Skip to main content
Knowledge bases provide agents with access to domain-specific information, documents, and organizational knowledge. AMAS supports two types of knowledge bases, each optimized for different use cases.

​
Available Knowledge Base Types

Milvus

Vector Database ✅Production-ready vector database with semantic search capabilities. Upload documents and perform semantic search over embeddings.Best for: Custom document collections, uploaded files, fine-grained control

UKnow

Cloud Storage ✅Connect Google Drive, Dropbox, OneDrive, and SharePoint via OAuth. Search cloud documents without manual uploads.Best for: Existing cloud storage, live document access, no file management

​
Comparison

FeatureMilvusUKnow
Data SourceUploaded filesCloud storage (OAuth)
SetupUpload documents to collectionsConnect cloud account
Search TypeSemantic vector searchSemantic search via UKnow API
File ManagementManual upload/deleteAutomatic sync from cloud
CustomizationFull control (chunking, embeddings)Managed by UKnow
Best ForCustom knowledge basesExisting cloud documents

​
When to Use Each Type

​
Use Milvus When:

You need full control over document processing and chunking
Building custom knowledge bases from uploaded files
Want to fine-tune embedding models and retrieval parameters
Need to manage document lifecycle (add, update, delete)

​
Use UKnow When:

Documents already exist in Google Drive, SharePoint, etc.
Want automatic sync with cloud storage
Need to search across multiple cloud providers
Prefer OAuth-based access over file uploads

​
Configuration Basics

​
Milvus Configuration

knowledge_bases:
  - name: "company_docs"
    knowledge_base_type: "milvus"
    id: "kb_abc123"
    config:
      top_k: 10
      metric_type: "COSINE"

​
UKnow Configuration

knowledge_bases:
  - name: "sharepoint_docs"
    knowledge_base_type: "uknow"
    enabled: true
    config:
      username: "[email protected]"
      drive_key: "oauth_key_123"
      path_filter: "/documents/"

​
Using Multiple Knowledge Bases

Agents can use multiple knowledge bases simultaneously: 
agents:
  - name: "comprehensive_assistant"
    agent_type: "llm_agent"
    
    knowledge_bases:
      # Vector database for uploaded docs
      - name: "internal_kb"
        knowledge_base_type: "milvus"
        id: "kb_001"
        config:
          top_k: 5
      
      # Cloud storage for live docs
      - name: "sharepoint"
        knowledge_base_type: "uknow"
        enabled: true
        config:
          username: "[email protected]"
          drive_key: "key_123"
          path_filter: "/"
    
    system_prompt: |
      You have access to both internal documentation and live SharePoint files.
      Search both knowledge bases to provide comprehensive answers.

​
How Knowledge Bases Work

1

Agent Receives Query

User asks a question that requires domain knowledge
2

Semantic Search

Agent searches configured knowledge bases using semantic similarity
3

Retrieve Relevant Documents

Top-k most relevant documents/chunks are retrieved
4

Context Injection

Retrieved content is added to agent’s context
5

Generate Response

Agent generates response using both its knowledge and retrieved context

​
Best Practices

Start Small: Begin with one knowledge base type and expand as needed
Combine Types: Use Milvus for stable internal docs and UKnow for dynamic cloud content
Tune Retrieval: Adjust top_k and search parameters based on your use case
Token Limits: More knowledge bases = more retrieved content. Monitor token usage and adjust retrieval settings accordingly.

​
Next Steps

Milvus Setup

Configure vector database knowledge bases

UKnow Setup

Connect cloud storage accounts

Agent Configuration

Learn about complete agent configuration

Agent Capabilities

Explore all agent capabilities