ai4finance-foundation-finrobot

name	ai4finance-foundation-finrobot
description	Produce FinRobot-inspired public-equity research with QVeris-routed company data. Use for evidence-backed company profiles, financial statement analysis, valuation, earnings-call review, analyst-expectation comparison, catalyst tracking, and structured investment research reports.

FinRobot-Inspired Equity Research with QVeris

Use the upstream FinRobot project as a methodology reference. Keep calculations, synthesis, and report generation local.

Data boundary

Acquire all live company, market, transcript, estimate, filing, and news data through QVeris. Require QVERIS_API_KEY for real calls. Use Discover → Inspect → Call and the capability plan in qveris.manifest.json; do not call data providers directly.

Workflow

1. Resolve the issuer, security, exchange, currency, fiscal year, and reporting period before collecting evidence.
2. Discover and call company profile, income statement, balance sheet, cash flow, derived ratios, earnings transcript, ratings and targets, and current news capabilities.
3. Record search_id, tool_id, provider, retrieval time, fiscal period, filing date, and currency for each data block.
4. Reconcile periods and units before calculating growth, margins, cash conversion, leverage, returns on capital, and valuation.
5. Separate reported facts, analyst estimates, management guidance, and local inference.
6. Compare the latest results with prior periods and available expectations. Identify both confirming and contradictory evidence.
7. Use structured XBRL as an optional verification layer. QVeris currently lacks native full filing text; disclose that gap and do not fabricate quotations.
8. Produce a balanced report covering business profile, financial quality, valuation, catalysts, risks, and evidence quality.

Quality gates

• Reject mismatched issuer identities, currencies, or fiscal periods.
• Label delayed or stale data.
• Never turn a price target into a guaranteed outcome.
• Keep primary facts and model-generated interpretation visibly separate.

Output contract

Return an as-of timestamp, company identity, QVeris tool trace, period-aligned financial tables, calculations, transcript and news evidence, expectation comparison, valuation assumptions, risk section, missing evidence, and research-only disclaimer.

Validation

node scripts/qveris_candidate_runner.mjs validate .
node --test tests/smoke.test.mjs
node scripts/qveris_candidate_runner.mjs preflight .
QVERIS_API_KEY=... node scripts/qveris_candidate_runner.mjs live .