Concurrency Queue

Caps concurrent agent operations. Additional requests queue with priority-aware ordering + starvation prevention.

Flow

graph TB A[Agent request] --> B{Slot available?<br/>running < maxConcurrent} B -->|yes| C[Acquire slot<br/>Status: RUNNING] B -->|no| D[Queue with priority<br/>Status: WAITING] D --> E[Sorted: USER > SCHEDULED > BACKGROUND] E --> F{Slot freed?} F -->|yes| G[Drain highest priority] F -->|no| H[Wait with timeout + cancellation] G --> C C --> I[Op completes → release → drain next]

Priority levels

Level	Value	Use case
USER	2	Direct user actions (chat, commands)
SCHEDULED	1	Automated tasks (standup, health check)
BACKGROUND	0	Background indexing, embedding generation

Higher priority dequeued first. Within same priority: FIFO.

Starvation prevention

Background timer runs every 15 s. Any item waiting > 60 s gets bumped up one priority level. BACKGROUND eventually promotes to SCHEDULED, SCHEDULED to USER — no indefinite starvation.

Queue depth

maxQueueDepth = maxConcurrent × 10

At maxConcurrentStreams = 3, queue holds up to 30 waiting items. Beyond → immediate reject.

Cancellation

• AbortSignal — pass an AbortSignal for cooperative cancellation.
• Timeout — set timeoutMs deadline; item rejected if it hasn’t acquired a slot by then.

Commands

Command	Effect
Show Agent Queue Status	Running + waiting items, per-item cancel
Cancel All Queued Agent Requests	Cancels all waiting items; running ops NOT interrupted

Status bar

Emits onDidChange with { running, waiting, maxConcurrent }. Displays:

• Agent: 2/3 — 2 slots in use out of 3
• Agent: 3/3 (2 queued) — full, 2 waiting

Settings

Setting	Default	Range	Purpose
codebuddy.agent.maxConcurrentStreams	3	1–10	Max concurrent agent streams

Live-reloads on change. Raising the limit drains queued items into newly available slots.