market-view

A queryable property graph of the equity universe + an LLM orchestration layer that answers multi-hop, explainable financial questions with citations.

Why a graph

Traditional equity research relies on time-series forecasting and per-company analysis. These methods produce numbers but not explanations. They cannot answer questions like "why is NVDA down 6% while AMD is up 2%?" or "if Iran cuts off Hormuz, which of my holdings face the largest second-order risk?" — those are relational, multi-hop, contextual questions. A property graph answers them; a single time series cannot.

Why a graph + LLM

A graph alone returns nodes and edges. An LLM alone hallucinates. The combination — graph for ground truth, LLM for natural-language reasoning over the graph — gives both depth and explainability:

• Graph provides the LLM with: entities to ground claims to, relationships to traverse, provenance for every assertion.
• LLM provides the graph with: a natural-language interface, inference over patterns the schema didn't anticipate, coherent narrative composition.

Operating principle

Every factual claim in an LLM response must cite at least one node or edge in the graph. If the LLM cannot cite, the LLM cannot claim.

Where to go next

• Quickstart — bring up the local stack and run your first query.
• Architecture — full diagram, free-tier hosting plan, and the 6-stage LLM pipeline.
• ADR 0001 — the architecture decisions that shape this build.