Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

A direct sub-agent chat gives no sign that the main agent is not in it, and the handoff count it already serves is never shown

Open
  • Is something wrong?
#460 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
3/5
Estimated time
1-2 days
Newbie friendliness
72/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
python
Domain
frontend

Research direction

Start in the ui-web instance surface, especially the instance composer, and inspect how the subagents.instances response is consumed. Use pending_handoff_count to show the non-zero count and add the proposed explanation of direct-chat context and handoff timing. Done means both messages appear correctly without backend changes.

Written by the indexing model from the issue text.

Description

Problem

Chatting with a sub-agent instance produces no visible reaction from the main agent, and the page says nothing about why. The behaviour is deliberate and documented in the code -- "Nothing here touches the session transcript. A direct chat exists to keep these exchanges out of the main agent's context" (raven/agent/subagent/manager.py:984) -- but nothing on screen tells the reader that, so it reads as the main agent having missed the message.

There is a handoff: DirectChatHandoff appends an entry as each direct turn lands, and the rendered block is prepended to the user's next message to the main agent (raven/agent/subagent/direct_chat.py). So the main agent does learn, after the fact, on the next thing the user says to it. That is a reasonable design; it is just invisible.

The count is already served and already unused. subagents.instances returns pending_handoff_count (raven/rpc/methods/instances.py:318, modelled at raven/rpc/models.py:2183), and no caller in ui-web/src reads it.

Proposal

Two small things, either independently useful:

  1. Say once, on the instance surface, that this exchange stays out of the main agent's context and reaches it as a summary on the next message to it. One line, where the instance composer is.
  2. Show pending_handoff_count when it is non-zero -- "N exchanges will be reported to the main agent" or similar. The number exists; nothing renders it.

Alternatives considered

  • Mirror direct turns into the session transcript: this is exactly what the design refuses, and for a good reason (context cost, and the main agent being steered by details it was deliberately not given). Not proposed.
  • Show nothing and let people learn the model: what happens today, and it produced this report.

Area

ui-web (the instance surface); no backend change needed for either item.

Additional context

The handoff already handles two edges worth knowing about, which is part of why surfacing it is cheap and honest: an instance created but never spoken to is an entry in its own right ("the case worth reporting is precisely the one where nothing has been said to it yet"), and a turn still answering is reported through live_direct_turns rather than being invisible until it lands -- a fix whose comment records that "a five-minute direct chat was reported to the main agent as an instance with no turns yet (2026-09-08)".

Dominant language
Python
Stars
4.1k
Forks
94
Avg merge
9h 54m
Merged PRs (30d)
370

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from EverMind-AI/Raven

All issues in EverMind-AI/Raven

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.