Chat

Chat is the main way you interact with HiPMMAI. Conversations live inside a project, so they can draw on that project's knowledge base.

Start a conversation

There are two ways to create a chat — the difference is whether it uses a knowledge base:

• Plain chat: sidebar → Chats → ＋ New chat — not tied to a project; you just talk to the model, no knowledge-base retrieval.
• Knowledge-base chat: sidebar → Knowledge base → open a project → ＋ New chat at the top — the chat is bound to that project and auto-retrieves its knowledge base. See Knowledge base.

Then type your message and send.

Pick a model

The model selector in the chat toolbar chooses which model answers. It lists the models from your channels, plus the default model open on every account. You can switch models between turns.

Attachments

Attach a file to a single message (up to 10 MB) to ask about it inline — useful for a quick one-off document you don't want to add to the knowledge base.

For anything you'll reference repeatedly, upload it to the project's knowledge base instead.

When the knowledge base is used

In a chat inside a project, relevant passages from that project's knowledge base are retrieved and used to ground the answer automatically. You'll get answers that cite your own documents rather than generic knowledge.

If a question isn't answerable from your documents, the assistant falls back to the model's general knowledge — so you can mix grounded and open questions in the same conversation.

Multi-turn context

Conversations are multi-turn: earlier messages stay in context, so you can refine ("make it shorter", "now compare it to last quarter") without repeating yourself.

Going further

• Turn on Agent mode when you want the assistant to do things — run tools, generate files, execute code.
• Preview and download rich output via Artifacts.