Knowledge Bases

Knowledge bases are how Synkora gives agents durable, organization-specific context.

What Goes Into a Knowledge Base

Typical sources include:

• PDFs and office documents
• markdown and text files
• uploaded files from the dashboard
• web content
• synced content from external systems

What Synkora Does With That Data

At a high level, the pipeline is:

1. ingest document
2. parse and normalize
3. split into chunks
4. embed and index
5. retrieve at query time
6. rerank and ground the response

Storage Model

Synkora uses PostgreSQL as the system of record and supports vector/search backends such as:

• pgvector
• Qdrant
• Pinecone
• Elasticsearch

Choose the provider that fits your operating model, scale, and latency requirements.

Knowledge Base Design Advice

• keep one knowledge base focused on one domain or team
• avoid mixing unrelated documents into the same retrieval scope
• use smaller, cleaner source sets before tuning prompts
• pair knowledge with tools when the agent needs to both know and act

Related Pages

• Create a RAG Agent
• Document Processing
• Advanced RAG