Memory leak during long-running session with multiple bash tool calls

[lildengzi]

Description

envirment:
tools value

Node v26.1.0
Shell fish, inside kitty terminal
RAM 14GB + 14GB swap
CPU AMD Ryzen 7 6800H
description:
During a ~2 hour interactive session involving many consecutive bash tool invocations (setting up a macOS KVM), the opencode process RSS grew monotonically from ~300MB to a peak of 8.6GB (reported by systemd for the parent kitty scope, opencode as dominant child).
The bash subprocess group alone peaked at 468MB, and open_code.db grew to ~70MB. No OOM kill occurred only due to 14GB swap + CachyOS aggressive reclaim. Meanwhile a QEMU VM consuming 4.9GB RSS was also running on the same host, making total pressure ~13.5GB.
No explicit errors in ~/.local/share/opencode/log/opencode.log (1.8MB, 10093 lines).

influence:
Near-OOM on a 14GB host with swap thrashing. Lighter hosts would trigger OOM kill.
guess:
Buffered bash tool output or tool-call history not released between invocations.

Plugins

No response

OpenCode version

1.17.8

Steps to reproduce

1. Start a long session with heavy tool-call throughput (e.g. quickemu macOS setup)
2. Make many consecutive bash calls over ~2 hours
3. Monitor with ps aux --sort=-%mem — RSS grows monotonically

Screenshot and/or share link

No response

Operating System

OS CachyOS (Arch Linux), kernel 7.0.12 LTO

Terminal

kitty

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• Memory Megathread #20695: Memory Megathread — the central tracking issue for memory problems; maintainers are collecting heap snapshots there to diagnose and fix memory growth issues
• OOM kill: message history accumulation causes 8.5GB memory usage during task execution #29730: OOM kill — message history accumulation causes 8.5GB memory usage during task execution (very similar symptoms: ~8.5GB peak, ~70MB db, long session with many tool calls)
• server mode: long-running opencode serve accumulates anonymous JS heap/swap; 26.8GiB cgroup peak, WKFastMalloc/JSJITCode #33213: long-running opencode accumulates heap/swap with 26GB+ cgroup peak

If your issue matches what is described in #20695, please consider contributing a heap snapshot there following the instructions in that thread — it would help the maintainers trace and fix the root cause.