What is a financial knowledge graph?
A financial knowledge graph is a connected view of the market: which companies compete, who supplies whom, which ETFs hold what, and which macro factors a name is exposed to. It turns an answer about one ticker into an answer about the neighborhood it lives in.
Why connections beat endpoints
Data commoditizes and models commoditize, but a connected view is where an answer stops being about a single ticker. Knowing that a name sits in a sector, competes with three peers, is held by two ETFs, and is exposed to specific macro factors lets every other engine ask "what else should I look at?"
Relationships are typed edges: in_sector, held_by_etf, peer_of, supplied_by, led_by, exposed_to, recent_filing, and traded_by_congress, each with a human phrasing.
The cascade: macro to company
Financial intelligence is a network, not a set of endpoints. A policy shift or a macro move cascades down a sector to specific names. LyraMind's Context Graph walks up from a company to the macro and policy drivers that move it, then out to the peers that share those drivers, producing impact paths that read driver to sector to company.
This top-down structure is the connected view a generic LLM wrapper cannot reconstruct from a single ticker query.
How LyraMind implements it
Offline, the graph is a curated seed of well-known relationships plus edges derived from it, clearly labeled as seed. It is deterministic and records nothing. Live sources merge over the seed behind one register_edges contract: Polygon (peers and sector, key-gated), SEC EDGAR (industry and recent filings, flag-gated with a required contact user-agent), and congressional trades via Financial Modeling Prep (key-gated).
Because every source uses the same seam, the graph queries are unchanged whoever populates the nodes. Missing keys or network errors simply leave the seed untouched, so offline runs and tests behave identically.
Related concepts
The knowledge graph supplies peer cohorts for the AI Readiness benchmark, drivers for the institutional simulation's predicted questions, and the connections surfaced by a company's Digital Twin.
- Typed edges connect companies to sectors, ETFs, peers, suppliers, executives, and macro factors
- The cascade walks company to sector to macro/policy drivers, then out to shared peers
- Offline it is a clearly labeled curated seed; deterministic and records nothing
- Live sources merge over the seed via one register_edges seam
- Polygon (peers, sector), SEC EDGAR (industry, filings), and FMP (congressional trades) plug in
- Missing keys leave the seed untouched, keeping offline runs reproducible
Questions
Is the graph populated with live data?
The default is a curated seed of well-known relationships. Polygon, SEC EDGAR, and FMP sources merge live edges when their keys or flags are configured, using the same register_edges contract.
What is the cascade?
A top-down traversal that maps how a macro or policy shift flows through a name's sector to the company and its peers, as explicit impact paths.
Does the graph give a buy or sell view?
No. It describes relationships in the data. It carries no recommendation.