UKnow cloud storage integration allows your agents to search and retrieve information directly from your cloud storage accounts without manual file uploads. Connect multiple cloud storage providers and give agents access to your organization’s documents.
Overview
UKnow provides semantic search capabilities across:
Google Drive - Search across all your Google Drive files and folders
Dropbox - Access Dropbox documents and shared folders
OneDrive - Search personal and business OneDrive accounts
SharePoint - Connect to SharePoint document libraries
UKnow connections are managed via OAuth authentication and require the UKnow service to be configured with valid API credentials.
How It Works
OAuth Connection : Users authenticate with their cloud storage provider via UKnow OAuth
Drive Selection : Choose which drives/libraries to include in search
Path Filtering : Optionally filter to specific folders or paths
Semantic Search : Agents query documents using natural language
Real-time Access : No file uploads needed - agents access live cloud data
Configuration
Basic Configuration
agents :
- name : "document_assistant"
agent_type : "llm_agent"
knowledge_bases :
- name : "company_sharepoint"
knowledge_base_type : "uknow"
enabled : true
config :
username : "[email protected] "
drive_key : "abc123xyz"
path_filter : "/documents/"
drive_ids : []
Configuration Fields
Unique identifier for this knowledge base instance
Enable or disable this knowledge base
UKnow-specific configuration object Email address of the cloud storage account This is the email used for OAuth authentication with the cloud provider.
OAuth drive key/token for authentication Obtained through UKnow OAuth flow. Stored securely and never exposed.
config.path_filter
string
default: "/documents/"
Path filter to limit search scope Examples:
"/documents/" - Only search in documents folder (default)
"/projects/2024/" - Search specific project folder
"/" - Search entire drive
List of specific drive/library IDs to search Leave empty to search all available drives. Use specific IDs to limit scope to certain SharePoint sites or shared drives. Example: ["guid-1234-5678", "guid-abcd-efgh"]
Advanced search configuration search_options.search_type
string
default: "similarity"
Search algorithm type Options:
"similarity" - Standard semantic similarity search
"similarity_score_threshold" - Filter by minimum score
"mmr" - Maximum Marginal Relevance (diverse results)
Number of results to fetch Range: 1-100
search_options.lambda_mult
Lambda multiplier for MMR search Range: 0.0-1.0
Controls diversity vs relevance trade-off in MMR mode
Usage Examples
SharePoint Document Library
agents :
- name : "hr_assistant"
agent_type : "llm_agent"
knowledge_bases :
- name : "hr_policies"
knowledge_base_type : "uknow"
enabled : true
config :
username : "[email protected] "
drive_key : "sharepoint_key_123"
path_filter : "/HR Policies/"
drive_ids : [ "sharepoint-site-guid-1234" ]
system_prompt : |
You are an HR assistant with access to company policies and procedures.
Search the knowledge base to answer employee questions accurately.
Google Drive with Multiple Folders
agents :
- name : "sales_assistant"
agent_type : "react_agent"
tools : [ "web_search" ]
knowledge_bases :
- name : "sales_materials"
knowledge_base_type : "uknow"
enabled : true
config :
username : "[email protected] "
drive_key : "gdrive_key_456"
path_filter : "/Sales Materials/"
search_options :
search_type : "mmr"
fetch_k : 10
lambda_mult : 0.7
system_prompt : |
You are a sales assistant with access to product documentation and sales materials.
Use the knowledge base to find relevant information for customer inquiries.
Multiple Cloud Storage Connections
agents :
- name : "knowledge_agent"
agent_type : "llm_agent"
knowledge_bases :
- name : "sharepoint_docs"
knowledge_base_type : "uknow"
enabled : true
config :
username : "[email protected] "
drive_key : "sp_key_123"
path_filter : "/documents/"
- name : "google_drive_files"
knowledge_base_type : "uknow"
enabled : true
config :
username : "[email protected] "
drive_key : "gd_key_456"
path_filter : "/shared/"
system_prompt : |
You have access to multiple cloud storage sources.
Search across all available knowledge bases to find relevant information.
Managing Connections
UKnow connections are managed through the platform interface or API. Each connection represents an authenticated cloud storage account.
Connection Properties
Email : Cloud storage account email
Drive Key : OAuth authentication token
Display Name : Friendly name for the connection
Drive IDs : Selected drives/libraries to search
Path Filter : Folder path restriction
Sync Status : Connection readiness (syncing, ready)
Connection Lifecycle
OAuth Authentication
User authenticates with cloud provider via UKnow OAuth flow
Drive Discovery
System fetches available drives/libraries from the account
Configuration
User selects drives and sets path filters
Sync Status Check
System verifies connection is ready for search
Agent Integration
Connection is used in agent knowledge base configuration
Search Behavior
How Agents Search UKnow
When an agent needs information:
Query Processing : Agent’s question is sent to UKnow API
Semantic Search : UKnow performs vector similarity search across configured drives
Path Filtering : Results filtered by path_filter if specified
Drive Filtering : Results limited to drive_ids if specified
Ranking : Results ranked by relevance score
Context Injection : Top results provided to agent as context
Search Types
Similarity Search
Score Threshold
MMR (Diverse Results)
Default search mode Returns documents most semantically similar to the query. search_options :
search_type : "similarity"
fetch_k : 5
Best for: General knowledge retrieval Filtered similarity search Only returns results above a minimum relevance score. search_options :
search_type : "similarity_score_threshold"
fetch_k : 10
Best for: High-precision requirements Maximum Marginal Relevance Balances relevance with diversity to avoid redundant results. search_options :
search_type : "mmr"
fetch_k : 10
lambda_mult : 0.7 # Higher = more diverse
Best for: Broad research, avoiding duplicates
Best Practices
Path Filtering : Use specific path filters to improve search relevance and reduce noise. Instead of searching entire drives, target specific folders.
Drive Selection : For SharePoint, select specific document libraries rather than searching all sites to improve performance.
Search Type : Use similarity for most cases. Use mmr when you need diverse results (e.g., research tasks).
Fetch K : Start with 5-10 results. Increase if agents need more context, but be mindful of token limits.
OAuth Tokens : Drive keys are sensitive credentials. Never hardcode them in configurations. Use the platform’s secure connection management.
Sync Status : Ensure connections show sync_status: ready before using in production. Syncing connections may return incomplete results.
Limitations
Read-Only : UKnow knowledge bases are read-only. Agents cannot create, update, or delete cloud files.
OAuth Required : All connections require valid OAuth authentication through UKnow.
API Dependency : Requires UKnow API to be configured with UKNOW_API_BASE and UKNOW_API_KEY environment variables.
Sync Delays : New files may take time to appear in search results depending on cloud provider sync.
Troubleshooting
Connection Issues
Issue Cause Solution sync_status: syncingConnection still indexing Wait for sync to complete, check status via API Missing credentialsAPI keys not configured Verify UKNOW_API_BASE and UKNOW_API_KEY environment variables No results foundPath filter too restrictive Broaden path filter or check folder exists Authentication failedExpired OAuth token Re-authenticate connection via platform
Search Quality Issues
Check path_filter is targeting correct folders
Verify drive_ids include the right libraries
Try adjusting search_type (e.g., use MMR for diversity)
Check connection sync_status is “ready”
Cloud provider may have sync delay
Verify document is in filtered path
Reduce fetch_k value
Use more specific path_filter
Limit drive_ids to relevant libraries only
Next Steps