Build ChatGPT chatbots, surprisingly fast 🚀 Use our Discord server https://discord.gg/botpress to interact!

Botpress

Hi

I have found a solution that works for me, maybe it can be useful for anyone else:

First I create two arrays of strings:
- workflow.arrayPreguntasUser
- workflow.arrayRespuestasChatbot

Then, in a execute code before sending the request to the LLM:

const productInfo = []
workflow.arrayPreguntasUser.push(workflow.pregDirectaCohere)
workflow.arrayRespuestasChatbot.push(workflow.respuestaCohere)

for (let i = 0; i < workflow.arrayPreguntasUser.length; i++) {
    const chatCohereUser =
    {
        role: "User",
        message: workflow.arrayPreguntasUser[i]
    }

    const chatCohereChatbot =
    {
        role: "Chatbot",
        message: workflow.arrayRespuestasChatbot[i]
    }
    productInfo.push(chatCohereUser,chatCohereChatbot)

}

This way I always have all the user's questions and all the chatbot's answers to pass to the LLM in order to have complete context of the conversations so I can get more accurate answers.

Hope this can help.

Thanks!!!