nanograph industry intel — SPIKE schema bootstrap

Stand up a SPIKE knowledge-graph schema on nanograph. One worked ontology ships pre-built (AI industry intel); for anything else the user brings, design the ontology together using the AI set as a pattern. This skill is ontology-focused: data ingestion, research, and seed generation are handled by the nanograph-ops skill, not here.

• AI path — use the template as-is (schema + real 2026 seed data). Ready to query in under a minute.
• Custom path — collaborate with the user to design domain-specific enums (entity kinds, sub-fields, actor roles, source/artifact types), adapt schema.pg, initialize an empty database. Works for Biotech, Fintech, Space, Crypto, Geopolitics, or any other domain.
• Ask the user for 2–3 example source types they already read — just enough to shape SourceEntity.type and artifactType. Not a full source list, not an ingestion plan.
• Two reference files: references/domain-examples.md explains the AI ontology as a worked pattern; references/schema-adaptation.md covers keep-vs-change rules and nanograph syntax gotchas.

Prerequisites

brew install nanograph          # or see https://nanograph.io
nanograph version               # needs 1.2.0+

An embedding API key in .env.nano is required before nanograph load, not just at query time. Every SPIKE node (Signal, Element, Pattern, Insight, KnowHow) has embedding: Vector(3072)? @embed(brief) @index, and nanograph embeds inline during load for any @embed(...) field whose source value is set and whose vector column is null. No API key → load fails with a cryptic embedding initialization failed error.

The template defaults to Gemini (GEMINI_API_KEY) — also covers multimodal embedding. OPENAI_API_KEY works too if you switch [embedding].provider in nanograph.toml. For CI or offline validation of just the schema/graph structure, set provider = "mock" temporarily.