Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.offloadapi.com/llms.txt

Use this file to discover all available pages before exploring further.

resultSchema tells Offload what structure you want in the final result field when a case completes successfully.

​
What The API Validates

The create-case handler only validates that resultSchema is an object. It does not currently validate:
  • whether the object is a complete JSON Schema document
  • whether the schema is internally consistent
  • whether every required field is realistically extractable from the conversation
Use a JSON Schema-shaped object anyway, but treat validation as best-effort prompt guidance rather than strict schema enforcement.

​
Good Example

{
  "type": "object",
  "properties": {
    "signedW9Received": { "type": "boolean" },
    "taxId": { "type": "string" },
    "receivedAt": { "type": "string" }
  },
  "required": ["signedW9Received"]
}

​
Good Schema Design Rules

  • Keep the schema focused on data your application will actually use.
  • Prefer explicit booleans and strings over vague free-form blobs.
  • Use optional properties for data that may not always be available.
  • Put workflow instructions in goal or constraints, not in the schema.

​
What Not To Put In The Schema

  • attachment download URLs
  • raw file bytes
  • instructions like “follow up twice”
  • fields that can only be known from internal business systems
Attachments already flow through the top-level attachments array.

​
Useful Patterns

​
Document Collection

{
  "type": "object",
  "properties": {
    "received": { "type": "boolean" },
    "documentType": { "type": "string" },
    "receivedAt": { "type": "string" }
  },
  "required": ["received"]
}

​
Scheduling Confirmation

{
  "type": "object",
  "properties": {
    "confirmed": { "type": "boolean" },
    "scheduledTime": { "type": "string" },
    "timezone": { "type": "string" }
  },
  "required": ["confirmed"]
}

​
Invoice Follow-Up

{
  "type": "object",
  "properties": {
    "invoiceReceived": { "type": "boolean" },
    "invoiceNumber": { "type": "string" },
    "amount": { "type": "string" }
  },
  "required": ["invoiceReceived"]
}

​
Practical Advice

  • Start with a small schema.
  • Add fields only when you can act on them downstream.
  • Expect result to be absent until the case reaches COMPLETED.
  • Be prepared for schema-shaped output quality to depend on the conversation quality and the clarity of your requested fields.