| name | liteforms-web |
| description | Set up, configure, and launch the Liteforms Web avatar chat app with OpenClaw Gateway as its LLM brain. |
| metadata | {"openclaw":{"requires":{"bins":["git","node","npm","openclaw"]}}} |
Liteforms Web
Set up Liteforms Web locally and connect it to the user's local OpenClaw Gateway.
What Liteforms Needs
- Repository:
https://github.com/Looking-Glass/liteforms-web.git
- App directory:
liteforms-web
- Development command:
npm run dev
- Local app URL: usually
http://localhost:3000
- OpenClaw Gateway OpenAI-compatible base URL:
http://127.0.0.1:18789/v1
- Liteforms LLM provider id:
openclaw
- Recommended Liteforms model value:
openclaw/default
- Browser session config key:
liteforms.sessionConfig
Liteforms already includes OpenClaw Gateway as an LLM provider. Do not patch Liteforms source unless the installed repo no longer exposes the OpenClaw provider.
Workflow
-
Verify prerequisites:
git --version
node --version
npm --version
openclaw --version
Require Node.js 20 or newer for Liteforms Web.
-
Clone or update the repo:
git clone https://github.com/Looking-Glass/liteforms-web.git
cd liteforms-web
If liteforms-web already exists, enter it and run:
git pull --ff-only
-
Install Liteforms dependencies.
If package-lock.json is present, use a clean, deterministic install:
npm ci
Use npm install only when you intentionally need to update the lockfile.
-
Configure OpenClaw Gateway for Liteforms:
openclaw config set gateway.http.endpoints.chatCompletions.enabled true --strict-json
Start or restart the Gateway after changing this setting:
openclaw gateway
If the Gateway refuses to start on a fresh install, run the user's normal OpenClaw setup/onboarding flow first, or use openclaw gateway --allow-unconfigured only for ad-hoc local bootstrap.
-
Enable fast mode for the OpenClaw session Liteforms will use.
Prefer a session-level setting by sending this through an OpenClaw chat/TUI/web session that uses the same agent/model Liteforms should talk to:
/fast on
If the user wants the default OpenAI/OpenAI-Codex model to always use fast mode, configure the model default instead. Preserve existing agents.defaults.models entries when editing config. Example shape:
{
"agents": {
"defaults": {
"models": {
"openai/gpt-5.5": {
"params": {
"fastMode": true
}
}
}
}
}
}
-
Get the Gateway token only when Liteforms needs it.
Liteforms shows an "OpenClaw Gateway token" field during onboarding. The value is gateway.auth.token or OPENCLAW_GATEWAY_TOKEN.
openclaw config get gateway.auth.token
Treat this value as a local secret: do not paste it into public logs, issue trackers, or summaries. If automating a browser, write it directly into Liteforms local storage or the form field without echoing it back to the user.
-
Start Liteforms:
npm run dev
Read the URL printed by Next.js. It is usually http://localhost:3000, but Next.js may choose another port if 3000 is busy.
-
Open the app in a browser tab.
If the browser starts on the onboarding screen, choose:
- Model provider:
OpenClaw Gateway
- Model:
openclaw/default
- Base URL:
http://127.0.0.1:18789/v1
- OpenClaw Gateway token: the local Gateway token, if token auth is enabled
- Voice provider:
Kokoro local, unless the user wants another TTS provider
- Speech input provider:
Distil-Whisper, unless the user wants another STT provider
Browser Automation Shortcut
When browser automation is available and the user wants the setup prefilled, set Liteforms local storage before interacting with the app. Replace GATEWAY_TOKEN with the token only in the browser execution context; do not print it.
localStorage.setItem(
"liteforms.sessionConfig",
JSON.stringify({
version: 1,
llm: {
provider: "openclaw",
model: "openclaw/default",
baseUrl: "http://127.0.0.1:18789/v1",
endpointMode: "openai-compatible",
credential: "GATEWAY_TOKEN"
},
tts: { provider: "kokoro" },
asr: { provider: "distil-whisper" }
})
);
If Gateway auth is disabled, omit the credential field.
Verification
If a build failure appears, first reproduce it from a clean tree:
git clean -xfd && npm ci && npm run build
If the failure still reproduces there, treat it as a real repo issue and patch or pin from that exact error.
After launch:
-
Confirm the Gateway is reachable:
openclaw gateway health
-
In Liteforms, confirm Settings shows:
- Model provider:
OpenClaw Gateway
- Model:
openclaw/default
-
Send a short chat message. If Liteforms reports a 404 for /v1/chat/completions, rerun the chat-completions config command and restart the Gateway. If it reports 401, re-enter the Gateway token.
Safety Notes
- Do not commit
.env, .env.local, browser storage dumps, or Gateway tokens.
- Do not expose the Gateway beyond loopback unless the user deliberately configured Gateway auth and bind settings.
- Keep Liteforms provider credentials browser-local unless the user explicitly asks for server-side deployment work.