Marketplace / Blueprint 04
LIVEKnowledge / knowledge

Knowledge Hub & RAG

Turn private documents and structured data into grounded, searchable AI context.

Review my build

Interactive platform map

Architecture in context

The focused blueprint, its required foundation, and declared recommendations.

Core Selected Automatic Required path
Blueprint 04 / Knowledge

Knowledge Hub & RAG

Turn private documents and structured data into grounded, searchable AI context.

2 vCPU4 GB RAM5 services
01 / Problem

What this replaces

General-purpose models cannot reliably answer from private organizational material, and keyword search misses semantic relationships.

02 / Outcome

What your team gains

A private retrieval layer combining structured PostgreSQL data, pgvector similarity search, full-text search and Open WebUI knowledge workflows.

03 / Capability

What is inside the blueprint

Local or remote embedding generation
Vector storage in PostgreSQL
Exact vector similarity
HNSW approximate indexes
IVFFlat approximate indexes
Metadata filtering
PostgreSQL full-text search
Hybrid lexical + vector retrieval
Multiple vector stores supported by Open WebUI
Cross-encoder reranking
Multiple extraction engines
Agentic retrieval
Full-document context mode
Knowledge access controls
04 / Architecture

How it fits the platform

Documents/data -> extraction/chunking -> EmbeddingGemma/embedding endpoint -> PostgreSQL/pgvector + full-text indexes -> retrieval -> Open WebUI/agents.

Included services

PostgreSQL, pgvector, EmbeddingGemma, llama.cpp, Open WebUI

Platform requirements
05 / Delivery

From prerequisites to operation

Prerequisites
  1. Source documents/data
  2. Chunking/metadata strategy
  3. Embedding endpoint
  4. Storage sizing
Deployment
  1. Create knowledge schema and vector/full-text indexes
  2. Deploy embedding endpoint
  3. Configure ingestion pipeline
  4. Connect Open WebUI knowledge backend
  5. Test retrieval and citations
Configuration
  1. Chunk size/overlap
  2. Embedding model/dimensions
  3. Metadata taxonomy
  4. Top-k and score thresholds
  5. Hybrid weighting/reranking
  6. Document access scope
Operations
  1. Source freshness
  2. Re-indexing
  3. Retrieval quality evaluation
  4. Embedding model migration
  5. Database backups
06 / Combinations

What this unlocks with other layers

Knowledge Hub & RAG + Persistent AI Memory + Knowledge Graph & Institutional Intelligence

Institutional Memory

The AI can retrieve documents, recall prior work and trace relationships and provenance.

Research Intelligence + Knowledge Hub & RAG + Knowledge Graph & Institutional Intelligence

Research-to-Knowledge

Discovery becomes a retained, structured research corpus instead of disappearing after one chat.

07 / Technology

Technology behind this capability

Open WebUIRUNNING - v0.11.0 in censusPostgreSQLRUNNING - three separate DB instances in censuspgvectorRUNNING in enterprise DB imageEmbeddingGemma + llama.cppRUNNING - low-utilization local embedder