The Problem
Your ERP knows everything.
Nobody can ask it anything.
ERPs store the operational truth of a business — invoices, inventory, payroll, client history, supplier relationships. But to extract anything useful from them, you need to know the data model, the filter system, the report builder, or have a developer write a query for you.
The people who most need the data — business owners, operations managers, finance teams — are exactly the people who can't get it. ERPWhisperer removes this bottleneck entirely. Ask in plain English. Get the answer.
You say
Show all unpaid invoices from Q3 2024 where the customer is from Lahore and the amount is above 50,000
Generated query
frappe.get_list("Sales Invoice",
filters=[
["status", "=", "Unpaid"],
["posting_date", ">=", "2024-07-01"],
["posting_date", "<=", "2024-09-30"],
["customer_city", "=", "Lahore"],
["grand_total", ">", 50000]
]
)
filters=[
["status", "=", "Unpaid"],
["posting_date", ">=", "2024-07-01"],
["posting_date", "<=", "2024-09-30"],
["customer_city", "=", "Lahore"],
["grand_total", ">", 50000]
]
)
You say
Which supplier has caused the most purchase order delays in the last 6 months?
Generated query
frappe.db.sql("""
SELECT supplier, COUNT(*) as delays
FROM `tabPurchase Order`
WHERE schedule_date < delivery_date
AND creation >= %s
GROUP BY supplier
ORDER BY delays DESC
LIMIT 1
""", [six_months_ago])
SELECT supplier, COUNT(*) as delays
FROM `tabPurchase Order`
WHERE schedule_date < delivery_date
AND creation >= %s
GROUP BY supplier
ORDER BY delays DESC
LIMIT 1
""", [six_months_ago])
Features
What ERPWhisperer does
Natural Language Queries
Ask questions in plain English or Urdu. The system translates to Frappe ORM or raw SQL, executes, and returns formatted results.
Schema-Aware Translation
Trained on the Frappe/ERPNext data model. Knows the difference between a Sales Invoice and a Payment Entry and how they relate.
Automatic Visualisation
Results are automatically presented as the right format — table, bar chart, timeline — based on the nature of the data returned.
Permission-Aware
Respects Frappe's role-based access control. Users only see data they have permission to see — the AI doesn't bypass security.
Saved Queries
Frequently asked questions become one-click reports. "This month's late invoices" becomes a permanent dashboard button for any non-technical user.
Frappe App Plugin
Ships as a native Frappe application. Install it into any ERPNext instance with standard bench commands. No custom infrastructure required.
Stack
How it's built
Frappe
App Framework
LLM API
NL Translation
Python
Core Logic
MariaDB
ERP Database
Redis
Query Cache
Vue.js
Frappe UI
Chart.js
Visualisation
RBAC Layer
Permissions
Currently in development
Schema mapping complete · NL→ORM translation at 81% accuracy · Building Frappe app scaffold