mirror of
https://github.com/openclaw/openclaw.git
synced 2026-05-06 07:50:43 +00:00
docs(experimental): explain local model lean mode for operators
This commit is contained in:
@@ -29,13 +29,53 @@ Treat them differently from normal config:
|
||||
|
||||
## Local model lean mode
|
||||
|
||||
`agents.defaults.experimental.localModelLean: true` is a pressure-release valve
|
||||
for weaker local-model setups. It trims heavyweight default tools like
|
||||
`browser`, `cron`, and `message` so the prompt shape is smaller and less brittle
|
||||
for small-context or stricter OpenAI-compatible backends.
|
||||
`agents.defaults.experimental.localModelLean: true` is a pressure-release valve for weaker local-model setups. When it is on, OpenClaw drops three default tools — `browser`, `cron`, and `message` — from the agent's tool surface for every turn. Nothing else changes.
|
||||
|
||||
That is intentionally **not** the normal path. If your backend handles the full
|
||||
runtime cleanly, leave this off.
|
||||
### Why these three tools
|
||||
|
||||
These three tools have the largest descriptions and the most parameter shapes in the default OpenClaw runtime. On a small-context or stricter OpenAI-compatible backend that is the difference between:
|
||||
|
||||
- Tool schemas fitting cleanly in the prompt vs. crowding out conversation history.
|
||||
- The model picking the right tool vs. emitting malformed tool calls because there are too many similar-looking schemas.
|
||||
- The Chat Completions adapter staying inside the server's structured-output limits vs. tripping a 400 on tool-call payload size.
|
||||
|
||||
Removing them does not silently rewire OpenClaw — it just makes the tool list shorter. The model still has `read`, `write`, `edit`, `exec`, `apply_patch`, web search/fetch (when configured), memory, and session/agent tools available.
|
||||
|
||||
### When to turn it on
|
||||
|
||||
Enable lean mode when you have already proved the model can talk to the Gateway but full agent turns misbehave. The typical signal chain is:
|
||||
|
||||
1. `openclaw infer model run --gateway --model <ref> --prompt "Reply with exactly: pong"` succeeds.
|
||||
2. A normal agent turn fails with malformed tool calls, oversized prompts, or the model ignoring its tools.
|
||||
3. Toggling `localModelLean: true` clears the failure.
|
||||
|
||||
### When to leave it off
|
||||
|
||||
If your backend handles the full default runtime cleanly, leave this off. Lean mode is a workaround, not a default. It exists because some local stacks need a smaller tool surface to behave; hosted models and well-resourced local rigs do not.
|
||||
|
||||
Lean mode also does not replace `tools.profile`, `tools.allow`/`tools.deny`, or the model `compat.supportsTools: false` escape hatch. If you need a permanent narrower tool surface for a specific agent, prefer those stable knobs over the experimental flag.
|
||||
|
||||
### Enable
|
||||
|
||||
```json5
|
||||
{
|
||||
agents: {
|
||||
defaults: {
|
||||
experimental: {
|
||||
localModelLean: true,
|
||||
},
|
||||
},
|
||||
},
|
||||
}
|
||||
```
|
||||
|
||||
Restart the Gateway after changing the flag, then confirm the trimmed tool list with:
|
||||
|
||||
```bash
|
||||
openclaw status --deep
|
||||
```
|
||||
|
||||
The deep status output lists the active agent tools; `browser`, `cron`, and `message` should be absent when lean mode is on.
|
||||
|
||||
## Experimental does not mean hidden
|
||||
|
||||
|
||||
Reference in New Issue
Block a user