Skip to content
Permalink

Comparing changes

Choose two branches to see what’s changed or to start a new pull request. If you need to, you can also or learn more about diff comparisons.

Open a pull request

Create a new pull request by comparing changes across two branches. If you need to, you can also . Learn more about diff comparisons here.
base repository: topcheer/ggcode
Failed to load repositories. Confirm that selected base ref is valid, then try again.
Loading
base: v1.3.61
Choose a base ref
...
head repository: topcheer/ggcode
Failed to load repositories. Confirm that selected head ref is valid, then try again.
Loading
compare: v1.3.62
Choose a head ref
  • 3 commits
  • 14 files changed
  • 2 contributors

Commits on Jun 7, 2026

  1. fix: batch sub-agent stream text/reasoning to prevent TUI message flood 

    Sub-agent NotifyStreamText/NotifyReasoning were called on every LLM
streaming token (~50-100/s per agent) with no batching, sending each
token as a separate Bubble Tea message. With 2-3 concurrent sub-agents
this produced 300-600 messages/second that overwhelmed the serial
Update() loop, causing severe TUI stuttering.

Changes:
- Add per-agent text/reasoning buffers to subagent.Manager with an 80ms
  ticker that flushes accumulated chunks in batches (~12.5 Hz vs ~100 Hz)
- Add StartStreamBatcher() called after TUI callbacks are wired
- Flush remaining buffers on Shutdown()
- Remove safego.Go wrapper from sendProgramMsgs — program.Send is already
  thread-safe (buffered channel), the goroutine only added scheduling
  overhead and message reordering risk
- Update tests to use FlushStreamBatch() and non-blocking mock sends

Result: ~85% reduction in message volume with 3 concurrent sub-agents
(600+/s → ~80/s), eliminating TUI freeze during sub-agent work.

Co-Authored-By: ggcode <noreply@ggcode.dev>
    Junjun Zhang and ggcode committed Jun 7, 2026
    Configuration menu
    Copy the full SHA
    a8794c9 View commit details
    Browse the repository at this point in the history

  2. fix: batch swarm teammate text/reasoning to prevent TUI message flood 

    Same issue as sub-agents: teammate_text/teammate_reasoning events were
emitted on every LLM streaming token (~50-100/s per teammate) through
the Manager.emit() → SetOnUpdate callback → sendProgramMsgs path. With
4+ teammates working concurrently this produced 400-800 messages/second.

Changes:
- Add per-teammate text/reasoning buffers to swarm.Manager with an 80ms
  ticker (same interval as sub-agent batching)
- teammate_text and teammate_reasoning events are buffered in emit();
  all other event types (tool_call, idle, shutdown, etc.) pass through
  immediately
- StartStreamBatcher() called after TUI SetOnUpdate wiring
- Flush remaining buffers on Shutdown()
- Export FlushStreamBatch() for test use

Combined with the sub-agent batching fix, total message volume with
3 sub-agents + 4 teammates drops from ~1400/s to ~90/s.

Co-Authored-By: ggcode <noreply@ggcode.dev>
    Junjun Zhang and ggcode committed Jun 7, 2026
    Configuration menu
    Copy the full SHA
    d910c2d View commit details
    Browse the repository at this point in the history

  3. release: v1.3.62 

    Co-Authored-By: ggcode <noreply@ggcode.dev>
    Junjun Zhang and ggcode committed Jun 7, 2026
    Configuration menu
    Copy the full SHA
    c4e24d4 View commit details
    Browse the repository at this point in the history
Loading