workflows: spell the value back once the caller refuses a confirmation

livekit-agents/livekit/agents/beta/workflows/address.py

@@ -59,6 +59,7 @@ def __init__(
 
         assert isinstance(instructions, (str, Instructions))  # for type checking
         self._current_address = ""
+        self._spell_read_back = False
         self._require_confirmation = require_confirmation
         self._require_explicit_ask = require_explicit_ask
 
@@ -135,9 +136,16 @@ async def _update_address_impl(
         current_tools.append(confirm_tool)
         await self.update_tools(current_tools)
 
+        read_back = (
+            f"Repeat the address field by field, spelling the street name letter by "
+            f"letter: {[' '.join(street_address.replace(' ', '')), *address_fields[1:]]}"
+            if self._spell_read_back
+            else "Repeat the address back to the user."
+        )
+        self._spell_read_back = True
         return (
             f"The address has been updated to {address}\n"
-            f"Repeat the address field by field: {address_fields} if needed\n"
+            f"{read_back}\n"
             f"Prompt the user for confirmation, do not call `confirm_address` directly"
         )
 

livekit-agents/livekit/agents/beta/workflows/dob.py

@@ -91,6 +91,7 @@ def __init__(
         self._require_explicit_ask = require_explicit_ask
         self._current_dob: date | None = None
         self._current_time: time | None = None
+        self._spell_read_back = False
 
         super().__init__(
             instructions=Instructions(
@@ -195,9 +196,15 @@ async def _update_dob_impl(
             formatted_time = self._current_time.strftime("%I:%M %p")
             response += f" at {formatted_time}"
 
+        read_back = (
+            f"Repeat the date back one part at a time, the month, the day, then the year: "
+            f"{dob.strftime('%B')}, {dob.day}, {dob.year}"
+            if self._spell_read_back
+            else "Repeat the date back to the user in a natural spoken format."
+        )
+        self._spell_read_back = True
         response += (
-            "\nRepeat the date back to the user in a natural spoken format.\n"
-            "Prompt the user for confirmation, do not call `confirm_dob` directly"
+            f"\n{read_back}\nPrompt the user for confirmation, do not call `confirm_dob` directly"
         )
 
         return response

livekit-agents/livekit/agents/beta/workflows/email_address.py

@@ -60,6 +60,7 @@ def __init__(
 
         assert isinstance(instructions, (str, Instructions))  # for type checking
         self._current_email = ""
+        self._spell_read_back = False
         self._require_confirmation = require_confirmation
         self._require_explicit_ask = require_explicit_ask
 
@@ -118,9 +119,15 @@ async def _update_email_impl(self, email: str, ctx: RunContext) -> str | None:
         current_tools.append(confirm_tool)
         await self.update_tools(current_tools)
 
+        read_back = (
+            f"Repeat the email character by character: {separated_email}"
+            if self._spell_read_back
+            else "Repeat the email back to the user."
+        )
+        self._spell_read_back = True
         return (
             f"The email has been updated to {email}\n"
-            f"Repeat the email character by character: {separated_email} if needed\n"
+            f"{read_back}\n"
             f"Prompt the user for confirmation, do not call `confirm_email_address` directly"
         )
 

livekit-agents/livekit/agents/beta/workflows/name.py

@@ -95,7 +95,7 @@ def __init__(
         self._collect_first_name = first_name
         self._collect_last_name = last_name
         self._collect_middle_name = middle_name
-        self._verify_spelling = verify_spelling
+        self._spell_read_back = verify_spelling
         self._require_confirmation = require_confirmation
         self._require_explicit_ask = require_explicit_ask
 
@@ -268,17 +268,16 @@ async def _update_name_impl(
         current_tools.append(confirm_tool)
         await self.update_tools(current_tools)
 
-        if self._verify_spelling:
-            return (
-                f"The name has been updated to {full_name}\n"
-                f"Spell out the name letter by letter for verification: {full_name}\n"
-                f"Prompt the user for confirmation, do not call `confirm_name` directly"
-            )
-
+        read_back = (
+            f"Spell out the name letter by letter for verification: {' '.join(full_name.replace(' ', ''))}"
+            if self._spell_read_back
+            else "Repeat the name back to the user."
+        )
+        self._spell_read_back = True
         return (
             f"The name has been updated to {full_name}\n"
-            f"Repeat the name back to the user and prompt for confirmation, "
-            f"do not call `confirm_name` directly"
+            f"{read_back}\n"
+            f"Prompt the user for confirmation, do not call `confirm_name` directly"
         )
 
     def _build_confirm_tool(

livekit-agents/livekit/agents/beta/workflows/phone_number.py

@@ -80,6 +80,7 @@ def __init__(
         extra = extra_instructions if extra_instructions else ""
 
         self._current_phone_number = ""
+        self._spell_read_back = False
         self._require_confirmation = require_confirmation
         self._require_explicit_ask = require_explicit_ask
 
@@ -152,9 +153,15 @@ async def _update_phone_number_impl(self, phone_number: str, ctx: RunContext) ->
         current_tools.append(confirm_tool)
         await self.update_tools(current_tools)
 
+        read_back = (
+            f"Read the number back digit by digit: {' '.join(cleaned)}"
+            if self._spell_read_back
+            else "Read the number back to the user in groups."
+        )
+        self._spell_read_back = True
         return (
             f"The phone number has been updated to {cleaned}\n"
-            f"Read the number back to the user in groups.\n"
+            f"{read_back}\n"
             f"Prompt the user for confirmation, do not call `confirm_phone_number` directly"
         )
 

tests/test_workflow_readback.py

@@ -0,0 +1,122 @@
+from __future__ import annotations
+
+from types import SimpleNamespace
+from typing import Any
+
+import pytest
+
+from livekit.agents import beta
+
+pytestmark = pytest.mark.unit
+
+
+def _audio_ctx() -> Any:
+    return SimpleNamespace(
+        speech_handle=SimpleNamespace(input_details=SimpleNamespace(modality="audio"))
+    )
+
+
+@pytest.mark.asyncio
+async def test_email_is_spelled_once_a_confirmation_is_refused() -> None:
+    task = beta.workflows.GetEmailTask()
+    ctx = _audio_ctx()
+
+    first = await task._update_email_impl("shayne.cole@gmail.com", ctx)
+    second = await task._update_email_impl("shayne.cole@gmail.com", ctx)
+
+    assert first is not None and second is not None
+    assert first != second
+    assert " ".join("shayne.cole@gmail.com") in second
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize(
+    ("first_name", "last_name"),
+    [("Shayne", "Cole"), ("Anne-Marie", "O'Neill"), ("Ana", "García")],
+)
+async def test_name_is_spelled_on_subsequent_updates(first_name: str, last_name: str) -> None:
+    task = beta.workflows.GetNameTask(first_name=True, last_name=True)
+    ctx = _audio_ctx()
+
+    first = await task._update_name_impl(ctx, first_name=first_name, last_name=last_name)
+    second = await task._update_name_impl(ctx, first_name=first_name, last_name=last_name)
+    third = await task._update_name_impl(ctx, first_name=first_name, last_name=last_name)
+
+    assert first is not None and second is not None
+    assert first != second
+    spelled = " ".join(f"{first_name}{last_name}")
+    assert spelled not in first
+    assert spelled in second
+    assert "  " not in second
+    assert third == second
+    assert task._first_name == first_name
+    assert task._last_name == last_name
+
+
+@pytest.mark.asyncio
+async def test_name_with_verify_spelling_is_spelled_from_the_start() -> None:
+    task = beta.workflows.GetNameTask(first_name=True, verify_spelling=True)
+    ctx = _audio_ctx()
+
+    first = await task._update_name_impl(ctx, first_name="Shayne")
+    second = await task._update_name_impl(ctx, first_name="Shayne")
+
+    assert first == second
+    assert first is not None
+    assert "S h a y n e" in first
+
+
+@pytest.mark.asyncio
+async def test_phone_is_read_digit_by_digit_once_a_confirmation_is_refused() -> None:
+    task = beta.workflows.GetPhoneNumberTask()
+    ctx = _audio_ctx()
+
+    first = await task._update_phone_number_impl("415-555-0626", ctx)
+    second = await task._update_phone_number_impl("415-555-0626", ctx)
+
+    assert first is not None and second is not None
+    assert first != second
+    assert " ".join("4155550626") in second
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize(
+    ("street", "unit", "locality", "country"),
+    [
+        ("1 Main St", "", "Springfield", "US"),
+        ("88 Harbor Lane", "Apartment 3", "Portland Oregon 97209", "United States"),
+        ("5 Rue de l’Ancienne Comédie", "App C4", "75006 Paris", "France"),
+    ],
+)
+async def test_address_is_spelled_on_subsequent_updates(
+    street: str, unit: str, locality: str, country: str
+) -> None:
+    task = beta.workflows.GetAddressTask()
+    ctx = _audio_ctx()
+
+    first = await task._update_address_impl(street, unit, locality, country, ctx)
+    second = await task._update_address_impl(street, unit, locality, country, ctx)
+    third = await task._update_address_impl(street, unit, locality, country, ctx)
+
+    assert first is not None and second is not None
+    assert first != second
+    spelled = " ".join(street.replace(" ", ""))
+    assert spelled not in first
+    assert spelled in second
+    assert "  " not in second
+    assert third == second
+    fields = [street, unit, locality, country] if unit else [street, locality, country]
+    assert task._current_address == " ".join(fields)
+    assert str([spelled, *fields[1:]]) in second
+
+
+@pytest.mark.asyncio
+async def test_dob_is_read_part_by_part_once_a_confirmation_is_refused() -> None:
+    task = beta.workflows.GetDOBTask()
+    ctx = _audio_ctx()
+
+    first = await task._update_dob_impl(1990, 5, 17, ctx)
+    second = await task._update_dob_impl(1990, 5, 17, ctx)
+
+    assert first is not None and second is not None
+    assert first != second