feat(chat): add onFinish hook to RAGChat for final response handling

[ronal2do]

• Implemented an onFinish callback in the RAGChat class to allow users to access and modify the final output of the chat.
• Updated the ChatOptions type to include the new onFinish property.
• Added a new test suite to validate the functionality of the onFinish hook, ensuring it is called with the correct output after a chat interaction.

closes #96

[CahidArda]

Hi @ronal2do,

In the current implementation, onComplete is async, therefore onFinish doesn't really modify the output.

Can you describe your use case a bit more:

• We return the response in ragChat.chat, what would be the use case for a hook?
• Do you need to modify the output in your use case?

[ronal2do]

@CahidArda indeed is async, I have it in a custom implementation.

application case, user wants to save on the database the response or add annotations

this is a piece of a higher level abstraction

const result = await chat(messages, {
    onFinish: async ({ response }) => {
      try {
        await saveChat({
          id,
          messages: [...coreMessages, ...response.messages],
        });
      } catch (error) {
        console.error('Failed to save chat');
      }
    },
  });

another example, but for a custom abstraction with the annotations

   const response = createDataStreamResponse({
      status: 200,
      statusText: 'OK',
      headers: {
        'Custom-Header': 'value',
      },
      async execute(dataStream) {
        // Write annotation
        dataStream.writeMessageAnnotation({ type: 'status', value: 'processing' });
        // Merge another stream
        const result = chat.chat(lastMessage, {
          experimental_activeTools: activeTools,
          onFinish: async ({ text, steps }) => {
            steps.forEach(async (step) => {
              if (step.finishReason === "tool-calls") {
                dataStream.writeMessageAnnotation({ type: 'status', value: step.toolCalls[0].toolName });
              }
            });
          },
        });
        result.mergeIntoDataStream(dataStream);
      },
      onError: error => `Custom error: ${error.message}`,
    });