Allow modificatons of completionParams via sendMessage

[markNZed]

Describe the feature

For options like temperature it seems OpenAPI supports changing this per chatCompletion message. For example, on a server we might have one API instance the is sending messages for different clients with different temperature ?

[transitive-bullshit]

Makes sense and happy to accept a PR for this.

completionParams being part of the constructor is purely because of historical reasons as the package has gone through quite a few iterations.

Thanks!

[markNZed]

How would you like the interface to look ? Would we add a public function for updating the params or add options to sendmessage ?

[transitive-bullshit]

Just adding it as an additional option to sendMessage opts and having it override the defaults like

const completionParams = { ...this._completionParams, ...opts.completionParams }

[transitive-bullshit]

@markNZed this is now supported in https://github.com/transitive-bullshit/chatgpt-api/releases/tag/v5.1.1 🔥

[markNZed]

I finally got to testing this and ran into problems. Do we need to be able to dynamically set: maxResponseTokens and maxModelTokens. Is it possible to do this with the OpenAI API ? A minor issues was a change in the artialResponse.delta starting with special characters that no longer went through JSON.stringify(partialResponse.delta) but that is not a big deal.