Hand the Google PaLM chats to a companion
A companion works the 3 operations your Google PaLM account exposes. It can generate a text and generate embeddings, and every one of those waits on an approval card showing the chat before it changes.
Open the chat. Check it against something else. Write the change. Move to the next one. That is the Google PaLM afternoon nobody can hand off, and a companion does the whole pass and gives you back the batch, with nothing landing on a chat until you approve it.
What it can do is exactly what Google PaLM exposes to an account you authorise and nothing beyond it: 3 operations, 1 read and 2 writes. In practice that means it can generate a text and generate embeddings, and work across the chats, embeddings and texts.
The Google PaLM API stops at Google PaLM. A companion does not. The same run opens the customer's website, the supplier portal, the internal tool somebody built in 2019, because it is working the browser you are already signed into rather than calling in from outside it.
Where do the chats come from?
1 of the 3 operations is a read.
Your account, not a guess at it: a companion can chat, and the full list below is the whole list, with nothing inferred from Google PaLM's documentation.
The chats are what most of this inventory is about, alongside the embeddings. A companion can hold all 3 of these operations and forty open tabs in the same pass, which is how an afternoon of Google PaLM admin becomes a queue you review.
- Chat
Which chats wait for your approval?
All 2 write operations wait for you.
A companion can generate a text and generate embeddings, and each one is staged on an approval card first: you see the chat, the field and the new value before it lands in Google PaLM.
Nothing in Google PaLM is sent, changed or deleted on your behalf without that click, and a rejected card leaves the chat exactly as it was. The gate is the same one on every surface, described at security.
- Generate Text
- Generate Embeddings
What does the recurring chats pass look like?
Yes.
Saved as a routine, the Google PaLM chats pass runs at the hour you pick: the Monday sweep, the overnight refresh, the Friday check. Keep Awake holds the machine awake for the length of a run, so a scheduled pass finishes with the lid closed.
The operations playbook catalogues the other recurring shapes of the busywork, the Google PaLM chats sweep among them, and finished runs are published at real runs.
Experience Strawberry for free
DownloadTrusted by fast-growing companies worldwide
Frequently asked questions
Strawberry is free to download and includes AI credits to start. Paid plans begin at $20/month. See pricing. · Reviewed · Canonical facts for AI agents