forked from Selig/openclaw-skill
Selig
4c966a3ad2
6 custom skills (assign-task, dispatch-webhook, daily-briefing, task-capture, qmd-brain, tts-voice) with technical documentation. Compatible with Claude Code, OpenClaw, Codex CLI, and OpenCode.
49 lines
2.8 KiB
JSON
Executable File
49 lines
2.8 KiB
JSON
Executable File
{
|
||
"title": "Context Window & Compaction",
|
||
"content": "Every model has a **context window** (max tokens it can see). Long-running chats accumulate messages and tool results; once the window is tight, OpenClaw **compacts** older history to stay within limits.\n\n## What compaction is\n\nCompaction **summarizes older conversation** into a compact summary entry and keeps recent messages intact. The summary is stored in the session history, so future requests use:\n\n* The compaction summary\n* Recent messages after the compaction point\n\nCompaction **persists** in the session’s JSONL history.\n\nSee [Compaction config & modes](/concepts/compaction) for the `agents.defaults.compaction` settings.\n\n## Auto-compaction (default on)\n\nWhen a session nears or exceeds the model’s context window, OpenClaw triggers auto-compaction and may retry the original request using the compacted context.\n\n* `🧹 Auto-compaction complete` in verbose mode\n* `/status` showing `🧹 Compactions: <count>`\n\nBefore compaction, OpenClaw can run a **silent memory flush** turn to store\ndurable notes to disk. See [Memory](/concepts/memory) for details and config.\n\nUse `/compact` (optionally with instructions) to force a compaction pass:\n\n## Context window source\n\nContext window is model-specific. OpenClaw uses the model definition from the configured provider catalog to determine limits.\n\n## Compaction vs pruning\n\n* **Compaction**: summarises and **persists** in JSONL.\n* **Session pruning**: trims old **tool results** only, **in-memory**, per request.\n\nSee [/concepts/session-pruning](/concepts/session-pruning) for pruning details.\n\n* Use `/compact` when sessions feel stale or context is bloated.\n* Large tool outputs are already truncated; pruning can further reduce tool-result buildup.\n* If you need a fresh slate, `/new` or `/reset` starts a new session id.",
|
||
"code_samples": [
|
||
{
|
||
"code": "/compact Focus on decisions and open questions",
|
||
"language": "unknown"
|
||
}
|
||
],
|
||
"headings": [
|
||
{
|
||
"level": "h2",
|
||
"text": "What compaction is",
|
||
"id": "what-compaction-is"
|
||
},
|
||
{
|
||
"level": "h2",
|
||
"text": "Configuration",
|
||
"id": "configuration"
|
||
},
|
||
{
|
||
"level": "h2",
|
||
"text": "Auto-compaction (default on)",
|
||
"id": "auto-compaction-(default-on)"
|
||
},
|
||
{
|
||
"level": "h2",
|
||
"text": "Manual compaction",
|
||
"id": "manual-compaction"
|
||
},
|
||
{
|
||
"level": "h2",
|
||
"text": "Context window source",
|
||
"id": "context-window-source"
|
||
},
|
||
{
|
||
"level": "h2",
|
||
"text": "Compaction vs pruning",
|
||
"id": "compaction-vs-pruning"
|
||
},
|
||
{
|
||
"level": "h2",
|
||
"text": "Tips",
|
||
"id": "tips"
|
||
}
|
||
],
|
||
"url": "llms-txt#context-window-&-compaction",
|
||
"links": []
|
||
} |