Skip to content

fix(openclaw): batch config set calls into single exec #3319

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
AhmedTMM merged 3 commits into from
Apr 18, 2026
Merged

fix(openclaw): batch config set calls into single exec #3319

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
e33d40f
fix(openclaw): batch config set calls into single exec
AhmedTMM Apr 17, 2026
87dd1fc
Merge branch 'main' into fix/openclaw-batch-config
AhmedTMM Apr 18, 2026
17ed357
Merge branch 'main' into fix/openclaw-batch-config
AhmedTMM Apr 18, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion packages/cli/package.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
{
"name": "@openrouter/spawn",
"version": "1.0.15",
"version": "1.0.16",
"type": "module",
"bin": {
"spawn": "cli.js"
Expand Down
80 changes: 36 additions & 44 deletions packages/cli/src/shared/agent-setup.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -478,45 +478,31 @@ async function setupOpenclawConfig(
await uploadConfigFile(runner, fallbackConfig, "$HOME/.openclaw/openclaw.json");
}

// Always set the model after onboard — `openclaw onboard` may pick its own
// default (e.g. arcee/trinity-large-thinking) instead of openrouter/auto.
const modelResult = await asyncTryCatchIf(isOperationalError, () =>
runner.runServer(
"export PATH=$HOME/.npm-global/bin:$HOME/.bun/bin:$HOME/.local/bin:$PATH; " +
`openclaw config set agents.defaults.model.primary ${shellQuote(modelId)} >/dev/null`,
),
);
if (!modelResult.ok) {
logWarn("Model config failed (non-fatal)");
}

// Disable Docker sandboxing — when Docker is installed on the VM, openclaw
// auto-detects it and runs agents inside containers, which hangs the session.
await asyncTryCatchIf(isOperationalError, () =>
runner.runServer(
"export PATH=$HOME/.npm-global/bin:$HOME/.bun/bin:$HOME/.local/bin:$PATH; " +
"openclaw config set agents.defaults.sandbox.mode off >/dev/null",
),
);

// Configure browser via CLI (openclaw config set) — the supported way to set
// browser options. Redirect stdout to suppress doctor warnings on each call.
const browserResult = await asyncTryCatchIf(isOperationalError, () =>
runner.runServer(
"export PATH=$HOME/.npm-global/bin:$HOME/.bun/bin:$HOME/.local/bin:$PATH; " +
"openclaw config set browser.executablePath /usr/bin/google-chrome-stable >/dev/null; " +
"openclaw config set browser.noSandbox true >/dev/null; " +
"openclaw config set browser.headless true >/dev/null; " +
"openclaw config set browser.defaultProfile openclaw >/dev/null",
),
);
if (!browserResult.ok) {
logWarn("Browser config setup failed (non-fatal)");
}

// Write channel stubs so the dashboard renders channel cards properly,
// even when the user hasn't configured them yet. Without stubs the
// dashboard shows "Unsupported type: . Use Raw mode."
// Batch all `openclaw config set` calls into ONE exec to reduce Sprite
// connection overhead. Previously 4 separate exec calls, each triggering a
// "Config overwrite" log line from OpenClaw. On Sprite (container-exec, not
// persistent SSH), many sequential execs exhaust the connection and cause
// "connection closed" / "context deadline exceeded" on later steps.
//
// Each individual config set is chained with `;` (not `&&`) so a failure
// in one doesn't skip the rest — these are all non-fatal preferences.
const configCmds = [
// Model — openclaw onboard writes arcee/trinity-large-thinking to the
// agent-specific config (agents.main.model.primary) which overrides
// the defaults path. Set BOTH so our model always wins.
`openclaw config set agents.defaults.model.primary ${shellQuote(modelId)} >/dev/null`,
`openclaw config set agents.main.model.primary ${shellQuote(modelId)} >/dev/null`,
// Disable Docker sandboxing — auto-detected Docker hangs the session
"openclaw config set agents.defaults.sandbox.mode off >/dev/null",
"openclaw config set agents.main.sandbox.mode off >/dev/null",
// Browser (requires Chrome installed above)
"openclaw config set browser.executablePath /usr/bin/google-chrome-stable >/dev/null",
"openclaw config set browser.noSandbox true >/dev/null",
"openclaw config set browser.headless true >/dev/null",
"openclaw config set browser.defaultProfile openclaw >/dev/null",
];

// Channel stubs so the dashboard renders channel cards
const channelNames = [
"telegram",
"whatsapp",
Expand All @@ -526,11 +512,17 @@ async function setupOpenclawConfig(
"googlechat",
"bluebubbles",
].filter((ch) => !enabledSteps || enabledSteps.has(ch));
if (channelNames.length > 0) {
const stubCmds = channelNames.map((ch) => `openclaw config set channels.${ch}.enabled true >/dev/null`).join("; ");
await asyncTryCatchIf(isOperationalError, () =>
runner.runServer("export PATH=$HOME/.npm-global/bin:$HOME/.bun/bin:$HOME/.local/bin:$PATH; " + stubCmds),
);
for (const ch of channelNames) {
configCmds.push(`openclaw config set channels.${ch}.enabled true >/dev/null`);
}

const batchResult = await asyncTryCatchIf(isOperationalError, () =>
runner.runServer(
"export PATH=$HOME/.npm-global/bin:$HOME/.bun/bin:$HOME/.local/bin:$PATH; " + configCmds.join("; "),
),
);
if (!batchResult.ok) {
logWarn("Some config settings may have failed (non-fatal)");
}

// Configure Telegram channel if a bot token was provided.
Expand Down
Loading