Knowledge Vault & Deploy Anywhere

Ingest documents, databases, and internal wikis into a secure vector knowledge base. Deploy on-premise, hybrid, or air-gapped.

Capabilities

• Semantic search across all connected knowledge sources
• On-premise, hybrid, and air-gapped deployment options
• Automatic knowledge sync with change detection

What Lives in the Vault

The Knowledge Vault unifies:

• Internal wikis (Notion, Confluence, LarkSuite docs, Bitrix24)
• Database tables (PostgreSQL, MySQL, MongoDB, BigQuery, SQLite)
• Project management exports (ClickUp, Notion databases)
• Uploaded documents (PDF, DOCX, MD, TXT)
• Custom data via REST connectors

Semantic Search

Agents query the vault using semantic search (pgvector under the hood) so they retrieve meaning, not exact keyword matches. A question about "how do we handle customer refunds" surfaces the refund policy doc even if it never uses the word "refunds."

flowchart LR
  Q[User question] --> E[Embedding model]
  E --> V[pgvector search]
  V --> C[Top-k chunks]
  C --> A[Agent context]
  A --> R[Grounded answer]

Sync & Change Detection

Connected sources are polled on configurable schedules. When a source changes, only changed records are re-embedded. This keeps re-index costs bounded as your knowledge base grows.

Deployment Modes

Mode	Where data lives	Where models run
Cloud SaaS	AgentBrain-managed region	Customer's BYOK or AB hosted
On-premise	Your datacenter (Kubernetes / bare metal)	Your GPUs or BYOK
Hybrid	Mixed: data on-prem, control plane in cloud	Customer choice
Air-gapped	No internet required, fully sovereign	Open-source models on your GPUs

Tier Availability

Tier	Knowledge Vault
Starter	Not included
Business	Included (KB / KG / Vault)
Enterprise	Included with advanced data connectors
Custom	Included with custom rules and isolated systems

Related

• Deployment Options
• Resources: Semantic Search Guide