Documentation

ChatCompletion implements Pipeline

Implements an OpenAI ChatCompletion pipeline using the GPT-3.5-Turbo model as its default.

This pipeline uses an underlying driver (default: GuzzleDriver) to communicate with the OpenAI API. It expects a SUAprompt instance to set the conversation messages, including system instructions, user inputs, and assistant responses. The output methods provide access to the response as text, JSON, or an array, and streaming methods are also supported.

Table of Contents

Interfaces

Pipeline
Defines the required methods for Ai Model Pipelines, which are required for compatibility with LaravelNeuro Agent Networking features.

Properties

$accessToken  : mixed
The OpenAI access token retrieved from configuration.
$driver  : GuzzleDriver
The driver instance used for HTTP communication.
$model  : mixed
The model identifier used for the API request.
$prompt  : mixed
The prompt data, structured as an array of messages.

Methods

__construct()  : mixed
ChatCompletion constructor.
array()  : array<string|int, mixed>
Retrieves the API response as an associative array.
driver()  : Driver
Accesses the injected Driver instance.
driverClass()  : string
Retrieves the class name of the default associated driver.
getDriver()  : Driver
Retrieves the current driver.
getModel()  : mixed
Retrieves the current model identifier.
getPrompt()  : mixed
Retrieves the current prompt.
json()  : string
Retrieves the API response as a JSON-formatted string.
output()  : string
Executes the API request and returns the generated text output.
promptClass()  : string
Retrieves the class name of the default associated prompt.
setModel()  : self
Sets the model for the pipeline.
setPrompt()  : self
Sets the prompt for the pipeline.
stream()  : Generator
Executes a streaming API request and yields output chunks.
streamArray()  : Generator
Executes a streaming API request and yields the output as decoded arrays.
streamJson()  : Generator
Executes a streaming API request and yields JSON-encoded output chunks.
streamText()  : Generator
Executes a streaming API request and yields text output chunks.
text()  : string
Retrieves the text output from the API response.

Properties

$accessToken

The OpenAI access token retrieved from configuration.

protected mixed $accessToken

$model

The model identifier used for the API request.

protected mixed $model

$prompt

The prompt data, structured as an array of messages.

protected mixed $prompt

Methods

__construct()

ChatCompletion constructor.

public __construct([Driver $driver = new GuzzleDriver() ]) : mixed

Retrieves configuration values for the GPT-3.5-Turbo model and API endpoint, initializes the driver (defaulting to GuzzleDriver), sets required HTTP headers, and validates that all required configuration values are present.

Parameters
$driver : Driver = new GuzzleDriver()

An instance implementing the Driver contract.

Tags
throws
InvalidArgumentException

if any required configuration value is missing.

array()

Retrieves the API response as an associative array.

public array() : array<string|int, mixed>
Return values
array<string|int, mixed> —

The decoded API response.

driver()

Accesses the injected Driver instance.

public driver() : Driver
Return values
Driver —

the Driver instance stored in this instance of the class.

driverClass()

Retrieves the class name of the default associated driver.

public driverClass() : string
Return values
string

getModel()

Retrieves the current model identifier.

public getModel() : mixed
Return values
mixed —

The model identifier.

getPrompt()

Retrieves the current prompt.

public getPrompt() : mixed
Return values
mixed —

The prompt data.

json()

Retrieves the API response as a JSON-formatted string.

public json() : string
Return values
string —

The JSON-encoded API response.

output()

Executes the API request and returns the generated text output.

public output() : string

This method is an alias for text().

Return values
string —

The generated text output.

promptClass()

Retrieves the class name of the default associated prompt.

public promptClass() : string
Return values
string

setModel()

Sets the model for the pipeline.

public setModel(mixed $model) : self

Updates both the pipeline and the underlying driver's model.

Parameters
$model : mixed

The model identifier.

Return values
self

setPrompt()

Sets the prompt for the pipeline.

public setPrompt(SUAprompt $prompt) : self

Expects a SUAprompt instance. Iterates through the prompt elements to build a structured messages array where:

  • "role" elements are transformed into system messages.
  • "agent" elements become assistant messages.
  • "user" elements become user messages. The resulting messages array is then passed to the driver using the "messages" key.
Parameters
$prompt : SUAprompt

A SUAprompt instance.

Tags
throws
InvalidArgumentException

if the provided prompt is not a SUAprompt.

Return values
self

stream()

Executes a streaming API request and yields output chunks.

public stream() : Generator

Modifies the request to enable streaming mode and yields the raw stream output.

Return values
Generator —

Yields the streaming output.

streamArray()

Executes a streaming API request and yields the output as decoded arrays.

public streamArray() : Generator
Return values
Generator —

Yields decoded output as an array.

streamJson()

Executes a streaming API request and yields JSON-encoded output chunks.

public streamJson() : Generator
Return values
Generator —

Yields JSON-formatted output chunks.

streamText()

Executes a streaming API request and yields text output chunks.

public streamText() : Generator

Iterates over the streaming output and yields the text content from delta messages.

Return values
Generator —

Yields text output chunks.

text()

Retrieves the text output from the API response.

public text() : string

Extracts and returns the content from the first choice in the response.

Return values
string —

The text content from the API response.


        

        
    




    

    

                                

                

                                
    

        On this page

        
    


                

                            

            

    

        

            
Search results