Dmytro Petryshchuk

Ima

Coming soon...

• Resources on building an empathetic AI NPC
  • Paper: Building an Emphatic AI Coach & Agent (PDF, Github repo)
  • Paper: Informing the design of question-asking conversational agents for reflection (Link)
• Memory system for AI agents: EverMemOS
  • Paper: EverMemOS (Paper on arXiv, Pdf of paper, Github repo)

  • For EverMemOS, they use docker to run 4 databases.

    • MongoDB to store conversation data, user profiles, and memories
    • Elasticsearch for searching text rapidly
    • Milvus for vectorDB
    • Redis for caching

    Realistically, we only need to have one database that stores all of that and lets us search easily

    And for Ima, since we're hosting on Railway, we can just use their Redis for caching.

    I'm looking at PostgreSql and pgvector which is a PostgreSql extension.

• Previous LLM based game: Intra
  • Github repo
  • Blog post