AI Platform
Enterprise AI Platform
A centralized AI platform that powers every application in WOXOX with governance, security and enterprise scalability.
Not a chatbot. Infrastructure shared by every WOXOX module — apps never call LLMs directly.
AI control plane
- User Request
- Authentication
- Permissions
- Model Router
- Prompt Manager
- Knowledge Retrieval
Modules → AI Platform → Providers
Why Centralized AI
Stop every app from talking to LLMs on its own
Traditional SaaS multiplies prompts, cost, and risk. WOXOX routes every request through one governed AI Platform.
Traditional SaaS
- Every app talks directly to AI
- Duplicated prompts
- No governance
- No monitoring
- High cost
WOXOX
- One AI Platform
- Governed AI
- Reusable prompts
- Shared knowledge
- Lower cost
AI Platform Overview
Infrastructure for every WOXOX application
Copilots, agents, search, and automation are consumers of this plane — not separate AI products. Deep architecture lives on the Platform page.
AI Request Lifecycle
Every request. One governed path.
User Request → Authentication → Permissions → Model Router → Prompt Manager → Knowledge → Context → LLM → Validation → Governance → Response → Audit.
Step 1 of 12
User Request
Copilot, agent, or API invoke enters the platform.
Model Router
The right model for every task
Multi-model support, cost and latency optimisation, fallback, and provider selection — OpenAI, Anthropic, Google, Azure OpenAI, open-source, and future providers.
Multi-model support
Route each task to the right model class.
Cost optimisation
Budgets, caching, and cheaper models when quality allows.
Latency optimisation
Prefer low-latency paths for interactive copilots.
Model fallback
Automatic failover when a provider degrades.
Provider selection
Policy-driven choice per tenant and workload.
- OpenAI
- Anthropic
- Azure OpenAI
- Open-source models
- Future providers
Prompt Manager
Reusable prompts with enterprise change control
Templates, variables, versioning, approval, testing, localization, and prompt analytics.
Reusable prompts
Shared libraries across every WOXOX module.
Templates
Structured templates for copilots, agents, and batch jobs.
Variables
Safe injection of entity fields and context slots.
Versioning
Semver prompts with rollout and rollback.
Approval
Change control before production promotion.
Testing
Eval suites and canary traffic.
Localization
Locale-aware prompt variants.
Prompt analytics
Quality, cost, and latency by prompt version.
Knowledge & RAG
Grounded answers with source attribution
Knowledge base, documents, vectors, embeddings, semantic search, context assembly, and RAG pipelines.
Knowledge Base
Curated enterprise knowledge for grounded answers.
Documents
Files and policies indexed under tenant permissions.
Vector Database
Embeddings store for semantic retrieval.
Embeddings
Consistent embedding pipeline via the Model Router.
Semantic Search
Find meaning across modules and files.
Context Assembly
Ranked snippets assembled for the LLM call.
RAG Pipelines
Configurable retrieval → generate → cite flows.
Source Attribution
Citations back to entities and documents.
Memory Layer
Context that respects tenancy and retention
Conversation, workspace, entity, business context, history, preferences, and retention policies.
Conversation Memory
Short-term dialogue state for copilots.
Workspace Memory
Tenant-scoped preferences and facts.
Entity Memory
Memory attached to deals, tickets, employees, and more.
Business Context
Live OS entities injected under permission checks.
AI History
Prior invocations for continuity and audit.
Personal Preferences
User-level tone and format preferences.
Retention Policies
TTL and legal hold aligned to compliance.
AI Agents
Orchestrated agents — never raw LLM callers
Sales, HR, Finance, Support, Marketing, Operations, Document, and Analytics agents act only through platform APIs and the Workflow Engine.
Sales Agent
Pipeline follow-ups and deal coaching via CRM entities.
HR Agent
Hiring and people workflows through People modules.
Finance Agent
Close helpers and variance explanations on finance data.
Support Agent
Ticket triage and draft replies with RAG.
Marketing Agent
Campaign drafts grounded in approved knowledge.
Operations Agent
Project and inventory exception handling.
Document Agent
Classification, extraction, and contract risk notes.
Analytics Agent
Narrative insights on cross-module metrics.
Agent orchestration uses the shared Workflow Engine and kernel permissions — plugins still cannot bypass the AI Platform.
Governance
The rule every module must follow
Permissions, policies, approval, rate limits, cost controls, audit, compliance, and safety.
Applications and plugins never call LLMs directly.
All requests pass through the AI Platform for authentication, routing, knowledge, governance, and audit.
Permissions
Kernel RBAC/ABAC evaluated before retrieval and tools.
Policies
Content, residency, and allowed-model policies per tenant.
Approval
Human-in-the-loop for sensitive agent actions.
Rate Limits
Per-app, per-user, and per-tenant throttles.
Cost Controls
Budgets, alerts, and hard stops.
Audit Logs
Immutable records of prompts, tools, and outputs.
Compliance
Evidence for SOC 2, GDPR, and HIPAA programs.
Safety
Filters, redaction, and blocked-topic enforcement.
Security
Enterprise AI security primitives
Encryption, prompt security, PII protection, secrets, and compliance — SOC 2, GDPR, HIPAA.
Encryption
TLS in transit; encrypted storage for prompts and artifacts.
Prompt Security
Injection defenses and template allow-lists.
PII Protection
Detection, masking, and residency-aware handling.
Secrets
Provider keys live only in the AI Platform vault — never in apps.
Compliance
Mapped controls for enterprise assurance.
SOC 2
Operational controls for trust reviews.
GDPR
Subject rights and retention for AI history.
HIPAA
Safeguards for healthcare administrative AI use.
Developer APIs
Invoke AI the only supported way
REST, SDK, streaming, webhooks, authentication, rate limits, examples, and playground.
REST API
Invoke, embed, agent run, and prompt APIs.
SDK
Typed clients for JS/TS, Python, and Node.
Streaming API
Token streaming for copilots and UIs.
Webhooks
Agent and job completion events.
Authentication
OAuth 2.0 and scoped platform tokens.
Rate Limits
Documented quotas with burst controls.
Examples
Reference apps for invoke and RAG.
Playground
Safe sandbox to test prompts and models.
import { Woxox } from '@woxox/sdk'
const woxox = new Woxox({ apiKey: process.env.WOXOX_API_KEY })
// Never call OpenAI/Anthropic SDKs from an app or plugin
const result = await woxox.ai.invoke({
capability: 'summarize',
entity: { type: 'ticket', id: 't_42' },
})AI Marketplace
Extend the AI Platform safely
Prompt packs, agents, knowledge packs, industry AI, developer extensions, and verified AI.
Enterprise Use Cases
Every module consumes the same AI plane
CRM, HR, Finance, Operations, Projects, Support, Documents, and Analytics — each workflow invokes the Centralized AI Platform.
AI Performance
Built for scale, cost, and visibility
Caching, streaming, async processing, queues, observability, tracing, monitoring, metrics, and cost analytics.
Caching
Prompt and embedding caches for repeat workloads.
Streaming
Low-latency token streams for interactive UX.
Async Processing
Long jobs off the request path.
Queues
Durable queues for agents and batch RAG.
Observability
Unified traces across router → LLM → tools.
Tracing
End-to-end spans for every AI request.
Monitoring
SLO alerts on latency, errors, and refusals.
Metrics
Quality, safety, and utilization dashboards.
Cost Analytics
Spend by module, prompt, model, and tenant.
Pricing Preview
Included, enterprise, and consumption AI
Compare how AI capability is packaged — always on the same control plane.
Included AI
Core copilots on Growth plans
- · Platform invoke
- · Standard models
- · Basic governance
Enterprise AI
Regulated & global orgs
- · Custom policies
- · Private models
- · Advanced agents
FAQ
AI Platform questions
Is Woxox AI a chatbot?
No. Woxox AI is an Enterprise AI Platform — infrastructure shared by every module. Copilots and agents are consumers of the platform, not the product itself.
Can apps or plugins call OpenAI or other LLMs directly?
No. Applications and plugins must never communicate directly with LLMs. All requests authenticate through the AI Platform for routing, governance, and audit.
Where do provider API keys live?
Only in the AI Platform secrets vault. Modules receive platform tokens — never raw provider credentials.
How does this relate to the Business OS Platform?
The AI Platform is Layer 4 of the Woxox architecture. Identity, permissions, and entities come from the Platform Kernel and Data Layer.
Woxox Business OS
Ready to run enterprise AI as infrastructure?
Explore the AI Platform architecture or book a demo — every WOXOX module already consumes this control plane.