A knowledge graph for business AI models accounts, deals, contacts, tickets, and documents as typed entities and relationships — enabling multihop queries vector RAG cannot answer.
Vector-only RAG retrieves chunks — it does not join entities across CRM, contacts, tickets, and contracts. Gyri combines hybrid search with a typed graph and multihop GraphQL for business AI.
Related: RAG vs graph · Multihop GraphQL
Similarity search misses relationships — "which open deals share this churn signal across support and email?"
Traditional enterprise search returns documents, not correlated operational narratives.
RAG pipelines embed text but do not persist typed conclusions agents and humans can reuse.
CRM, email, Slack, and docs federated into a workspace graph — not doc-only search or single-app plugins.
Claim-level citations with links to source records — trace any answer back to source. Verify before customer-facing or legal use.
Typed insights in a knowledge graph accumulate across agents and quarters — not session resets.
Claude, Cursor, and internal agents search, synthesize, and write insights back to CRM and workflows.
| Capability | Gyri | Vector-only RAG | Keyword search |
|---|---|---|---|
| Typed entity graph | ✅ | ❌ | ❌ |
| Multihop GraphQL | ✅ | ❌ | ❌ |
| Hybrid keyword + graph retrieval | ✅ | ⚠️ | ⚠️ |
| Cited multihop synthesis | ✅ | ❌ | ❌ |
| Agent graph write-back | ✅ | ❌ | ❌ |
Named comparisons: RAG vs knowledge graph · Multihop GraphQL for BI
Traverse contacts → opportunities → threads → tickets in one graph query.
Correlate stale deals with support spikes and email silence — multihop, cited.
Cross-source entity linking for contracts, comms, and financial docs in one workspace.
Free trial — graph, search, and cited synthesis together.
Vector RAG retrieves similar chunks. Knowledge graphs join entities across systems — e.g. linking churn tickets to open renewals and email silence in one query.
Yes. Gyri exposes multihop GraphQL for operators and agents to traverse federated business entities.
Yes. Gyri combines keyword retrieval with graph traversal and cited synthesis.