`@anvia/graph`

@anvia/graph is the provider-neutral knowledge-graph layer for Anvia. It owns typed graph schemas, fact extraction, raw-text ingestion, retrieval contracts, graph search tools, and bounded exploration. Database connections, provisioning, persistence, and query execution stay in an adapter such as @anvia/neo4j or @anvia/memgraph.

Install

pnpm add @anvia/graph @anvia/core zod

Install one graph adapter alongside it:

pnpm add @anvia/neo4j
# or
pnpm add @anvia/memgraph

Portable workflow

text

raw text → deterministic chunks → typed fact extraction
         → chunk/entity embeddings → managed graph write

query → vector or hybrid seeds → bounded traversal → optional source evidence

The same schema, ingestion helpers, retrieval options, result types, and Agent tool factory work with either first-party adapter. Provider-specific configuration is limited to the client and graph registration.

Main entry points

API	Purpose
`defineGraphSchema()`	Define typed nodes, relationships, identities, and strict property schemas.
`extractGraphFacts()`	Extract schema-valid entities, relationships, and source mentions.
`ingestGraphText()`	Chunk, extract, embed, and replace one raw-text source document.
`ingestGraphDocuments()`	Apply the same workflow to a batch of documents.
`prepareGraphDocuments()`	Prepare graph and reusable vector records without writing them.
`createGraphSearchTool()`	Expose any compatible graph retriever as an Agent tool.
`GraphExplorer`	Read bounded overviews and neighborhoods for visualization.

@anvia/graph ​

Install ​

Portable workflow ​

Main entry points ​

Continue ​

`@anvia/graph`

Install

Portable workflow

Main entry points

Continue