Description

When OpenViking is configured as the contextEngine in OpenClaw (replacing the default lossless-claw), the /compact command becomes unavailable with the error:

Compaction skipped: legacy_compact_unavailable

Environment

• OpenClaw version: 2026.3.22
• OpenViking version: 0.2.9
• OpenViking plugin version: 2026.2.6-3

Configuration

"plugins": {
  "slots": {
    "memory": "none",
    "contextEngine": "openviking"
  }
}

Steps to Reproduce

1. Configure OpenViking as contextEngine in OpenClaw
2. Have a long conversation that approaches context window limit
3. Execute /compact command
4. Observe error: legacy_compact_unavailable

Expected Behavior

Either:

1. /compact should work with OpenViking as contextEngine, OR
2. Documentation should clearly state that OpenViking contextEngine replaces the compression functionality and /compact is not supported

Actual Behavior

User has no way to compress context when using OpenViking as contextEngine, leading to context overflow errors when the context window is exceeded.

Impact

When context exceeds the model's context window (e.g., GLM-5's 202,752 tokens), users get:

HTTP 400: InternalError.Algo.InvalidParameter: Range of input length should be [1, 202752]

The only workaround is /reset which loses all conversation context.

Suggested Solutions

1. Implement context compaction support in OpenViking contextEngine
2. Or provide clear documentation about this limitation and recommend using OpenViking only as memory slot, not contextEngine
3. Or implement automatic context pruning when approaching context window limits

Related

This was discovered during upgrade from lossless-claw to OpenViking contextEngine. The trade-off between conversation compression (LCM) and semantic memory injection (OpenViking) was not clearly documented.