la idea es que se soporten N comando, osea que si el comando no es de promper lo envie en un nuevo decorador a LolaPY, con esto nos aseguramos que podamos soportar comandos tales como /phone set .............

tener en cuenta que es importante que se envie si el usuario envía si esta o no logueado en el asistente

When debugging using SDK debug=True, logtail from chats made with Chatwoot is missing.

Add a setting to automaticaly downgrade the agnet model from gpt4 to 3.5.
This option must be available on SDK.

La idea es :

se crean eventos de terceros que requieren llegar a el asistente, estos eventos serian por ejemplo : finalización del registro, inicio de compra ......

en Lola prompter se expondrá un api que reciba estos eventos, donde le llegue en los parámetros el Lead

ejemplo del decorador para el SDK de python seria:

@lola.on_notification('notification_name')

A method that allows extracting entities from the conversational context of the chat. Initially intended for chat3.5 to keep costs down.
It should process the dialog and prompt looking for entities defined by the wizard developer.

Allow to handle multiple timeouts markers per user session.

Problem: only 1 set_timeout is allowed per session id, for example here notify_afk will overwrite send_promo.

ctx.set_timeout(5, 'send_promo')
ctx.set_timeout(50, 'notify_afk')

Both timeputs marker should coexists.

Google Cluod STT allows to specify 3 languages for Speech to Text.
Main language, and 2 alternative languages.

Lola will expose this settings in the Assistant structure. Language field is by default:
Main lang: en-US
Secondary lang: es-ES

There are two set methods defined for LolaStateManager, the second one overwrites the first one, making the first method unusable.

Steps to Reproduce:

s = ctx.state.get()

# This will return error, it doesn't expect arguments
ctx.state.set(s)

Proposed Solution

Change one of the methods name.

class LolaStateManager:


# rest of code...

    def set(self, state):
        url = f'{self.prompter_url}/api/state/update'
        headers = {'x-lola-auth': self.lola_token, 'Content-Type': 'application/json'}
        data = {'lead': self.lead, 'state': state}
        response = requests.post(url, headers=headers, json=data)
        response.raise_for_status()
        return
    

# This one should be renamed to "reset" for example
    def set(self):
        url = f'{self.prompter_url}/api/state/reset'
        headers = {'x-lola-auth': self.lola_token, 'Content-Type': 'application/json'}
        data = {'lead': self.lead}
        response = requests.post(url, headers=headers, json=data)
        response.raise_for_status()
        return

When using the reset method of LolaStateManager, the client_cmd_enabled value is inadvertently cleared. Consequently, this requires users to re-trigger the /login [TOKEN] in-chat command to reactivate the commands.

Enhancing the developer experience can be achieved by separating the reset of the client_cmd_enabled value from the final state or alternatively making that reset optional.

Add a way to mark URLs so the platform can convert it to an image.

For example for this image:
https://cdn.com/image.png

A developer can mark it with:
https://cdn.com/image.png#image

then if Lola Platform finds out a url pattern ending in #image in the AI response, it will be transform into an Image repsonse.

Provide a way to pause an assistant.
A paused assistant will answer a pre defined message without AI processing.
This message may or may not be configurable througth an attribute in <settings/>

Current settings allows developer to enable or diable audio response on audio incoming messages.
Change this settings to a configurable rule:

If audio input msg -> response: text+audio, only text, only audio