multi-agent-research¶

A 3-agent team — lead triages, researcher digs, writer produces prose. Uses the team topology kind so workers route through the lead.

What this shows¶

reyn agent new to provision named agents with role prompts.
reyn topology new --kind team to constrain who can talk to whom.
The expected delegation chain when the lead receives a request that needs both research and drafting.

Topology¶

       lead (leader)
       /         \
researcher       writer

team kind permits leader↔member edges only. researcher and writer cannot talk directly — they go through lead.

Setup¶

reyn agent new lead       --role "team lead. Triage requests; delegate research to researcher and drafting to writer; synthesize a final answer."
reyn agent new researcher --role "deep technical research. Cite primary sources. No prose polish."
reyn agent new writer     --role "concise prose. Strict word budgets. No headings unless asked."

reyn topology new launch --kind team \
    --leader lead \
    --members lead,researcher,writer

Inspect:

reyn topology show launch

Expected (4 permitted edges, all leader↔member):

permitted edges (4):
  lead → researcher
  lead → writer
  researcher → lead
  writer → lead

Run¶

reyn chat lead
> Investigate DuckDB v1's breaking changes and produce a 200-word summary.

Expected delegation chain¶

lead receives the request, classifies it as research+writing.
lead emits messages_to_agents → researcher (gather facts).
researcher returns findings to lead.
lead emits messages_to_agents → writer (draft to 200 words, here are facts).
writer returns prose to lead.
lead synthesizes and replies to the user.

User sees an interim "(working on it)" then the final 200-word summary.

Files in this example¶

setup.sh — the three commands above as a runnable script.
topology.expected.txt — sample reyn topology show launch output.
transcript.example.txt — sketch of the chat session.