AI memory

  • September 20, 2017 at 19:25 #22017
    Moodie Younis
    Participant

    Hi,

    Im really enjoying Utility and path so far. But the further i go into designing my AI the more i want to ensure its capable of achieving its intended result for my product.

    Im constructing a memory for my AI and eager for some thoughts on best approach. Currently im storing aspects of my AI’s memory in a set of variables in a script attached on the AI itself. Im wondering if this is the best approach… if i wanted to SAVE this memory and reinstantiate that AI later in time.

    My AI evolves over long periods of time (months)… and this is not a continuous process, there will be sessions where someone loads up an app to interact with it, socialize it, feed it ect… This probably touches on alot of concepts but would really love some feedback / thoughts on this.

    September 22, 2017 at 10:57 #22028
    Rami
    Keymaster

    Hi Moodie,

    Thanks for your post.

    You are of course free to store the “memory” in any way that you see fit.

    If I were to come up with some recommendations, it would be to utilize the standard IAIContext-implementing object (the Context), since that is what is already provided to each AI entity.

    In regards to ‘saving’, one option is to utilize Apex AI’s serialization, which basically requires you to add the attribute [ApexSerialization] to the fields you want to store in JSON. Of course the fields that you apply to need to be serializable (primitative data types or [ApexSerialization] classes). Additionally, it is out of scope for this support to go into details in this regard, but basically the Apex SerializationMaster can save and load objects that are Apex serializable, and JSON files that have been formatted correctly.

You must be logged in to reply to this topic.

Topic Tags