Skip to main content

​
Input validation

When you define an input schema, each request is validated before executing the handler. If validation fails, the handler does not execute and a 400 is returned: 
{
  "error": "Validation failed",
  "details": [
    {
      "path": ["email"],
      "message": "Invalid email",
      "code": "invalid_string"
    }
  ]
}

​
Example with full schema

index.ts
import { define, z } from "@jelou/functions";

export default define({
  name: "create-ticket",
  description: "Creates a support ticket from WhatsApp",
  input: z.object({
    name: z.string().min(1).describe("Customer name"),
    email: z.string().email().describe("Contact email"),
    subject: z.string().max(200).describe("Ticket subject"),
    priority: z.enum(["low", "medium", "high"]).default("medium"),
  }),
  output: z.object({
    ticketId: z.string(),
    status: z.string(),
  }),
  handler: async (input, ctx) => {
    ctx.log("Creating ticket", { customer: input.name, priority: input.priority });
    return { ticketId: "TKT-2024-0042", status: "open" };
  },
});

​
Supported types

You can use any Zod type inside z.object(): 
input: z.object({
  name: z.string().min(1),
  age: z.number().int().positive(),
  email: z.string().email(),
  active: z.boolean().default(true),
  role: z.enum(["admin", "user", "guest"]),
  tags: z.array(z.string()).optional(),
  metadata: z.record(z.string(), z.unknown()).optional(),
})

​
Coercion in GET requests

For GET requests, query parameters are strings. Use z.coerce to convert types automatically:
index.ts
import { define, z } from "@jelou/functions";

export default define({
  name: "search-orders",
  description: "Search orders by status",
  input: z.object({
    q: z.string(),
    limit: z.coerce.number().default(10),
    page: z.coerce.number().default(1),
    active: z.coerce.boolean().default(true),
  }),
  handler: async (input, ctx) => {
    ctx.log("Searching", { q: input.q, limit: input.limit });
    return { results: [], total: 0 };
  },
});

curl "https://search-orders.fn.jelou.ai/?q=pending&limit=5&page=2"

​
.describe() annotations

Use .describe() on each field to document the parameters. These descriptions appear automatically in the MCP schema, which helps AI agents understand how to use your function: 
input: z.object({
  phone: z.string().min(10).describe("Phone number with country code, e.g.: 593987654321"),
  includeHistory: z.boolean().default(false).describe("Whether to include the conversation history"),
})

​
Output validation

When you define an output schema, the value returned by the handler is validated after execution. If it does not match:
  • A warning is logged
  • The response is sent normally with status 200
Output validation never blocks the response. It is a development tool to detect inconsistencies.
output: z.object({
  name: z.string(),
  balance: z.number(),
})
If the handler returns { name: "Maria", balance: "150" } (balance as string), you will see a warning in the logs but the client receives the response unchanged.

​
Validation error format

Each error in the details array contains:
FieldTypeDescription
pathstring[]Path to the field with the error, e.g.: ["email"] or ["address", "city"]
messagestringHuman-readable error message
codestringZod error code (e.g.: invalid_string, too_small, invalid_enum_value)
{
  "error": "Validation failed",
  "details": [
    {
      "path": ["name"],
      "message": "String must contain at least 1 character(s)",
      "code": "too_small"
    },
    {
      "path": ["priority"],
      "message": "Invalid enum value. Expected 'low' | 'medium' | 'high', received 'urgent'",
      "code": "invalid_enum_value"
    }
  ]
}