Chatting with StretchGPT and Managing Conversations

This is the day-to-day guide to using StretchGPT: starting a conversation, sending messages, getting the most from multi-turn context, and keeping your history tidy. Everything here is scoped to you — your conversations are private to your account and your organization.

Starting a conversation

1. Click New chat.
2. Optionally choose an assistant family (see Understanding Assistant Families). The default is Core.
3. Type your first message and press Send.
4. StretchGPT names the conversation automatically from your first message (trimmed to about 60 characters). If you send nothing yet, it's simply titled New chat.

You can also start a conversation with a first message in one step — the conversation is created and the first answer generated together.

Sending messages and multi-turn context

StretchGPT remembers the recent turns of a conversation — up to the last 20 completed user/assistant exchanges — and feeds them to the model as context. That's why follow-ups work without repeating yourself.

1. Type your next message and send it.
2. StretchGPT reads the recent history plus your new message, retrieves grounding evidence if available, and replies.
3. Only completed messages count as history — blocked or errored turns are skipped so they don't pollute context.

Keep each message under 4,000 characters. If you paste something longer, StretchGPT asks you to shorten it.

A realistic example

You're drafting a release note in a Core conversation:

1. "Draft a short release note for a feature that adds CSV export to reports." → StretchGPT drafts it.
2. "Make it more formal and add a line about where to find the export button." → Because turn 1 is in context, StretchGPT revises the same note.
3. "Now give me a one-line version for the changelog." → It condenses what you've been building.

Three short messages, one coherent thread — that's the multi-turn workflow.

Managing your conversations

• Find them: your conversations are listed with the most recently active at the top.
• Open one: click it to see the full message history.
• Rename: give a conversation a clearer title so you can find it later.
• Archive/delete: remove a conversation from your list when you're done with it. Archiving hides it from your active list.

Steps to rename

1. Open the conversation (or use its menu in the list).
2. Choose Rename.
3. Enter a new title and confirm.

Steps to archive

1. Open the conversation's menu.
2. Choose Archive (or Delete).
3. It disappears from your active conversation list.

Handling blocked answers

If your workspace isn't fully activated, StretchGPT still saves your message and creates an assistant reply marked blocked, with a short explanation of what's needed (for example, a model provider isn't connected, or live execution isn't enabled). This keeps your history intact and tells you exactly why there's no live answer yet. Once an operator activates the runtime, new messages get live answers. See Activating Live Chat.

Tips

• One topic per conversation. Because StretchGPT carries context, keeping a thread focused produces better answers than one giant mixed thread.
• Rename important threads early. Auto-titles are handy, but a clear name makes a thread easy to find weeks later.
• Start fresh to switch families. A conversation keeps its family for life; open a new chat to change it.
• Use grounding for facts. For anything StretchGroup-specific, check the [n] citations before you rely on it.

Troubleshooting

My conversation disappeared. It was probably archived. Archived conversations are removed from the active list. Check with your workspace if you need it restored.

My message won't send — it says it's too long. Messages are capped at 4,000 characters. Trim it or split it across two messages.

The reply says it's blocked. Your workspace's live runtime isn't activated. That's an operator step — see Activating Live Chat and The Safety Envelope and Execution Gate.

Answers seem to forget earlier context. StretchGPT keeps roughly the last 20 completed turns. In a very long thread, older details may age out of context — restate the key detail, or start a focused new conversation.