual value never silently restores readiness using a provider fallback.
Malformed file structure, duplicate JSON keys at any depth, or colliding ticker
names (case/whitespace normalized) exit 2 before provider calls or reports.
order_ready is the aggregate screening readiness field, present in JSON,
CSV and Markdown. It is true only when timing is TRIGGERED, verdict is
CLEAN-PASS or CONDITIONAL-PASS, final t1_blocked is false, and no
pre_order_blockers remain. WAIT, ASSUMPTION-REQUIRED, PASS-CAUTION and missing
Step-1/safety/event evidence are not ready. A utility with an open FFO/debt
blocker is never ready, even if its legacy T1 flag is false.
event_t1_blocked means the event cap alone blocks the first tranche.
t1_blocked retains its legacy meaning: final T1 restriction after verdict
synthesis (it may be false while a pre-order blocker remains). Neither replaces
order_ready. A ready screening result is not a broker authorization or trade
instruction; human review and existing order gates still apply.