Intermediate · Agents
Memory and sub-agents Keep the help desk's long conversations within a budget of tokens: old tool results cleared first, notes the model saves that outlast a summary of the turns before, and researchers that search each service's pages in a conversation of their own; a review of an assistant's memory; and notes that last from one conversation to the next.
You build context.py, then memory.py, then researchers.py, then desk_memory.py, then notebook.py
Stages 3 Build, 1 Catch, 1 Ship
Time About 3 hours
Runs on Your machine Loading the workspace…
Loading the workspace…
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . The help desk's 16 long conversations, written for the lab: in each, the caller says when they can take a call, asks 10 of the dataset's opening questions from its test dialogues, of seven words or more, 5 at a time, drawn at random with the passages that hold each one's answer, and asks twice for a callback about one of their tickets, with the advisers' free times.
Data: MultiDoc2Dial , by Song Feng, Siva Sankalp Patel, Hui Wan, Sachindra Joshi, CC BY 3.0 . Each of the 16 conversations held three times by each of three agents: Tools and planning's, which sends the whole conversation; this lab's second stage's, with a budget of 8,192 tokens; and the same with this lab's third stage's researchers. With Qwen3.5-4B at 4 bits and nomic-embed-text-v1.5 at 8 bits, through llama.cpp's server, release b11734, on machines with 4 CPU cores and no GPU, on 22 October 2026, each run with a seed of its own, the same for the three agents, at the sampling settings Qwen3.5-4B's model card gives for its non-thinking mode. Each part of each run, with whether the desk did it, the seconds it took, its requests, their tokens of prompt, those the server had cached and the largest prompt, the tools the model called, what the memory cleared and summarised, the researchers' requests and tokens, and a request the server refused.
Model: Qwen3.5-4B , by Qwen, Apache License 2.0 .
Model: nomic-embed-text-v1.5 , by Nomic AI, Apache License 2.0 .