Opening the thread

PROTRAILBLAZER

LLM Memory · Context & Recall
CTX TOKENS0
LIMIT1,000
TURNS IN0
DROPPED0
SUMMARY TOK0
MEM INJECTED0
RETRIEVED0
conceptual · token counts illustrative
Conversation the full thread you see
Active context the only thing the model receives
0 / 1,000 tokens 

      

    

  


  

    

      

        Saved memory
        selected facts, not transcripts
        
      

      

    

    

      

        Prior chats
        retrieval store, searched per request
        
      

      

    

  









  

    

      
System prompt

      
Recent turns

      
Summary of old turns

      
Saved memory

      
Retrieved prior chat

    

    
Conceptual architecture

    
    
  


  

    
    
    
  


  

    
Conversation

    

      
      
      
      
    

    
Context limit