feat: Export getLlamaForOptions

[Thomas101]

Add getLlamaForOptions to the list of exported functions

Description of change

This PR exports getLllamaForOptions from the main import. We've been using this in Aibrow to hard check for GPU engine support at runtime. For example...

import { getLlamaForOptions } from 'node-llama-cpp'

try {
  await getLlamaForOptions({
    gpu: false, // 'cuda' 'vulkan' 'metal' ...
    build: 'never',
    vramPadding: 0
  },
  { skipLlamaInit: true })
  console.log('Supported')
} catch (ex) {
  console.log('Not supported')
}

Pull-Request Checklist

• Code is up-to-date with the master branch
• npm run format to apply eslint formatting
• npm run test passes with this change
• This pull request links relevant issues as Fixes #0000 N/A
• There are new or updated unit tests validating the change N/A
• Documentation has been updated to reflect this change N/A
• The new commits and pull request title follow conventions explained in pull request guidelines (PRs that do not follow this convention will not be merged)

[giladgd]

@Thomas101 getLlamaForOptions is an internal function and is not intended for external use.
It exists purely for additional internal parameters used for internal operations, that when used outside of the intended lifecycle can malfunction and cause side effects.

To check which GPU types are supported by the current machine at runtime you can use the getLlamaGpuTypes method.
To check whether loading the available binaries for a given configuration works you can use the dryRun option of the getLlama method.

Here's how you can achieve the same result in a more stable manner using the public APIs:

import {getLlamaGpuTypes, getLlama, LlamaLogLevel} from "node-llama-cpp";

async function getWorkingGpuTypes() {
    const gpuTypes = await getLlamaGpuTypes("supported");
    const testResults = await Promise.all(
        gpuTypes.map(async (gpuType) => {
            try {
                await getLlama({
                    gpu: gpuType,
                    build: "never",
                    logLevel: LlamaLogLevel.disabled,

                    // this will only load the binary and avoid loading the llama.cpp backend.
                    // to also validate loading the llama.cpp backend,
                    // omit this option and then call `.dispose()` on the returned `Llama` instance afterwards
                    dryRun: true
                });
                return gpuType;
            } catch (err) {
                return null;
            }
        })
    );

    return testResults.filter((gpuType) => gpuType != null);
}

Let me know if this solves your issue, or you still need something not possible with the above solution.
I'll close this PR for now since getLlamaForOptions is intended to be internal.

[Thomas101]

Thanks, that's a much better way to do it. I don't think the getLlamaGpuTypes call was available when we originally patched this in at the beginning of the year :)