Table of Contents

• Table of Contents
• Summary
• 1. detectPic Image Recognition
• 2. invoiceValidation Invoice Verification
• 3. queryOcrTypes Get supported OCR recognition types
• 4. ocrImages OCR recognition
• 5. Call the chatCompletions API of the large language model
• 6. completionsByPrompt Invokes the large model generation interface via prompt templates
• 7. Retrieve content via semantic search
• Reference object com.fxiaoke.functions.model.InvoiceData
  • Field description
• Changelog
• Background
• Applicable Scenarios
• Prerequisites
• Steps
• Notes

Summary

The overview of this document is as follows: This document introduces the API definitions of Fx.ai, usage methods, and related notes.

1. detectPic Image Recognition

Fx.AI.detectPic(<String modelId>, <String path>)

Request parameters

Request Body

Parameter Name	object	Description	Required
modelId	String	Unique identifier of the object recognition model	Y
path	String	nPath of the image	Y

Code examples

def(boolean error, Map data, String errorMessage) = Fx.AI.detectPic("5dabce96e75d9594e1dc05f6", "N_202006_09_87e2be5f5c364bda9cb5ed61a21ddd25")
if (error) {
  log.error("detectPic " + errorMessage)
  //1. Terminate execution with error
  //Fx.message.throwException("detectPic " + errorMessage)
  //2. Terminate execution using return
  //return;
  //3. Continue execution
} 
//dosomething

Notice

• Enterprises can only use the model if they possess it.

2. invoiceValidation Invoice Verification

Invoice Verification

Fx.AI.invoiceValidation(<InvoiceData arg>)

Request parameters

Request Body

Parameter Name	object	Description	Required
arg	InvoiceData	Invoice verification entity object	Y

Return Parameters

Parameter Name	object	Description
code	integer
message	string
data	object

data	object	Description
N_202308_29_e157da5c641f446f92dbc9e2ba8c8885	object

N_202308_29_e157da5c641f446f92dbc9e2ba8c8885	object	Description
success	boolean
errorCode	string
errorMsg	string
ocrJson	null
ocrType	null

Return an example

{
	"code": 200,
	"message": "OK",
	"data": {
		"N_202308_29_e157da5c641f446f92dbc9e2ba8c8885": {
			"success": false,
			"errorCode": "216630",
			"errorMsg": "recognize error",
			"ocrJson": null,
			"ocrType": null
		}
	}
}

Code examples

InvoiceData invoice = InvoiceData.builder()
     .type("elec_normal_invoice")
     .code("011002100511")
     .number("24830566")
     .date("20210620")
     .amount("339.62")
     .checkCode("957725")
     .build()

def(Boolean error, Object data, String errorMessage) = Fx.AI.invoiceValidation(invoice)
if (error) {
  log.error("detectPic " + errorMessage)
  //1. Use error reporting to terminate execution
  //Fx.message.throwException("detectPic " + errorMessage)
  //2. Use return to terminate execution
  //return;
  //3. Continue execution
} 
//dosomething

Reference Interface

• Reference guide: InvoiceData

3. queryOcrTypes Get supported OCR recognition types

Get supported OCR recognition types

Fx.AI.queryOcrTypes()

Request parameters

Return Parameters

Parameter Name	object	Description
code	integer
message	string
data	array[object]

data	object	Description
type	string
desc	string
descKey	string

Return an example

{
    "code": 200,
    "message": "message",
    "data":
    [
        {
            "type": "type",
            "desc": "desc",
            "descKey": "descKey"
        }
    ]
}

Code examples

def (ue, ud, uem) = Fx.AI.queryOcrTypes()
if(ue){
  log.info(uem)
}
else{
  log.info(ud)
}

4. ocrImages OCR recognition

OCR recognition

Fx.AI.ocrImages(<string ocrType>, <array paths>)

Request parameters

Request Body

Parameter Name	object	Description	Required
ocrType	string		Y
paths	array[string]		Y

Return Parameters

Parameter Name	object	Description
code	integer
message	string
data	object

data	object	Description
N_202308_29_e157da5c641f446f92dbc9e2ba8c8885	object

N_202308_29_e157da5c641f446f92dbc9e2ba8c8885	object	Description
success	boolean
errorCode	string
errorMsg	string
ocrJson	null
ocrType	null

Return an example

{
	"code": 200,
	"message": "OK",
	"data": {
		"N_202308_29_e157da5c641f446f92dbc9e2ba8c8885": {
			"success": false,
			"errorCode": "216630",
			"errorMsg": "recognize error",
			"ocrJson": null,
			"ocrType": null
		}
	}
}

Code examples

def (ue, ud, uem) = Fx.AI.ocrImages("MultipleInvoice", ["N_202311_28_8c3e4c6e652748eb9c52142a32d2bcce"])
if (ue) {
  log.info(uem)
}
else {
  log.info(ud)
}

5. Call the chatCompletions API of the large language model

Calling the AI large model chatCompletions interface

Fx.AI.chatCompletions(<Map map>)

Request parameters

Request Body

Parameter Name	object	Description	Required
map	Map	Map parameter for invoking the chatCompletions interface of large language models	Y

map	Map	Map parameter for invoking the chatCompletions interface of large language models	Required
model	String	Specifies the large language model to be invoked, e.g., ERNIE-Bot, qwen-turbo	Y
messages	List[object]	Historical conversation messages with AI	Y
temperature	BigDecimal	Temperature is a value that controls the confidence of model predictions. Lowering temperature makes predictions more accurate and deterministic but potentially less creative. Increasing temperature yields more diverse completions but may result in less accurate or overly creative outputs. Range: 0-1, default: 0.7	--

messages	object	Description	Required
role	String	Role used in the conversation: system (system), assistant (large language model), user (user)	Y
content	String	Content of the conversation message	Y

Return Parameters

Parameter Name	object	Description
finish_reason	String	The reason why the model stopped generating text
message	String	The text content generated by the large language model

Return an example

{
    "finish_reason": "normal",
    "message": "Certainly. Here's a brief diary example:\n\n---\n\nFriday, April 15, 2023 Sunny\n\nToday my mood was as bright as the sunshine outside. In the morning, I went jogging with the first rays of sunlight, feeling full of energy. In the afternoon, I met with friends at a café, and time slipped away amidst laughter and conversation. In the evening, sitting alone at my desk reading favorite books, my heart was peaceful and content. A simple day, yet filled with happiness and warmth.\n\n---\n\nI hope this diary meets your requirements."
}

Code examples

//model: Currently supported large language models include:
//ERNIE-Bot, ERNIE-Bot4.0 (Baidu Wenxin)
//generalv2, generalv3 (iFLYTEK)
//qwen-turbo, qwen-plus (Alibaba Tongyi Qianwen)
//hunyuan-pro, hunyuan-standard (Tencent Hunyuan)
//ep-20240618123932-5zk5j (Doubao-pro-4k), ep-20240618123950-llzrs (Doubao-pro-32k) (ByteDance Doubao)

//messages: Conversation messages with the model are of List type, containing two fields: role and content
//- role: participant role
//system: context set by the system
//assistant: content returned by the model
//user: content sent by the user
//- content: specific text content

//temperature: A value controlling the model's prediction confidence. Lower temperature yields more accurate and deterministic results but may lack creativity. Higher temperature produces more diverse completions but may result in less accurate or creative outputs. Range: 0-1

Map request = [
  "model": "ERNIE-Bot",
  "messages": [
    [
      "role": "user",
      "content": "Hello ERNIE-Bot, could you generate a diary entry of about 100 words for me?"
    ]
  ],
  "temperature": 0.7
]

def(boolean error, Map data, String errorMessage) = Fx.AI.chatCompletions(request)
if (error) {
  log.error("chatCompletions " + errorMessage)
  //1. Use error to terminate execution
  //Fx.message.throwException("chatCompletions " + errorMessage)
  //2. Use return to terminate execution
  //return;
  //3. Continue execution
}
log.info(data)
//dosomething

Notice

• AI capabilities require purchasing resource packages: [AI Product Resource Pack]

6. completionsByPrompt Invokes the large model generation interface via prompt templates

Invoking Large Model Generation APIs via Prompt Templates

Fx.AI.completionsByPrompt(<String promptApiName>, <Map objectData>, <String bingObjectDataId>, <List otherObjectData>, <Map sceneVariables>)

Request parameters

Request Body

Parameter Name	object	Description	Required
promptApiName	String	Prompt template apiName	Y
objectData	Map	Bound main object data	--
bingObjectDataId	String	Bound main object data ID (mutually exclusive with objectData parameter; when both are provided, objectData takes precedence)	--
otherObjectData	List[object]	Other object data used in the prompt template (objectApinName represents object apiName, dataId represents data ID)	--
sceneVariables	Map	Custom scene variables for prompt template (keys must correspond with those defined in the prompt template)	--

otherObjectData	object	Description	Required
objectApinName	List	Object apiName	Y
dataId	List	Data ID	Y

Return Parameters

Parameter Name	object	Description
message	String	Output result of the large language model

Return an example

{
    "message": "{\n \"outputLanguage\": \"English\",\n \"outText\": \"Hello\"\n}"
}

Code examples

// Prompt template apiName
String promptApiName = 'prompt_GrQPP__c'
// Complete information of the bound main object data
Map objectData = [
  "name": "Customer Data 123",
  "object_describe_api_name": "AccountObj"
]
// Bound main object data ID. Either this parameter or objectData can be used (if both are provided, objectData takes priority)
String bingObjectDataId = "63b6b583b6b41e00018ae21d"
// Other object data used in the prompt template: objectApinName is the object API name, dataId is the data ID
List otherObjectData = [
  [
    "objectApinName": "LeadsObj",
    "dataId": "62a9b7c16d5aad00017bff4f"
  ],
  [
    "objectApinName": "SalesOrderObj",
    "dataId": "62a9b7c16d5aad00017bff4f"
  ]
]
// Custom scenario variables for the prompt template (keys must match those in the prompt template)
Map sceneVariables = [
  "content": "Hello",
  "targetLanguage": "English"
]

def(boolean error, Map data, String errorMessage) = Fx.AI.completionsByPrompt(promptApiName, objectData, bingObjectDataId, otherObjectData, sceneVariables)
if (error) {
  log.error("completionsByPrompt " + errorMessage)
  //1. Use error to terminate execution
  //Fx.message.throwException("completionsByPrompt " + errorMessage)
  //2. Use return to terminate execution
  //return;
  //3. Continue execution
}
log.info(data)
//dosomething

Notice

• AI capabilities require purchasing resource packages: [AI Product Resource Package]

7. Retrieve content via semantic search

Retrieving object data through semantic search

Fx.AI.RAGRetrieval(<String ragApiName>, <String query>)

Request parameters

Request Body

Parameter Name	Knowledge Base	Description	Required
ragApiName	String	Semantic search index API name	Y
query	String	User query	Y

Return Parameters

Parameter Name	object	Description
scoreThreshold	BigDecimal	Reference threshold
hits	List[object]

hits	object	Description
objectApiName	String	Object API name
id	String	Object data ID
content	String	Article content
retrievalType	String	Retrieval type
score	BigDecimal	Score

Return an example

{
    "scoreThreshold": 2.0,
    "hits": [
        {
            "objectApiName": "ServiceKnowledgeObj",
            "id": "6718e6c45bc3550001a216e7",
            "content": "RocketMQ and MySQL Knowledge Points\n\n\nContent of mysql.docx file:\nInterviewer: Please elaborate on RocketMQ's message acknowledgment mechanism? xxx",
            "score": 0.009232215
        },
        {
            "objectApiName": "ServiceKnowledgeObj",
            "id": "66bad2bd0a62fe00018dbd0b",
            "content": "Special Topic: Analysis and Optimization of Mass Data Scheduled Workflows xxx",
            "score": 0.0058653383
        }
    ]
}

Code examples

// ragAPIName is the newly created data in [Semantic Search Index] under [AI Platform Management] in the admin console
String ragAPIName = "rag_service_knowledge__c"

// Content to be retrieved
String query = "Connect to Tianrun"

def(boolean error, Map data, String message) = Fx.AI.RAGRetrieval(ragAPIName, query)
if (error) {
  log.error("chatCompletions " + errorMessage)
  //1. Use error to terminate execution
  //Fx.message.throwException("chatCompletions " + errorMessage)
  //2. Use return to terminate execution
  //return;
  //3. Continue execution
}
log.info(data)
//dosomething

Notice

• This is part of AI capabilities and requires purchasing a resource package: [AI Product Resource Package]

Owner：韩统武(已停用)

Reference object com.fxiaoke.functions.model.InvoiceData

Field description

Parameter Name	object	Description
date	java.lang.String	Invoice date:
number	java.lang.String	Invoice number
amount	java.lang.String	Invoice amount (excluding tax)
code	java.lang.String	Invoice code
type	java.lang.String	Invoice type
checkCode	java.lang.String	Invoice verification code (last 6 digits), optional

Changelog

Version	Date	Changes	Author
v1.0	2026-05-20	Initial version

Background

This document provides detailed information about the Fx.AI API functionality and usage, helping developers integrate relevant capabilities.

Applicable Scenarios

• Business scenario integration
• Data processing and synchronization
• System integration and extension

Prerequisites

• Access to Fxiaoke Open Platform
• Application authorization and configuration completed
• Basic knowledge of relevant business domain

Steps

Please refer to the detailed instructions for each API.

Notes

• Ensure prerequisites are met before calling APIs
• Pay attention to API call frequency limits
• Refer to error code documentation for exception handling