Getting Started

AIvis Overview

View as Markdown

AIvis is a privately deployed enterprise AI workspace. It brings chat, enterprise knowledge, model configuration, Agents, Actions, scheduled tasks, and organization governance into one controlled surface while keeping knowledge, indexes, files, credentials, task history, and audit data inside the enterprise network, cloud account, or data-center boundary.

AIvis Private Deployment and Data Privacy MapShows AIvis as a private enterprise deployment that keeps knowledge, files, indexes, credentials, task history, and audit data inside the enterprise-controlled boundary and sends only authorized minimum context to external models or tools.AIvis Private Deployment and Data Privacy MapDesigned so enterprise data stays in your environment while models, tools, and agents receive minimum authorized context1. Enterprise private deployment boundaryAIvis Web / APIdeployed in your environmentidentity and workspace checksEnterprise data planefiles, database, indexesknowledge and history retainedBackground servicesconnectors, indexing, schedulesruns are traceableCredential custodysecrets do not enter promptsPermission filteringuser, group, document scopeAudit factsruntime facts, not secret content2. AI capabilities inside the boundaryChat, search, multimodalunified user entrytext, web, image, voice, codeModel providersadmin-configured accessminimum approved context onlyKnowledgeconnectors, indexes, setsanswers use authorized chunksAgents and ActionsMCP / OpenAPI checked before callsScheduled Tasksrun with creator permissions and audit trail3. Enterprise privacy and governance postureData stays in boundaryexternal services receive minimum context onlyLeast privilegeprompts cannot expand accessFail closedmissing identity or ambiguous config is deniedFor private enterprise deployments: knowledge, connectors, indexes, files, audit records, and scheduled run history stay under enterprise control.

What AIvis is

AIvis is the unified entry point for teams using generative AI. Users can chat with approved models, knowledge, and Agents; administrators configure model access, connect knowledge sources, publish Agents, manage roles and groups, and control which integrations or tools are available.

For enterprise deployments, the goal is not to let models access data without boundaries. AIvis separates models, knowledge, tools, and people into reviewable configuration surfaces. Each capability must be enabled by administrators, then applied through roles, groups, Agent configuration, and runtime authorization.

In private deployments, AIvis emphasizes data residency and minimum disclosure by default. Enterprise knowledge bases, indexes, files, conversations, task history, audit records, and connector configuration are held by the enterprise deployment. When an external model or tool is used, AIvis sends only the minimum authorized context generated for that request, not the full knowledge base, raw secrets, or unauthorized documents.

Quickstart

PathUse whenNext step
Local or private deploymentYou need to validate deployment form, ports, images, and base services.Read Deployment Quickstart.
Production planningYou need to decide whether to enable connectors, indexing, knowledge retrieval, and background jobs.Read Deployment Modes.
Administrator setupYou already have a deployment and need the first administrator, roles, and responsibility boundaries.Read Admin Panel & Roles.
Agent launchYou need to combine models, knowledge, and Actions into reusable assistants.Read Manage Agents.

Major capabilities

CapabilityWhat it does
ChatProvide a unified chat entry point using administrator-enabled models, knowledge, and capabilities.
AgentsCombine instructions, models, knowledge, and Actions into reusable assistants for business workflows.
Internal knowledge searchIndex enterprise content from approved connectors and improve responses with retrieval.
Web SearchAdd public external information when the deployment allows it, without mixing it with internal knowledge governance.
Image, video, and voiceProvide multimodal generation or input/output inside configured model and service boundaries.
ActionsUse MCP Actions or OpenAPI Actions to call approved tools for auditable work.
Code InterpreterProvide controlled execution for calculation, data processing, and analysis tasks when supported by the deployment.
Scheduled TasksRun approved Agents once or on a recurring schedule and review execution history.
Model routing and providersManage model credentials, language models, custom inference providers, default models, and generation capabilities.
Privacy DictionaryDefine organization-specific sensitive entities and decide whether matches are tokenized, generalized, routed locally, blocked, or marked for review before model calls.
Governance and auditManage users, groups, roles, usage, identity, security, and connector data boundaries.

Why AIvis

AIvis is designed for teams that need AI capabilities inside real enterprise workflows:

  • More reliable internal knowledge: use connectors, indexing, document sets, and retrieval settings to ground answers in approved organization knowledge.
  • More controlled private deployment: choose deployment modes that match your security, network, data, and compliance requirements.
  • Finer-grained configuration: govern models, providers, connectors, Agents, Actions, scheduled tasks, roles, and groups independently.
  • Safer Agent execution: Agents can only use saved configuration for models, knowledge, and Actions; they cannot expand access on their own.
  • Clear expansion path: start with Lite for chat and Agent validation, then move to Standard when enterprise knowledge retrieval is required.
  1. Confirm the deployment mode: decide whether connectors, OpenSearch, indexing, and RAG are required.
  2. Establish administrators and roles: define Admin, Curator, and Global Curator responsibility boundaries.
  3. Configure models: add model credentials, enable language models, set the default model, and enable required generation capabilities.
  4. Connect knowledge: choose connectors, limit indexing scope, and validate document sets and retrieval.
  5. Publish Agents and configure scheduled tasks: combine instructions, models, knowledge, and Actions, then test before sharing or scheduling.
  6. Configure the Privacy Dictionary: publish sensitive entity rules for masking, local routing, blocking, and review before model calls.
  7. Establish governance: review users, groups, roles, usage, identity, security, and data-flow boundaries.

Deployment boundaries

Menu visibility and available controls depend on the deployment configuration. Lite mode does not include a vector database; use Standard mode to enable connectors and vector indexing.

Some capabilities are conditional. Craft appears only when the deployment declares it available; Tracing is available only in supported deployment forms; Query History can be disabled or hidden by an administrator; and connectors, indexing, and document sets depend on Standard mode and the relevant background services.

Next steps

If this is your first deployment, start with Deployment Quickstart. For production rollout, follow Deployment Modes, Model Credentials, Connectors & Indexing, Agents, Security, and Tracing.

For production environments, also review AIvis Security to confirm system architecture, Agent runtime, security strategy, data flows, data storage, and access controls.