m4Mindset docs

Mindset v3 / Build & Optimize / AMS SDK

View as Markdown

Knowledge Provisioning Options

Four ways to provide knowledge/context to agents

The Four Methods

Method 1: Tenant Uploads Files

How it works:

  • Tenant uses Context Builder or Context Configuration component
  • Tenant creates contexts manually
  • Tenant uploads files into those contexts
  • You provision these tenant-level contexts in agent sessions

When to use: Tenants have unique, proprietary knowledge they need to provide

Method 2: You Provide App-Level Contexts

How it works:

  • You create app-level contexts via Contexts API
  • You upload and manage files centrally
  • You provision these to tenants through agent sessions
  • Tenants don't need to upload anything

When to use: You have standardized knowledge all tenants should access

Method 3: You Create Tenant-Level Contexts

How it works:

  • Use Contexts API with externalTenantId parameter
  • Create contexts programmatically for specific tenants
  • Optional: Set up synced ingestion (e.g., from Google Drive)
  • Files automatically sync when changed

When to use: You want to create tenant-specific knowledge programmatically or sync from external sources

Method 4: RAG MCP

How it works:

  • Provision a RAG-type MCP instead of contexts
  • No context creation or file uploads needed
  • RAG MCP handles knowledge retrieval
  • Cannot be combined with contexts in same session

When to use: You have an existing RAG system or prefer external knowledge retrieval

Method Comparison

MethodWho ManagesContext LevelFile ManagementEmbed Context Components?
Tenant UploadsTenantTenant-levelManual upload by tenant✅ Yes - Required
App-Level ContextsYou (Platform)App-levelCentral management via API❌ No
Tenant-Level (API)You (Platform)Tenant-levelProgrammatic or synced❌ No
RAG MCPYou (Platform)N/A (no contexts)External RAG system❌ No

Important: Contexts vs RAG MCP

Quick Decision Guide

  1. Do tenants need to provide their own knowledge?

    YES → Use Method 1 (Tenant Uploads Files)

    NO → Continue to next step

  2. Do you have standardized knowledge for all tenants?

    YES → Use Method 2 (App-Level Contexts)

    NO → Continue to next step

  3. Do you need to create tenant-specific knowledge programmatically?

    YES → Use Method 3 (Tenant-Level via API)

    NO → Continue to next step

  4. Do you have an existing RAG system?

    YES → Use Method 4 (RAG MCP)