In the following video titled “0060 Writing the body request of the AI Manager prompt interface,” the presenter delves into the process of engaging with the backend through the AI manager interface. Initially, the importance of acquiring an API key and server base URL to access the backend code is underscored. Subsequently, the presenter elaborates on the creation of a function named “generateTextWithPrivateServer” to facilitate backend interaction. This function necessitates parameters such as a prompt to capture user input, a system message for guiding the interaction, and a completion function to manage the response. To handle asynchronous outcomes, the presenter introduces the concept of employing a completion handler and a result string. Additionally, the process of constructing the API network request using a payload is discussed, wherein text or images are encoded in Base64 format. Notably, the presenter emphasizes the variability in payload formatting based on the development environment and programming language.
Next, the presenter picks up from the previous session where they crafted the request body. Now, their focus shifts to employing this body through the creation of network requests. Initially, they construct a URL for data transmission by combining the server URL with the endpoints “V1/chat/completion” from their LM Studio. Subsequently, they instantiate a request object utilizing the prepared request body and incorporate HTTP headers, thus configuring it as an HTTP POST request with a JSON content type. Following this, they append the request body to the network request’s HTTP body and transform it into JSON serialization data. Acknowledging the potential for errors during this process, the presenter advocates for the inclusion of a catch block to handle errors effectively, ensuring comprehensive error handling. In the subsequent session, they intend to execute the actual network call.
In the following video titled “0062 write the network API request to the private server,” the presenter elaborates on executing a network call to a private server using Swift and URLSession data task. Following the preparation of the request body and URL, the presenter initiates a URLSession data task and configures the request as the task’s HTTP body. Subsequently, the presenter delves into error handling by intercepting the response and scrutinizing for potential errors. In the event of an error, the completion sends an error message, while in its absence, the response data is parsed as a JSON object. Demonstrating the extraction of specific data from the JSON response as strings using JSONSerialization, the presenter stores the retrieved data in the Json Choices array after removing superfluous options and brackets. Additionally, the presenter underscores the necessity of rectifying minor issues like missing brackets and semicolons during debugging, highlighting their significance in programming. The primary objective is to dispatch the task once all curly braces have been closed. Concluding, the presenter mentions the subsequent step, which involves invoking this function within a view controller.