Configure Compaction

You can configure compaction settings in opencode.jsonc:

{
  "$schema": "https://opencode.ai/config.json",
  "compaction": {
    "auto": true,
    "keep": {
      "tokens": 10000,
    },
    "buffer": 20000,
  },
}

• auto: whether to compact automatically near context limits. It defaults to true, and usually you want it to stay that way.
• keep.tokens: the approximate number of tokens of recent context that OpenCode tries to preserve (in addition to the summary).
• buffer: the safety margin, in tokens, below the model's limit; increasing it can trigger compaction earlier.