bugfix: unhandled rejection error

[hacking-racoon]

Background

• Several users have encountered an "unhandled rejection error," causing the application to enter an infinite loading state
• About 9 related issues are currently open in this repository.
• There has been no reply for more than 5 minutes after raising the question. #198 Backend gets 500 from Ollama that works fine #310 Deploy well, can use searXNG search, but cannot use perplexica search #316 Docker version is broken #346 Axios exception broke backend, and also broke thread #368 Backend Error: 404 Resource not Found #361 Discussion search stays "on wait" forever. And edit & run prompt again is not an available option #372 Exposed Ollama on local network, still get connection errors #501 API Calls to Perplexica Hang Without Response When Using Docker Setup  #503

Cause

• Model-related information is stored in local storage (e.g., chatModel, chatModelProvider, embeddingModel, embeddingModelProvider, etc.).
  

• An error occurred (unhandled rejection) because the local storage information was used as-is, even though the inference server had changed. (There is a case where available chat models cannot be checked)
  

Solution

• Logical changes (Included in the PR)
  Replaced local storage with session storage.
  Ensures storage is refreshed each time a new tab is opened.

• Guidance change proposal (just proposal)
  Adding a help message advising users to clear the storage in case of an "unhandled rejection" error. (readme.md)

@ItzCrazyKns . I hope this helps with your amazing project!

[ItzCrazyKns]

There's this piece of code present in the chat window file which checks if the model exists in the list of the available chat models if not it will select another one.

if (
  Object.keys(chatModelProviders).length > 0 &&
  !chatModelProviders[chatModelProvider]
) {
  chatModelProvider = Object.keys(chatModelProviders)[0];
  localStorage.setItem("chatModelProvider", chatModelProvider);
}

[hacking-racoon]

@ItzCrazyKns
During the system setup process, while testing by modifying the config.toml file, if the chatModelProvider stored in localStorage is set to custom_openai for any reason (There are cases where it is unintentionally configured depending on the test scenario)

the block containing localStorage.setItem('chatModel', chatModel); (line 130, ChatWindow.tsx) does not function as expected.

Therefore, I believe we should use sessionStorage.

Please use Chrome developer tools to modify it directly and check the behavior. Thank you.