Customer Support Schema

The Customer Support Schema is for support chatbots, agent-assist copilots, CRM AI, and helpdesk platforms. It keeps the normal MemoryOS SDK path:

add() -> store memory from support conversations
get() -> retrieve customer context before answering

When a tenant enables domain_schema: "support", MemoryOS still runs the General Engine, then adds a structured support overlay in support_memories.

What it remembers

Field	Meaning
`support_type`	The active support vertical for this customer or conversation
`customer_identity`	Account hints, tier, customer name, verified status, or account reference
`communication_preference`	Channel, tone, detail level, and technical comfort
`language_profile`	Preferred language and regional style
`current_open_issue`	The issue future support replies should start from
`issue_history`	Useful prior support issues without dumping every ticket
`resolution_preference`	Refund, replacement, callback, workaround, escalation, or self-serve preference
`sentiment_pattern`	Frustration, repeat complaint, high escalation risk, or easy-to-resolve pattern
`risk_signals`	Fraud, churn, legal, compliance, VIP, safety, or escalation signals
`support_context`	Vertical-specific JSONB context such as orders, invoices, bookings, plans, or transactions

Support types

Supported support types:

saas
ecommerce
banking_fintech
travel
telecom
edtech_support
general_info

Enable from dashboard

Most teams should enable the Support Schema from the workspace dashboard first.

Open Settings in the MemoryOS workspace dashboard.
Find Domain Schema.
Select Customer Support Schema.
Choose a support routing mode:
- single for one vertical such as SaaS, ecommerce, banking, travel, telecom, or EdTech support
- multi for helpdesk platforms such as Intercom, Crisp, or Zendesk
- auto for early testing
Save changes.

Enable Customer Support Schema from the MemoryOS tenant settings page

The API examples below do the same thing programmatically.

Routing modes

Support products are not all the same. A bank usually has one vertical. A helpdesk platform like Intercom, Crisp, or Zendesk may serve many verticals. MemoryOS supports three routing modes:

Mode	Best for	Behavior
`single`	SBI, Flipkart, MakeMyTrip, Airtel, Udemy-style tenants	Always use one fixed support type
`multi`	Intercom, Crisp, Zendesk-style helpdesk platforms	Classify each conversation within an allowed list
`auto`	Early testing or broad demos	Classify each conversation across all support types

Single vertical

Use this when your whole product belongs to one support vertical.

{
  "support_type_mode": "single",
  "support_type": "banking_fintech",
  "support_types_allowed": []
}

Multi-vertical helpdesk

Use this for support platforms that serve many customer businesses.

{
  "support_type_mode": "multi",
  "support_type": null,
  "support_types_allowed": [
    "saas",
    "ecommerce",
    "banking_fintech",
    "travel",
    "telecom",
    "general_info"
  ]
}

For example:

Customer A asks about a delayed order -> ecommerce
Customer B asks about a failed invoice -> saas
Customer C asks about a failed UPI payment -> banking_fintech

Classification happens inside the support extraction pass. It does not require a separate detection call.

Auto detect

Use this for experimentation when you do not want an allow-list.

{
  "support_type_mode": "auto",
  "support_type": null,
  "support_types_allowed": []
}

Enable with API

If you use the dashboard, no API key is needed because you are already logged in. If you call the API directly, use your tenant API key:

Authorization: ApiKey mem_your_api_key

Enable Support

PATCH /v1/tenant/domain-schema
Authorization: ApiKey mem_...
Content-Type: application/json

{
  "domain_schema": "support"
}

Executable curl example:

curl -X PATCH "https://api.memoryos.io/v1/tenant/domain-schema" \
  -H "Authorization: ApiKey mem_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{"domain_schema":"support"}'

Configure Routing

Single vertical:

curl -X PATCH "https://api.memoryos.io/v1/tenant/support-type" \
  -H "Authorization: ApiKey mem_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "support_type_mode": "single",
    "support_type": "saas",
    "support_types_allowed": []
  }'

Multi-vertical helpdesk:

PATCH /v1/tenant/support-type
Authorization: ApiKey mem_...
Content-Type: application/json

{
  "support_type_mode": "multi",
  "support_type": null,
  "support_types_allowed": ["saas", "ecommerce", "banking_fintech", "travel", "telecom", "general_info"]
}

Executable curl example:

curl -X PATCH "https://api.memoryos.io/v1/tenant/support-type" \
  -H "Authorization: ApiKey mem_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "support_type_mode": "multi",
    "support_type": null,
    "support_types_allowed": ["saas", "ecommerce", "banking_fintech", "travel", "telecom", "general_info"]
  }'

Check Current Config

GET /v1/tenant/domain-schema
Authorization: ApiKey mem_...

curl "https://api.memoryos.io/v1/tenant/domain-schema" \
  -H "Authorization: ApiKey mem_your_api_key"

The response includes:

{
  "domain_schema": "support",
  "support_type_configured": null,
  "support_type_mode": "multi",
  "support_types_allowed": ["saas", "ecommerce", "banking_fintech"]
}

Retrieval safety rules

Support memory is not a live system of record. It is remembered customer context. When Support retrieval returns system_prompt_addition, MemoryOS prepends guardrails:

Support memory usage rules:
- Treat this as remembered customer context, not live system truth.
- Do not claim you checked, changed, refunded, cancelled, verified, located, or triggered anything unless your own tool/API result confirms it.
- If live system data is needed, say you need to check the relevant support system or call the appropriate tool.

For banking_fintech, MemoryOS also adds:

- Never ask for or repeat OTPs, passwords, full account numbers, full card numbers, Aadhaar, or PAN values.

These rules are deliberately small and deterministic. They add no extra LLM call and only a few prompt tokens.

MemoryOS versus your support tools

Production support agents should combine MemoryOS with your operational tools.

Responsibility	Should live in
Customer remembers they prefer refunds	MemoryOS
Customer has an unresolved invoice issue	MemoryOS
Customer has contacted support three times	MemoryOS
Current order status	Your order system
Current invoice status	Your billing system
Account owner verification	Your CRM/auth system
Refund creation	Your payment/support tool
Ticket SLA timer	Your helpdesk system

The recommended production pattern:

Call MemoryOS get() for remembered support context.
Call your tools for live truth.
Give both to your model.
Store the conversation back with MemoryOS add().

PII and regulated data

The Support Schema includes extraction-time redaction for high-risk identifiers, especially for banking/fintech. Never send unnecessary sensitive data to MemoryOS. Prefer:

stable internal customer IDs in external_user_id
ticket IDs in metadata
account references or masked identifiers
concise support summaries instead of full operational records

Avoid sending:

raw payment card data
OTPs
passwords or access tokens
Aadhaar or PAN numbers
full bank account numbers
full passport numbers
full IMEI numbers

PNR values are allowed for travel support because they are operational booking references, not financial credentials.

Workspace dashboard endpoints

Use tenant endpoints for dashboards and support analytics.

Need	Endpoint
Customer support memory list	`GET /v1/tenant/customers`
Aggregate support stats	`GET /v1/tenant/support-stats`
Configure support routing	`PATCH /v1/tenant/support-type`
Current domain config	`GET /v1/tenant/domain-schema`

Normal model responses should use SDK get(), not dashboard endpoints.

​What it remembers

​Support types

​Enable from dashboard

​Routing modes

​Single vertical

​Multi-vertical helpdesk

​Auto detect

​Enable with API

​Enable Support

​Configure Routing

​Check Current Config

​Retrieval safety rules

​MemoryOS versus your support tools

​PII and regulated data

​Workspace dashboard endpoints

​Related pages