Add form-fill UIWorker example

A ReplyToolMixin UIWorker that fills inputs (fills) and toggles checkboxes /
presses submit (click) by voice — the state-changing half of the standard
action set.

examples/multi-worker/ui-worker/form-fill/README.md

@@ -0,0 +1,76 @@
+# form-fill
+
+The UIWorker fills form inputs and clicks buttons by voice. The page
+renders a job application with text fields, a textarea, checkboxes, and
+a submit button. Tell the worker your name, email, and the rest; when
+you're ready, say "submit."
+
+## What it shows
+
+- The **state-changing actions**: `set_input_value` for writing into
+  inputs, `click` for checkboxes and submit. Both are bundled into the
+  same `ReplyToolMixin` that pointing and deixis use — `fills` is a list
+  of `{"ref", "value"}` so the LLM can fill several fields in one turn
+  ("my name is John Smith" fills first AND last name in one call), and
+  `click` is a list so checkboxes and submit run in order.
+- That `FormWorker` is a one-line composition:
+  `class FormWorker(ReplyToolMixin, UIWorker)`. Same shape as pointing
+  and deixis; the visual fields (`highlight`, `select_text`) just stay
+  `null` here, and the prompt steers the LLM toward `fills` / `click`.
+
+## What it adds vs. `pointing` and `deixis`
+
+Those exercise the visual / attention-pointing fields of `reply`. This
+one exercises the state-changing fields (`fills`, `click`). Same
+composition, same mixin — different fields per turn, driven by the
+prompt.
+
+## Run
+
+Two terminals.
+
+**Terminal 1 — bot:**
+
+```bash
+cd examples/multi-worker/ui-worker/form-fill
+uv run python bot.py
+```
+
+The bot starts on `http://localhost:7860`.
+
+**Terminal 2 — client:**
+
+```bash
+cd examples/multi-worker/ui-worker/form-fill/client
+npm install            # one-time
+npm run dev
+```
+
+Open `http://localhost:5173` and click **Connect**.
+
+## What to try
+
+- _"My name is John Smith."_ — fills first and last name in one call.
+- _"My email is john at gmail dot com."_ — converts the spoken form to
+  `mark@daily.co` and fills the email field.
+- _"I have five years of experience and I love working on real-time
+  voice agents."_ — fills two fields in one call.
+- _"Agree to the terms."_ — clicks the terms checkbox.
+- _"What have I entered so far?"_ — reads back current values from
+  `<ui_state>` (no fills, no clicks).
+- _"Submit it."_ — clicks submit. If terms isn't ticked yet, the worker
+  clicks both in order: terms, then submit.
+
+## Requirements
+
+- `OPENAI_API_KEY`
+- `DEEPGRAM_API_KEY`
+- `CARTESIA_API_KEY`
+
+A `.env` in the example folder is the easiest way to set these (see
+`examples/multi-worker/env.example`).
+
+## What this example _doesn't_ show
+
+Selection-based deixis (see `deixis/`) or async task cards (see
+`async-tasks/`).

examples/multi-worker/ui-worker/form-fill/bot.py

@@ -0,0 +1,278 @@
+#
+# Copyright (c) 2026, Daily
+#
+# SPDX-License-Identifier: BSD 2-Clause License
+#
+
+"""Form-fill — the UIWorker changes form state by voice.
+
+The page renders a job-application form. The user dictates field values
+("my name is John Smith", "email is john at gmail dot com"), checks
+boxes ("I agree to the terms"), and submits ("send it"). The UIWorker
+maps each spoken value to the right input ref and writes it.
+
+Same skeleton as ``pointing`` / ``deixis``. ``FormWorker`` composes
+``ReplyToolMixin``: the ``reply(answer, scroll_to, fills, click)`` bundle
+covers the state-changing actions — ``fills`` writes input values
+(many at once), ``click`` toggles checkboxes and presses submit.
+
+Architecture::
+
+    Main worker (PipelineWorker, owns transport + RTVI):
+      transport.in → STT → user_agg → LLM → TTS → transport.out → assistant_agg
+        └── answer_about_screen(query) tool
+              └── params.pipeline_worker.job("ui", name="respond", payload={query})
+
+    FormWorker (ReplyToolMixin + UIWorker):
+      └── inherited: reply(answer, scroll_to, fills, click)
+
+Run::
+
+    uv run python bot.py
+
+Then open the client at ``http://localhost:5173`` (see ``README.md``).
+
+Requirements:
+
+- OPENAI_API_KEY
+- DEEPGRAM_API_KEY
+- CARTESIA_API_KEY
+"""
+
+import os
+
+from dotenv import load_dotenv
+from loguru import logger
+
+from pipecat.adapters.schemas.tools_schema import ToolsSchema
+from pipecat.audio.vad.silero import SileroVADAnalyzer
+from pipecat.frames.frames import LLMRunFrame
+from pipecat.pipeline.job_context import JobError
+from pipecat.pipeline.pipeline import Pipeline
+from pipecat.pipeline.runner import PipelineRunner
+from pipecat.pipeline.worker import PipelineParams, PipelineWorker
+from pipecat.processors.aggregators.llm_context import LLMContext
+from pipecat.processors.aggregators.llm_response_universal import (
+    LLMContextAggregatorPair,
+    LLMUserAggregatorParams,
+)
+from pipecat.runner.types import RunnerArguments
+from pipecat.runner.utils import create_transport
+from pipecat.services.cartesia.tts import CartesiaTTSService
+from pipecat.services.deepgram.stt import DeepgramSTTService
+from pipecat.services.llm_service import FunctionCallParams
+from pipecat.services.openai.llm import OpenAILLMService
+from pipecat.transports.base_transport import BaseTransport, TransportParams
+from pipecat.transports.daily.transport import DailyParams
+from pipecat.workers.ui import ReplyToolMixin, UIWorker
+
+load_dotenv(override=True)
+
+MAIN_NAME = "main"
+
+transport_params = {
+    "daily": lambda: DailyParams(audio_in_enabled=True, audio_out_enabled=True),
+    "webrtc": lambda: TransportParams(audio_in_enabled=True, audio_out_enabled=True),
+}
+
+
+VOICE_PROMPT = """\
+You are the voice layer of a form-fill assistant. A separate UI \
+layer sees the form and writes the spoken reply.
+
+For every user utterance involving the form (filling fields, \
+checking boxes, submitting), call ``answer_about_screen`` with the \
+user's request verbatim. The tool's response is the spoken reply, \
+already TTS-ready.
+
+Only respond directly for pure pleasantries (greetings, thanks, \
+goodbyes). Keep direct replies to one short spoken sentence."""
+
+
+# The UI wire-format guide (UI_STATE_PROMPT_GUIDE) is appended to the LLM's
+# system instruction automatically by UIWorker, so this prompt only needs the
+# app-specific behavior.
+UI_PROMPT = """\
+You help the user fill out a job application form by voice. The \
+current ``<ui_state>`` block is in your context. Each input has a \
+ref (e.g. ``e5``) and a label. Use the labels to decide which input \
+gets which value.
+
+## Tool: reply
+
+Every turn calls ``reply`` exactly once. One tool call per turn.
+
+``reply(answer, scroll_to=None, fills=None, click=None)``:
+
+- ``answer`` (REQUIRED): a short spoken reply confirming what you \
+did or asking for missing info. One short sentence. Plain language.
+- ``scroll_to`` (OPTIONAL): a single snapshot ref. Use when a field \
+the user wants to see is tagged ``[offscreen]``.
+- ``fills`` (OPTIONAL): a list of ``{"ref": "eN", "value": "..."}`` \
+objects. Each entry writes ``value`` into the input at ``ref``. \
+You can fill many fields in one turn (e.g. first name + last name \
++ email when the user says "my name is John Smith, mark at \
+daily dot co").
+- ``click`` (OPTIONAL): a list of refs to click. Use for \
+checkboxes (terms, newsletter) and the submit button. Order matters: \
+click checkboxes before submit.
+
+## Decision rules
+
+- **User dictates field values** → match each value to the input \
+whose label fits, set ``fills``, confirm in ``answer``.
+- **User says "check" / "agree" / "yes" for a checkbox** → resolve \
+the matching checkbox ref, set ``click=[ref]``.
+- **User says "submit" / "send it"** → confirm any required fields \
+are filled (especially the terms checkbox if needed), then \
+``click=[submit_ref]``. If terms isn't checked yet but the user said \
+submit, click both: ``click=[terms_ref, submit_ref]``.
+- **User asks "what have I entered?" / "what's left?"** → read the \
+current values from ``<ui_state>`` (the walker emits each input's \
+current value), summarize in ``answer``. No fills, no clicks.
+
+## Spelling and disambiguation
+
+When the user says something like "john at gmail dot com", convert \
+to ``mark@daily.co``. "five five five one two three four" → \
+``5551234``. "five years" → ``5``. Don't read these conversions \
+back to the user verbatim; just confirm naturally ("got it, your \
+email is mark@daily.co").
+
+## Examples
+
+(refs are illustrative; use the actual refs from the current \
+``<ui_state>``)
+
+- "My name is John Smith." → \
+``reply(answer="Got it, John Smith.", fills=[{"ref":"e5","value":"Mark"}, {"ref":"e7","value":"Backman"}])``
+- "Email is john at gmail dot com." → \
+``reply(answer="Email saved.", fills=[{"ref":"e9","value":"mark@daily.co"}])``
+- "I have five years of experience and I love working on \
+real-time voice agents." → \
+``reply(answer="Five years and your interest noted.", fills=[{"ref":"e15","value":"5"}, {"ref":"e17","value":"I love working on real-time voice agents."}])``
+- "I agree to the terms." → \
+``reply(answer="Terms accepted.", click=["e22"])``
+- "Submit it." (terms not yet checked) → \
+``reply(answer="Submitting.", click=["e22","e26"])``
+- "What have I entered?" → \
+``reply(answer="John Smith, mark@daily.co, 5 years experience. The cover letter and terms aren't done yet.")``"""
+
+
+class FormWorker(ReplyToolMixin, UIWorker):
+    """UIWorker that fills form fields and toggles controls via ``reply``.
+
+    Composes ``ReplyToolMixin``, which exposes a single
+    ``reply(answer, scroll_to=None, fills=None, click=None, ...)`` LLM
+    tool. ``fills`` writes values into inputs (many in one turn) and
+    ``click`` toggles checkboxes / presses submit — the state-changing
+    half of the standard action set.
+    """
+
+    def __init__(self):
+        llm = OpenAILLMService(
+            api_key=os.environ["OPENAI_API_KEY"],
+            settings=OpenAILLMService.Settings(system_instruction=UI_PROMPT),
+        )
+        super().__init__("ui", llm=llm)
+
+
+async def answer_about_screen(params: FunctionCallParams, query: str):
+    """Ask the screen-aware UI worker to fill the form / answer about it.
+
+    Args:
+        query (str): The user's request, passed verbatim.
+    """
+    logger.info(f"answer_about_screen('{query}')")
+    try:
+        async with params.pipeline_worker.job(
+            "ui", name="respond", payload={"query": query}, timeout=10
+        ) as t:
+            pass
+    except JobError as e:
+        logger.warning(f"ui job failed: {e}")
+        await params.result_callback("Something went wrong on my side.")
+        return
+
+    speak = (t.response or {}).get("speak")
+    await params.result_callback(speak or "I'm not sure how to answer that.")
+
+
+async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
+    logger.info("Starting form-fill bot")
+
+    runner = PipelineRunner(handle_sigint=runner_args.handle_sigint)
+
+    stt = DeepgramSTTService(api_key=os.environ["DEEPGRAM_API_KEY"])
+    tts = CartesiaTTSService(
+        api_key=os.environ["CARTESIA_API_KEY"],
+        settings=CartesiaTTSService.Settings(
+            voice=os.getenv("CARTESIA_VOICE_ID", "71a7ad14-091c-4e8e-a314-022ece01c121"),
+        ),
+    )
+    llm = OpenAILLMService(
+        api_key=os.environ["OPENAI_API_KEY"],
+        settings=OpenAILLMService.Settings(system_instruction=VOICE_PROMPT),
+    )
+    llm.register_direct_function(answer_about_screen, cancel_on_interruption=False, timeout_secs=30)
+
+    context = LLMContext(tools=ToolsSchema(standard_tools=[answer_about_screen]))
+    aggregators = LLMContextAggregatorPair(
+        context,
+        user_params=LLMUserAggregatorParams(vad_analyzer=SileroVADAnalyzer()),
+    )
+
+    pipeline = Pipeline(
+        [
+            transport.input(),
+            stt,
+            aggregators.user(),
+            llm,
+            tts,
+            transport.output(),
+            aggregators.assistant(),
+        ]
+    )
+
+    worker = PipelineWorker(
+        pipeline,
+        name=MAIN_NAME,
+        params=PipelineParams(enable_metrics=True, enable_usage_metrics=True),
+        idle_timeout_secs=runner_args.pipeline_idle_timeout_secs,
+    )
+
+    @transport.event_handler("on_client_connected")
+    async def on_client_connected(transport, client):
+        logger.info("Client connected")
+        context.add_message(
+            {
+                "role": "developer",
+                "content": (
+                    "Greet the user briefly. Tell them they can dictate field "
+                    "values and you'll fill them in. One short sentence."
+                ),
+            }
+        )
+        await worker.queue_frame(LLMRunFrame())
+
+    @transport.event_handler("on_client_disconnected")
+    async def on_client_disconnected(transport, client):
+        logger.info("Client disconnected")
+        await runner.cancel()
+
+    await runner.launch_worker(FormWorker())
+    await runner.launch_worker(worker)
+
+    await runner.run()
+
+
+async def bot(runner_args: RunnerArguments):
+    """Main bot entry point compatible with Pipecat Cloud."""
+    transport = await create_transport(runner_args, transport_params)
+    await run_bot(transport, runner_args)
+
+
+if __name__ == "__main__":
+    from pipecat.runner.run import main
+
+    main()

examples/multi-worker/ui-worker/form-fill/client/index.html

@@ -0,0 +1,88 @@
+<!doctype html>
+<html lang="en">
+  <head>
+    <meta charset="UTF-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <title>Form fill — UIAgent demo</title>
+    <link rel="stylesheet" href="./styles.css" />
+  </head>
+  <body>
+    <header>
+      <h1>Application form</h1>
+      <button id="connect" type="button">Connect</button>
+    </header>
+
+    <main>
+      <form id="application-form" aria-label="Job application">
+        <h2>Apply for: Software Engineer</h2>
+        <p class="hint">
+          Tell the assistant what to put in any field. Try: "my name
+          is John Smith, my email is john at gmail dot com". When
+          you're ready, say "submit".
+        </p>
+
+        <fieldset>
+          <legend>Your details</legend>
+
+          <div class="field">
+            <label for="first-name">First name</label>
+            <input id="first-name" name="first_name" type="text" autocomplete="given-name" />
+          </div>
+
+          <div class="field">
+            <label for="last-name">Last name</label>
+            <input id="last-name" name="last_name" type="text" autocomplete="family-name" />
+          </div>
+
+          <div class="field">
+            <label for="email">Email address</label>
+            <input id="email" name="email" type="email" autocomplete="email" />
+          </div>
+
+          <div class="field">
+            <label for="phone">Phone (optional)</label>
+            <input id="phone" name="phone" type="tel" autocomplete="tel" />
+          </div>
+        </fieldset>
+
+        <fieldset>
+          <legend>About you</legend>
+
+          <div class="field">
+            <label for="experience">Years of relevant experience</label>
+            <input id="experience" name="experience" type="text" inputmode="numeric" />
+          </div>
+
+          <div class="field">
+            <label for="cover">Why are you interested?</label>
+            <textarea id="cover" name="cover" rows="5"></textarea>
+          </div>
+        </fieldset>
+
+        <fieldset>
+          <legend>Confirm</legend>
+
+          <div class="field checkbox-field">
+            <input id="terms" name="terms" type="checkbox" />
+            <label for="terms">I agree to the terms of service.</label>
+          </div>
+
+          <div class="field checkbox-field">
+            <input id="newsletter" name="newsletter" type="checkbox" />
+            <label for="newsletter">Send me product updates.</label>
+          </div>
+        </fieldset>
+
+        <div class="actions">
+          <button id="submit" type="submit">Submit application</button>
+          <span id="form-status" aria-live="polite"></span>
+        </div>
+      </form>
+    </main>
+
+    <div id="status" aria-live="polite"></div>
+    <audio id="bot-audio" autoplay data-a11y-exclude></audio>
+
+    <script type="module" src="./main.js"></script>
+  </body>
+</html>

examples/multi-worker/ui-worker/form-fill/client/main.js

@@ -0,0 +1,202 @@
+/**
+ * Form fill — vanilla JS client.
+ *
+ * Same base wiring as pointing/deixis (PipecatClient
+ * + managed snapshot streaming + bot audio sink). Three command
+ * handlers: ``scroll_to``, ``set_input_value``, and ``click``.
+ *
+ * ``set_input_value`` writes a string into an ``<input>`` /
+ * ``<textarea>``. Crucially it dispatches ``input`` and ``change``
+ * events so React-controlled or other frameworks pick up the change
+ * naturally; the React standard handler does the same. ``click``
+ * is the catch-all for checkboxes, radios, and submit buttons.
+ *
+ * The submit button intercepts the form's submit event so the demo
+ * stays on-page after the agent submits. Real apps would let it
+ * through.
+ */
+
+import {
+  PipecatClient,
+  RTVIEvent,
+  findElementByRef,
+} from "@pipecat-ai/client-js";
+import { SmallWebRTCTransport } from "@pipecat-ai/small-webrtc-transport";
+
+const BOT_URL = "http://localhost:7860/api/offer";
+
+const connectButton = document.getElementById("connect");
+const status = document.getElementById("status");
+const botAudio = document.getElementById("bot-audio");
+const form = document.getElementById("application-form");
+const formStatus = document.getElementById("form-status");
+
+let client;
+let unsubscribes = [];
+
+function setStatus(text, autoHideMs = 0) {
+  status.textContent = text;
+  status.dataset.show = text ? "1" : "0";
+  if (text && autoHideMs > 0) {
+    setTimeout(() => {
+      if (status.textContent === text) status.dataset.show = "0";
+    }, autoHideMs);
+  }
+}
+
+function resolveTarget(payload) {
+  if (payload?.ref) {
+    const el = findElementByRef(payload.ref);
+    if (el) return el;
+  }
+  if (payload?.target_id) {
+    return document.getElementById(payload.target_id);
+  }
+  return null;
+}
+
+function handleScrollTo(payload) {
+  const el = resolveTarget(payload);
+  if (!el) return;
+  const behavior =
+    payload?.behavior === "instant" || payload?.behavior === "smooth"
+      ? payload.behavior
+      : "smooth";
+  el.scrollIntoView({ behavior, block: "center", inline: "nearest" });
+}
+
+/**
+ * Write ``payload.value`` into the targeted input/textarea.
+ *
+ * Skips ``disabled`` / ``readonly`` / ``type="hidden"`` targets so
+ * the agent can't bypass UI affordances. Dispatches ``input`` and
+ * ``change`` events so framework-controlled inputs (React, Vue, etc.)
+ * notice the change. Briefly flashes the field so the user sees
+ * what the agent wrote.
+ */
+function handleSetInputValue(payload) {
+  const el = resolveTarget(payload);
+  if (!el) return;
+  if (!(el instanceof HTMLInputElement || el instanceof HTMLTextAreaElement))
+    return;
+  if (el.disabled || el.readOnly) return;
+  if (el.type === "hidden") return;
+
+  const value = String(payload?.value ?? "");
+  const replace = payload?.replace !== false;
+  el.value = replace ? value : (el.value || "") + value;
+
+  el.dispatchEvent(new Event("input", { bubbles: true }));
+  el.dispatchEvent(new Event("change", { bubbles: true }));
+
+  // Visual confirmation: a brief background flash so the user
+  // notices the write happened.
+  el.classList.remove("fill-flash");
+  void el.offsetWidth;
+  el.classList.add("fill-flash");
+  setTimeout(() => el.classList.remove("fill-flash"), 1200);
+}
+
+/**
+ * Click the targeted element. Skips ``disabled`` targets so the
+ * agent can't bypass disabled affordances; the standard React
+ * handler does the same.
+ */
+function handleClick(payload) {
+  const el = resolveTarget(payload);
+  if (!el) return;
+  if ("disabled" in el && el.disabled) return;
+  el.click();
+}
+
+// Don't actually submit the form on the demo; the agent says "I
+// submitted it" and we show a status message instead.
+form.addEventListener("submit", (e) => {
+  e.preventDefault();
+  formStatus.textContent = "Submitted (demo only — no network call).";
+  formStatus.style.color = "#16a34a";
+});
+
+function onUICommand(command, handler) {
+  const listener = (data) => {
+    if (data.command !== command) return;
+    handler(data.payload);
+  };
+  client.on(RTVIEvent.UICommand, listener);
+  return () => client.off(RTVIEvent.UICommand, listener);
+}
+
+async function connect() {
+  connectButton.disabled = true;
+  setStatus("Connecting…");
+
+  client = new PipecatClient({
+    transport: new SmallWebRTCTransport(),
+    enableMic: true,
+    enableCam: false,
+  });
+
+  client.on(RTVIEvent.BotConnected, () => setStatus("Bot connected", 1500));
+  client.on(RTVIEvent.Disconnected, () => {
+    setStatus("Disconnected", 2000);
+    connectButton.dataset.state = "";
+    connectButton.textContent = "Connect";
+    connectButton.disabled = false;
+    teardownUI();
+  });
+
+  client.on(RTVIEvent.TrackStarted, (track, participant) => {
+    if (track.kind !== "audio") return;
+    if (participant?.local) return;
+    botAudio.srcObject = new MediaStream([track]);
+  });
+
+  unsubscribes = [
+    onUICommand("scroll_to", handleScrollTo),
+    onUICommand("set_input_value", handleSetInputValue),
+    onUICommand("click", handleClick),
+  ];
+
+  try {
+    await client.connect({ webrtcUrl: BOT_URL });
+    client.startUISnapshotStream();
+    connectButton.dataset.state = "connected";
+    connectButton.textContent = "Disconnect";
+    connectButton.disabled = false;
+    setStatus("Connected. Try: 'My name is John Smith'", 5000);
+  } catch (err) {
+    console.error("Connect failed:", err);
+    setStatus(`Connect failed: ${err.message ?? err}`, 4000);
+    teardownUI();
+    connectButton.disabled = false;
+  }
+}
+
+async function disconnect() {
+  connectButton.disabled = true;
+  setStatus("Disconnecting…");
+  try {
+    await client?.disconnect();
+  } finally {
+    teardownUI();
+    connectButton.dataset.state = "";
+    connectButton.textContent = "Connect";
+    connectButton.disabled = false;
+  }
+}
+
+function teardownUI() {
+  client?.stopUISnapshotStream();
+  unsubscribes.forEach((unsubscribe) => unsubscribe());
+  unsubscribes = [];
+  if (botAudio.srcObject) botAudio.srcObject = null;
+  client = undefined;
+}
+
+connectButton.addEventListener("click", () => {
+  if (connectButton.dataset.state === "connected") {
+    disconnect();
+  } else {
+    connect();
+  }
+});

examples/multi-worker/ui-worker/form-fill/client/package.json

@@ -0,0 +1,18 @@
+{
+  "name": "form-fill-client",
+  "private": true,
+  "version": "0.1.0",
+  "type": "module",
+  "scripts": {
+    "dev": "vite",
+    "build": "vite build",
+    "preview": "vite preview"
+  },
+  "dependencies": {
+    "@pipecat-ai/client-js": "1.9.0",
+    "@pipecat-ai/small-webrtc-transport": "^1.10.2"
+  },
+  "devDependencies": {
+    "vite": "^8"
+  }
+}

examples/multi-worker/ui-worker/form-fill/client/styles.css

@@ -0,0 +1,207 @@
+:root {
+  color-scheme: light;
+  font-family: system-ui, -apple-system, sans-serif;
+  --border: #d4d4d8;
+  --border-focus: #3b82f6;
+  --muted: #71717a;
+  --bg: #fafafa;
+}
+
+* {
+  box-sizing: border-box;
+}
+
+body {
+  margin: 0;
+  background: var(--bg);
+  color: #18181b;
+}
+
+header {
+  position: sticky;
+  top: 0;
+  z-index: 10;
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  padding: 1rem 1.5rem;
+  border-bottom: 1px solid var(--border);
+  background: #fff;
+}
+
+header h1 {
+  font-size: 1.125rem;
+  margin: 0;
+}
+
+#connect {
+  padding: 0.5rem 1rem;
+  border: 1px solid var(--border);
+  background: #fff;
+  border-radius: 6px;
+  cursor: pointer;
+  font-size: 0.875rem;
+}
+
+#connect:hover {
+  background: #f4f4f5;
+}
+
+#connect[data-state="connected"] {
+  background: #ef4444;
+  color: white;
+  border-color: #ef4444;
+}
+
+main {
+  max-width: 640px;
+  margin: 0 auto;
+  padding: 2rem 1.5rem 4rem;
+}
+
+form h2 {
+  margin: 0 0 0.5rem;
+  font-size: 1.5rem;
+}
+
+.hint {
+  margin: 0 0 1.5rem;
+  font-size: 0.9375rem;
+  line-height: 1.5;
+  color: var(--muted);
+}
+
+fieldset {
+  margin: 0 0 1.5rem;
+  padding: 1rem 1.25rem 1.25rem;
+  border: 1px solid var(--border);
+  border-radius: 8px;
+  background: #fff;
+}
+
+legend {
+  padding: 0 0.5rem;
+  font-size: 0.8125rem;
+  font-weight: 600;
+  text-transform: uppercase;
+  letter-spacing: 0.04em;
+  color: var(--muted);
+}
+
+.field {
+  display: flex;
+  flex-direction: column;
+  gap: 0.25rem;
+  margin-top: 0.875rem;
+}
+
+.field:first-child {
+  margin-top: 0;
+}
+
+.field label {
+  font-size: 0.875rem;
+  font-weight: 500;
+  color: #27272a;
+}
+
+.field input[type="text"],
+.field input[type="email"],
+.field input[type="tel"],
+.field textarea {
+  font: inherit;
+  font-size: 1rem;
+  padding: 0.5rem 0.625rem;
+  border: 1px solid var(--border);
+  border-radius: 6px;
+  background: #fff;
+  width: 100%;
+  scroll-margin-top: 5rem;
+  transition:
+    border-color 0.15s,
+    box-shadow 0.15s,
+    background 0.4s;
+}
+
+.field input:focus,
+.field textarea:focus {
+  outline: none;
+  border-color: var(--border-focus);
+  box-shadow: 0 0 0 3px rgba(59, 130, 246, 0.2);
+}
+
+/* Brief flash when the agent fills a field, so the user notices the
+   write. The CSS class is added by the set_input_value handler and
+   removed after the animation. */
+@keyframes field-fill-flash {
+  0% {
+    background: #fef3c7;
+  }
+  100% {
+    background: #fff;
+  }
+}
+
+.field input.fill-flash,
+.field textarea.fill-flash {
+  animation: field-fill-flash 1.2s ease-out;
+}
+
+.field textarea {
+  resize: vertical;
+}
+
+.checkbox-field {
+  flex-direction: row;
+  align-items: center;
+  gap: 0.5rem;
+}
+
+.checkbox-field label {
+  font-weight: 400;
+}
+
+.actions {
+  display: flex;
+  gap: 1rem;
+  align-items: center;
+}
+
+#submit {
+  font: inherit;
+  font-size: 0.9375rem;
+  font-weight: 500;
+  padding: 0.625rem 1.25rem;
+  background: #18181b;
+  color: white;
+  border: 1px solid #18181b;
+  border-radius: 6px;
+  cursor: pointer;
+}
+
+#submit:hover {
+  background: #27272a;
+}
+
+#form-status {
+  font-size: 0.875rem;
+  color: var(--muted);
+}
+
+#status {
+  position: fixed;
+  bottom: 1rem;
+  right: 1rem;
+  padding: 0.5rem 0.75rem;
+  border-radius: 6px;
+  font-size: 0.8125rem;
+  background: #18181b;
+  color: white;
+  opacity: 0;
+  transition: opacity 0.2s;
+  pointer-events: none;
+}
+
+#status[data-show="1"] {
+  opacity: 1;
+}

examples/multi-worker/ui-worker/form-fill/client/vite.config.js

@@ -0,0 +1,7 @@
+import { defineConfig } from "vite";
+
+export default defineConfig({
+  server: {
+    port: 5173,
+  },
+});