Hey all — solo dev here.

This is a full memory system that passes every hard test (30-day multi-hop, “ignore everything” traps, vague recall, etc.) at perfect 1.00 faithfulness / 1.00 context recall RAGAS testing with gpt-4.1-mini.

Would love your feedback.

Public LangSmith traces (click → Examples tab): https://smith.langchain.com/public/4b3ee453-a530-49c1-abbf-8...