Skip to content

System Configuration

The Admin UI provides a web interface for managing system-wide settings without editing configuration files.

Chat UI settings

Control the appearance and behavior of the chat interface:

SettingDescription
Chat UI titleTitle shown in the browser tab
LogoCustom logo (light and dark variants)
Placeholder textDefault text shown in empty chat
Follow-up questionsEnable AI-generated follow-up suggestions after each response

Authentication settings

SettingDescription
RegistrationEnable or disable new user signups
Login requiredRequire authentication to use the app
Default roleRole assigned to new accounts

Assistant defaults

Generation settings

SettingDescriptionDefault
ModelDefault LLM model for new assistants
TemperatureRandomness of responses0.7
Max tokensMaximum response length
System promptDefault instructions for assistants
Conversation titleAuto-generate conversation titlesEnabled

RAG settings

SettingDescriptionDefault
Retrieval count (k)Chunks to retrieve5
Similarity thresholdMinimum similarity score0.0
Hybrid searchEnable BM25 + dense fusiontrue
RerankingEnable cross-encoder rerankingtrue
Query transformDefault query transformationmulti_query
Reranker modelDefault rerankerBAAI/bge-reranker-v2-m3

Model & Provider Management

SectionPathDescription
Model providers/admin/model-providersAdd/edit LLM provider connections (API keys, endpoints)
Models/admin/modelsConfigure which models are available for selection

Knowledge Base Management

SectionPathDescription
Knowledge bases/admin/kbsCreate, configure, and manage KBs
Documents/admin/kbs/{id}Upload, reindex, and monitor document ingestion

User Management

SectionPathDescription
All users/admin/usersView and manage user accounts
Ban/unbanRestrict user access
Change rolePromote/demote users

OAuth Providers

SectionPathDescription
OAuth config/admin/oauthConfigure SSO providers (Google, GitHub, etc.)

After adding an OAuth provider, users see the option on the login page.

Agent Settings

Configure default agent behavior for assistants with tool calling enabled:

SettingDescriptionDefault
Max iterationsMaximum tool-calling rounds per response10
TimeoutMaximum agent execution time (seconds)300
MemoryEnable cross-session agent memoryfalse

MCP Configuration

Configure Model Context Protocol servers for extending agent capabilities:

SettingDescription
EnabledToggle MCP integration on/off
Server config pathPath to the MCP servers JSON configuration

Analytics

The admin dashboard shows key metrics:

MetricDescription
ConversationsTotal, today, and trending
MessagesVolume and response times
UsersActive users and registrations
Knowledge basesDocument counts and ingestion status

MIT License