how to set systemPrompt in code?

[eisneim]

the documentation an wraper is very hard to understand, lack of examples: https://withcatai.github.io/node-llama-cpp/guide/chat-prompt-wrapper

for example wizardCoder:

Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
{prompt}

### Response:

how to extend this class? the example didn't show how to set systemPrompt

class MyCustomChatPromptWrapper extends ChatPromptWrapper {
    public readonly wrapperName: string = "MyCustomChat";
    
    public override wrapPrompt(prompt: string, {systemPrompt, promptIndex}: {systemPrompt: string, promptIndex: number}) {
        if (promptIndex === 0) {
            return "SYSTEM: " + systemPrompt + "\nUSER: " + prompt + "\nASSISTANT:";
        } else {
            return "USER: " + prompt + "\nASSISTANT:";
        }
    }

    public override getStopStrings(): string[] {
        return ["USER:"];
    }

    public override getDefaultStopString(): string {
        return "USER:";
    }
}

and also how to write wraper for codefusion

<|role_start|>system<|role_end|>{system_message}
<|role_start|>human<|role_end|>{prompt}
<|role_start|>bot<|role_end|>

[giladgd]

You can pass a custom systemPrompt when you create a LlamaChatSession object.

Regarding your wizardCoder example, you can just customize the GeneralChatPromptWrapper with the role names you mentioned:

import {fileURLToPath} from "url";
import path from "path";
import {
    LlamaModel, LlamaContext, LlamaChatSession, GeneralChatPromptWrapper
} from "node-llama-cpp";

const __dirname = path.dirname(fileURLToPath(import.meta.url));

const model = new LlamaModel({
    modelPath: path.join(__dirname, "models", "codellama-13b.Q3_K_M.gguf"),
    promptWrapper: new GeneralChatPromptWrapper({
        instructionName: "Instruction",
        responseName: "Response"
    })
});
const context = new LlamaContext({model});
const session = new LlamaChatSession({
    context,
    systemPrompt: "Below is an instruction that describes a task. Write a response that appropriately completes the request."
});


const q1 = "Hi there, how are you?";
console.log("User: " + q1);

const a1 = await session.prompt(q1);
console.log("AI: " + a1);


const q2 = "Summerize what you said";
console.log("User: " + q2);

const a2 = await session.prompt(q2);
console.log("AI: " + a2);

Based on this example, you can read the source code of GeneralChatPromptWrapper.ts for an example of how it was implemented, so you can create something similar for codefusion.

You're also welcome to open a PR with a CodeFusionChatPromptWrapper so other users can just use it out of the box. You can read the development guide for instructions on how to do that.

[eisneim]

thank you for the answer!