Fix Memory Recall Failures in OpenHuman Integration #31

Goal
Ensure memory recall works correctly in OpenHuman using the memory server.

Description
Current state:

• Ingestion is working
• Recall is NOT working reliably
• Debug screen shows memory layer not returning correct data

This blocks:

core product functionality
user experience
validation of Neocortex + memory system

Focus:

debugging recall pipeline
validating retrieval correctness
ensuring integration works end-to-end

Acceptance Criteria

• Memory recall returns correct results from stored data
  Debug screen confirms:
• memory retrieval working
  Works across:
• local
• staging
• No empty/incorrect responses when data exists
• Logs confirm successful retrieval
• Tested with:
• multiple ingestion cases
• multi-query scenarios