puq Chat

puq Chat is in Alpha. Behavior and limits described here may change.

puq Chat is the puq.ai Assistant — a chat interface that can read, create, and edit your workflows, look up available pieces and connections, and run steps or whole workflows on your behalf, all from a conversation.

Opening Chat

You can use the assistant in two places:

  • Chat page — click Chat in the sidebar to open it full-page at /chat.
  • Workflow builder drawer — while editing a workflow, click the round teal chat button in the bottom-right corner of the canvas (Open puq.ai Chat). The assistant opens as a side panel next to the canvas so you can keep watching the workflow while you talk. Closing the drawer (the X in its header) returns it to page mode.

When the assistant is opened from inside a workflow, it works on that workflow’s draft; messages sent from the drawer are tied to the open workflow.

Starting and Managing Conversations

Each conversation is a separate thread.

  • New Chat — the plus/message icon in the header, or Start New Chat at the top of the history panel, clears the view and starts a new thread the moment you send your first message.
  • Chat History — the clock icon opens a list of your threads, grouped into Today, Yesterday, Previous 7 Days, Previous 30 Days, and Older, newest first. The list is paginated; use the arrows at the bottom to move between pages.
  • Thread titles are generated automatically from your first message when a thread is created — there is no manual rename.
  • Delete a conversation — hover a thread in the history list and click the trash icon, then confirm in the Delete Conversation dialog. Deletion cannot be undone.
  • Reasoning toggle — the eye icon in the header shows or hides the assistant’s intermediate “thinking” steps in the current view.
  • Usage — the gauge icon opens a dialog showing your daily output token usage (see Limits below).

What the Assistant Can Do

Inside a conversation, the assistant can:

  • Read and discuss your workflows — list your workflows, open one, and read its current steps, trigger, and connections.
  • Build and edit workflows — create a new draft workflow, set the trigger, and append, replace, or remove steps (including code, branch, and loop steps) based on what you ask for.
  • Look up building blocks — list available pieces (integrations) and read details about a specific piece before using it.
  • Run things for you — execute a single step or the whole workflow, and report back what happened.
  • Publish — publish the workflow once it’s ready.
  • Request a connection — if a step needs a connection you don’t have yet, the assistant pauses and asks you to pick or create one (see Connection Requests below) instead of guessing or failing silently.

When the assistant creates or changes a workflow, the change is saved to the workflow’s draft automatically, and if you’re chatting from the full-page view about a specific workflow, opening it jumps you straight into the builder with that draft loaded.

Reading a Response

While the assistant is working, the conversation shows:

  • Your messages and the assistant’s replies.
  • Thinking steps — a collapsible trace of its reasoning, shown only when the reasoning toggle is on.
  • Tool activity cards — one card per action the assistant takes (for example, reading a workflow, creating a step, or running it), each showing its inputs, result, and — for workflow edits — a diff of what changed.
  • A Stop button (replacing Send) while a reply is streaming, to cancel generation.

Connection Requests

When a step needs a connection you haven’t set up, an Input required card appears in the conversation with an Approval required badge, asking you to choose or confirm a connection. The request expires after 15 minutes — after that it can no longer be resumed, and you’ll need to ask again. You can also cancel the request from the same card.

Limits

puq Chat enforces a daily output token allowance per account. There is currently no way to select a different model for chat — every conversation uses the platform’s single configured assistant model.

  • Your usage is tracked against a daily output token limit that resets at midnight. Check the Usage Status dialog (gauge icon) for your current used/remaining tokens and reset time.
  • A warning banner appears above the message box once you’ve used 80% of your daily allowance.
  • When the daily limit is reached, sending a new message or resuming a paused request returns a CHAT_CREDIT_LIMIT error and is blocked until the allowance resets.
  • One active run per thread — a thread can only process one message at a time. Sending another message while the assistant is still replying, or while a connection request is pending, is rejected until the current run finishes.
  • Threads and messages are scoped to your account; you cannot see another user’s conversations.

Best Practices

  • Open the chat drawer from inside a workflow when you want the assistant to edit that workflow specifically.
  • Resolve or cancel a pending connection request promptly — it expires after 15 minutes.
  • Turn on the reasoning toggle when you want to understand why the assistant made a change, and off to keep the conversation compact.
  • Watch the usage gauge during heavy chat sessions so you’re not interrupted mid-task by the daily limit.