Personal assistant

How it would work

A message arrives

The user calls, chats, or sends a WhatsApp message.

Unpod knows who it is

The channel handle resolves to one person, with every past conversation.

The agent picks up from there

It answers with that context, and acts through your APIs.

The outcome is remembered

The next conversation, on any channel, starts informed.

The shape

Illustrative - not callable today

person = unpod.entities.upsert(
    identities=[{"channel": "whatsapp", "value": "+91981..."}],
)
conversation = unpod.conversations.latest(person_id=person.id)
unpod.messages.send(
    conversation_id=conversation.id,
    channel="whatsapp",
    text="Booked: a table for two at 8 tonight.",
)

Piece	Status	Where
WhatsApp and SMS conversations (dashboard flows only; no SDK resource yet)	In build	Channels
Conversations API	In build	Conversations
Cross-channel identity	In build	People & identity

Piece

Status

Where

WhatsApp and SMS conversations (dashboard flows only; no SDK resource yet)

In build

Channels

Conversations API

In build

Conversations

Cross-channel identity

In build

People & identity

Build it today

A voice-and-chat version runs now, with two separate memories:

Voice: a voice agent with Enable Memory on (Agent Studio → Advanced) recalls past calls from the same number.

Chat: the Chat API keeps a conversation stateful when you pass the same user id - use your own user’s id.

curl -X POST "https://inference.unpod.ai/v1/chat/completions" \
  -H "Authorization: Bearer $UNPOD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"public:PB_...","messages":[{"role":"user","content":"Book my usual table"}],"user":"user_42"}'

The two memories are not joined yet. Joining them is what cross-channel identity in the Context Engine is for.

Apps

Personal assistant

How it would work

The shape

What it waits on

Build it today

​How it would work

​The shape

​What it waits on

​Build it today

How it would work

The shape

What it waits on

Build it today