fix: clear tool output and attachments when pruning to prevent memory leak

[hendem]

Fixes #3013

Summary

Clear tool output and attachments when compaction prunes old tool results, actually freeing memory instead of just flagging.

Evidence

In a real session working on this repo:

Metric	Value
OpenCode process RAM	16.4 GB
Tool parts on disk	34,887 files
Session duration	~3 hours

Sample tool output sizes from this session:

• webfetch of docs page: 10 KB
• read of source file: 5-50 KB
• git diff / gh pr diff: 10-100 KB
• bash command outputs: 1-50 KB

All of these outputs stay in memory even after compaction marks them "old".

The Problem

In SessionCompaction.prune(), when tool outputs are pruned:

// BEFORE: Only sets a flag, output data stays in memory
part.state.time.compacted = Date.now()
await Session.updatePart(part)

The compacted timestamp is used by toModelMessage() to replace output with placeholder text like "(Old tool result content cleared)" - but the actual data never gets freed.

Over a long session:

• 34,887 tool calls × average output size = hundreds of MB to GBs retained
• Memory never decreases, even after compaction runs
• Eventually Bun runs out of memory and crashes

The Fix

// AFTER: Actually clear the data
part.state.time.compacted = Date.now()
part.state.output = ""              // Free the output string
part.state.attachments = undefined  // Free any attachments
await Session.updatePart(part)

Now when compaction runs, the memory is actually freed. The placeholder text is already shown to the LLM (that logic exists), we just were not clearing the source data.

Testing

Existing compaction tests pass. Added test verifying output/attachments are cleared after prune.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

No duplicate PRs found

[Copilot]

Pull request overview

This PR fixes a memory leak where compacted tool outputs and their attachments were retained indefinitely in storage. When the pruning mechanism marks old tool outputs as compacted, the fix now clears both the output string and attachments array to free memory.

Key changes:

• Clears tool output and attachments during compaction to prevent unbounded memory growth
• Adds comprehensive test coverage for the pruning behavior

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
packages/opencode/src/session/compaction.ts	Sets output to empty string and attachments to undefined when marking tool parts as compacted, with explanatory comment
packages/opencode/test/session/compaction.test.ts	Adds two new test cases: one verifying output/attachments are cleared during pruning, and one testing that pruning respects the disabled configuration

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[hendem]

Addressed review comment about test not verifying behavior when pruning is disabled (commit 488173440).

The test now:

1. Creates a tool part with large output (200,000 chars) and an attachment (similar to the first test)
2. Creates additional user messages to get past turn protection
3. Calls prune() with pruning disabled via config
4. Verifies that:
   • output.length remains 200,000 (unchanged)
   • attachments.length remains 1 (unchanged)
   • time.compacted is undefined (not set)

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[kevoconnell]

this would be really helpful

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[Th0rgal]

Server Mode Diagnostic Data

Running OpenCode v1.1.23 as a systemd daemon (opencode serve) on a 64GB Linux server. Here is data that may help with this PR.

Memory Growth (from systemd logs)

Runtime	Memory Peak	Swap
2.5h	263 MB	0
4.5h	699 MB	0
4.75h	56 GB	8.8 GB
2.25h	20.9 GB	0

Memory grows over time and eventually causes the process to become unresponsive.

Storage Accumulation

storage/part/     141MB (10,820 directories)
storage/message/   19MB (728 directories)
tool-output/       21MB (15 files)

The 10,820 part directories correlate with the unfree-d tool outputs described in this PR.

Timeout Pattern

When memory gets high, HTTP requests to /session/{id}/message stop responding. Downstream clients see 10-minute timeouts:

17:22:29 WARN Retry attempt=2 session=ses_xxx
17:32:30 WARN Retry attempt=3 (10 min gap = timeout)
17:42:31 ERROR elapsed_secs=1801 Failed after retries

Environment

• OpenCode v1.1.23, server mode (opencode serve --port 4096)
• Ubuntu Linux, systemd service
• Multiple concurrent sessions over extended periods
• Heavy tool usage (grep, read, bash)

Happy to provide more data if useful.

[kryptobaseddev]

can the PR be reviewed along with this one, it was fixing similar issues from a different persepctive: #14650