Clear regular prompt message if using Re-Ask Option

[rjwignar]

This fixes #587

Background

7cc1e2f3ec1aaf77014a56703d7d8e77c706f72d allowed web handlers and slash commands to be executed with the Re-ask option (issue described in #541) by passing the prompt text to the onPrompt call in MessagesView::onResubmitClick().

Before this change, retrying with a regular prompt message would trigger the regular re-try routine (since the prompt text wasn't passed to the onPrompt call in MessagesView::onResubmitClick(): https://github.com/tarasglek/chatcraft.org/blob/9f73deac651f7caa6144e905e326d72a543db002/src/Chat/ChatBase.tsx#L283-L297

After this change however, if re-asking with a regular prompt message, the updated message will be passed to onPrompt (as prompt) and add a duplicate message to the chat (line 278): https://github.com/tarasglek/chatcraft.org/blob/9f73deac651f7caa6144e905e326d72a543db002/src/Chat/ChatBase.tsx#L275-L278

Changes

I added a boolean parameter to ChatBase::onPrompt called retry (false by default):

    async (prompt?: string, imageUrls?: string[], retry: boolean = false) => {
      // ...
      try {
        // If retrying, set the prompt to null to avoid duplicate message
        if (retry) {
          prompt = "";
        }
        // If the prompt text exist, package it up as a human message and add to the chat
        // ...

If retry is True, the prompt will be cleared to prevent message duplication.

retry is False by default because I only want to explicitly set it to True when calling onPrompt() from onResubmitClick():

type MessagesViewProps = {
  // ...
  onPrompt: (prompt?: string, imageUrls?: string[], retry?: boolean) => void;
};
// ... 

function MessagesView({
          onResubmitClick={async (promptText?: string) => {
            await deleteMessages(message.id, "after");
            onPrompt(promptText, undefined, true); // pass prompt text and true for retry, don't include any imageURLs
          }}
          // ...
        })

I had to update the onPrompt signature in MessagesViewProps for syntax reasons. If there's a better way to implement this I'm open to suggestions.

Comparison to Production (updated with Image as Input observations)

Below are the observations I've made on the Re-Ask options before and after 7cc1e2f3ec1aaf77014a56703d7d8e77c706f72d, as well as for this Pull Request:

Task	Observations pre- 7cc1e2f3ec1aaf77014a56703d7d8e77c706f72d (tested using adcfd2d44c63447e6c9c5cfbd42c91e15116f990)	Observations post- 7cc1e2f3ec1aaf77014a56703d7d8e77c706f72d (tested using Production as of e4967a99b4cb7a6478ad5d10c738c3763ffdb126) (Prompt duplication bug observed)	PR Observations
Ask with "help" (to invoke /help)	Executes /help command	Executes /help command	Executes /help command
Re-ask with "help" (to invoke /help)	Response from LLM (undesired)	Executes /help command	Executes /help command
Ask with web handler	Executes web handler	Executes web handler	Executes web handler
Re-ask with web handler	Response from LLM (undesired)	Executes web handler	Executes web handler
Ask with slash command (/import)	Invokes /import	Invokes /import	Invokes /import
Re-ask with slash command (/import)	Response from LLM (undesired)	Invokes /import	Invokes /import
Ask with regular prompt	Response from LLM	Response from LLM	Response from LLM
Re-ask with regular prompt	Response from LLM	Response from LLM, duplicates new message (undesired)	Response from LLM, no message duplication
Ask with Image as Input (no prompt)	Appropriate LLM response	Appropriate LLM response	Appropriate LLM response
Re-ask with Image as Input (no prompt, same image 2)	Appropriate LLM Response	Appropriate LLM Response (no message duplication as no prompt present, expected)	Appropriate LLM (no message duplication as no prompt present, expected)
Ask with prompt and Image as Input	Appropriate LLM Response	Appropriate LLM Response	Appropriate LLM Response
Re-ask with prompt and Image as Input (same image 2)	Appropriate LLM Response (no message duplication, desired result)	Appropriate LLM Response (message duplication with text only, undesired 3)	Appropriate LLM Response (no message duplication, desired result)

My PR preserves the changes from 7cc1e2f3ec1aaf77014a56703d7d8e77c706f72d but prevent the message duplication when re-trying with a regular prompt message.